Subscribe
Share
Share
Embed
The Question is a collaborative learning podcast about Design Systems. Smart people like you sign up, answer a few niche questions about design systems for each episode, and then we all get together to unpack the data we've gathered. Each week, I'll invite a new co-host to help facilitate the conversation. After the deep dive, the co-host and I record a recap of what we learned. That means, for each episode, you can listen to the recap and the full deep dive!
If you're a design system practitioner, subscribe today (https://bencallahan.com/the-question) to receive an invitation to each episode. This only works if the community joins in!
Stay in learning mode ❤️
Ben Callahan (00:00)
course, I wholeheartedly agree. there should you we should not be using AI to write the
079 deep dive screen raw (00:00)
course, I wholeheartedly agree. there should you we should not be using AI to write the
079 deep dive speaker raw (00:00)
course, I wholeheartedly agree. there should you we should not be using AI to write the
Ben Callahan (00:07)
Hello, system thinkers. Welcome to the Question, a live collaborative learning show committed to broadening your perspective on design systems. I'm your host, Ben Callahan, founder of Sparkbox and Redwood's Design System Community, and you're listening to the episode 079 Deep Dive with Afyia Smith, where we're discussing the evolution of design system documentation. If you have a moment, please get subscribed to the show out at bencallahan.com/thequestion
Let's get into it.
Ben Callahan (00:40)
Afyia tell us a little bit about your background. I want to hear, because you're such a specialist in the space. And I don't know. And it's funny because we're going to look at the data here in a minute. And one of the questions we asked is who owns design system docs? And almost nobody has somebody who's fully dedicated to this role. So yeah, this is this is unique. So tell us a little bit about
079 deep dive screen raw (00:40)
Afyia, tell us a little bit about your background. I want to hear, because you're such a specialist in the space. And I don't know. And it's funny because we're going to look at the data here in a minute. And one of the questions we asked is who owns design system docs? And almost nobody has somebody who's fully dedicated to this role. So yeah, this is this is unique. So tell us a little bit about
079 deep dive speaker raw (00:40)
Afyia, tell us a little bit about your background. I want to hear, because you're such a specialist in the space. And I don't know. And it's funny because we're going to look at the data here in a minute. And one of the questions we asked is who owns design system docs? And almost nobody has somebody who's fully dedicated to this role. So yeah, this is this is unique. So tell us a little bit about
Ben Callahan (01:00)
How you kind of got into this deep specialization? Yeah. I've been in the design system space for maybe seven years now. And I started out as a UXUI designer over at EA Electronic Arts, was part of a really small design system team. We wore many hats, one of those being writing docs while building out the library. And I really fell in love with writing the docs. It gave me, because I was such a new designer, it was a way to understand components.
079 deep dive screen raw (01:00)
How you kind of got into this deep specialization? Yeah. I've been in the design system space for maybe seven years now. And I started out as a UXUI designer over at EA Electronic Arts, was part of a really small design system team. We wore many hats, one of those being writing docs while building out the library. And I really fell in love with writing the docs. It gave me, because I was such a new designer, it was a way to understand components.
079 deep dive speaker raw (01:00)
How you kind of got into this deep specialization? Yeah. I've been in the design system space for maybe seven years now. And I started out as a UXUI designer over at EA Electronic Arts, was part of a really small design system team. We wore many hats, one of those being writing docs while building out the library. And I really fell in love with writing the docs. It gave me, because I was such a new designer, it was a way to understand components.
Ben Callahan (01:30)
and I also got a seat at the table with the engineering team because they had implementation details that needed to be kind of added. And I just really, really enjoyed that process and quickly kind of dropped designing. I didn't it wasn't my passion, and moved full speed ahead into writing documentation. shortly after I left EA, I started to freelance with a few of the design system agencies that are out there.
079 deep dive screen raw (01:30)
and I also got a seat at the table with the engineering team because they had implementation details that needed to be kind of added. And I just really, really enjoyed that process and quickly kind of dropped designing. I didn't it wasn't my passion, and moved full speed ahead into writing documentation. shortly after I left EA, I started to freelance with a few of the design system agencies that are out there.
079 deep dive speaker raw (01:30)
and I also got a seat at the table with the engineering team because they had implementation details that needed to be kind of added. And I just really, really enjoyed that process and quickly kind of dropped designing. I didn't it wasn't my passion, and moved full speed ahead into writing documentation. shortly after I left EA, I started to freelance with a few of the design system agencies that are out there.
Ben Callahan (01:57)
Mostly on design systems, but specifically on documentation, for design system teams of various sizes and design systems of various phases and stages. So I'd be brought in to maybe help give a documentation workshop. So that's helping and teaching them how to write. Or they don't want anything to do with it. They just shove 50 components at me, leave me alone for three months, and I'm spinning up documentation. Also helping with like
079 deep dive screen raw (01:57)
Mostly on design systems, but specifically on documentation, for design system teams of various sizes and design systems of various phases and stages. So I'd be brought in to maybe help r give a documentation workshop. So that's helping and teaching them how to write. Or they don't want anything to do with it. They just shove 50 components at me, leave me alone for three months, and I'm spinning up documentation. Also helping with like
079 deep dive speaker raw (01:57)
Mostly on design systems, but specifically on documentation, for design system teams of various sizes and design systems of various phases and stages. So I'd be brought in to maybe help r give a documentation workshop. So that's helping and teaching them how to write. Or they don't want anything to do with it. They just shove 50 components at me, leave me alone for three months, and I'm spinning up documentation. Also helping with like
Ben Callahan (02:25)
You know, tool platform analysis, like, you know, what's what's better for our team, custom site, zeroheight, notion: things of that sort. Then governance came into play in contribution models. And as of late, AI readiness, which is what I was doing most recently at Epic Games. I was there for about five years, and I managed to be a documentation specialist on the design system team, which is
079 deep dive screen raw (02:25)
You know, tool platform analysis, like, you know, what's what's better for our team, custom site, zero height notion, things of that sort. Then governance came into play in contribution models. And as of late, AI readiness, which is what I was doing most recently at Epic Games. I was there for about five years, and I managed to be a documentation specialist on the design system team, which is
079 deep dive speaker raw (02:25)
You know, tool platform analysis, like, you know, what's what's better for our team, custom site, zero height notion, things of that sort. Then governance came into play in contribution models. And as of late, AI readiness, which is what I was doing most recently at Epic Games. I was there for about five years, and I managed to be a documentation specialist on the design system team, which is
Ben Callahan (02:54)
A really rare position. I really haven't even, I don't know anyone else that that's been able to do that. And I did that for the entire length of time that I was there. And grew, I did it for two of their design systems, Fortnite and Epic Design System. I helped build out the custom reference site. I learned Docs's code workflow because that's how we were maintaining the documentation, writing it and publishing it and reviewing it.
079 deep dive screen raw (02:54)
A really rare position. I really haven't even, I don't know anyone else that that's been able to do that. And I did that for the entire length of time that I was there. And grew, I did it for two of their design systems, Fortnite and Epic Design System. I helped build out the custom reference site. I learned Docs's code workflow because that's how we were maintaining the documentation, writing it and publishing it and reviewing it.
079 deep dive speaker raw (02:54)
A really rare position. I really haven't even, I don't know anyone else that that's been able to do that. And I did that for the entire length of time that I was there. And grew, I did it for two of their design systems, Fortnite and Epic Design System. I helped build out the custom reference site. I learned Docs's code workflow because that's how we were maintaining the documentation, writing it and publishing it and reviewing it.
Ben Callahan (03:21)
I built governance models and then I built an AI readiness framework at the point that the team started to experiment with AI tools. from there I started just playing around with component schemas because I thought that that is something that we need, a standard. How should we document these sets of components? and then I bumped into, or I actually I think Ben, you introduced me to PJ Onori and I saw what he was doing with.
079 deep dive screen raw (03:21)
I built governance models and then I built an AI readiness framework at the point that the team started to experiment with AI tools. from there I started just playing around with component schemas because I thought that that is something that we need, a standard. How should we document these sets of components? and then I bumped into, or I actually I think Ben, you introduced me to PJ Onori and I saw what he was doing with.
079 deep dive speaker raw (03:21)
I built governance models and then I built an AI readiness framework at the point that the team started to experiment with AI tools. from there I started just playing around with component schemas because I thought that that is something that we need, a standard. How should we document these sets of components? and then I bumped into, or I actually I think Ben, you introduced me to PJ Anori, and I saw what he was doing with.
Ben Callahan (03:50)
DSDS and I raised my hand and I emailed him. And I know he thought it was crazy because it was like three in the morning and it was I was high on coffee. So it was like a 10 paragraph email asking how I can contribute because you know he wanted GitHub and you know, that's just not my area of expertise. So I just emailed him and we've been emailing each other back and forth over the past couple of months or so, and I was able to collaborate on
079 deep dive screen raw (03:50)
DSDS and I raised my hand and I emailed him. And I know he thought it was crazy because it was like three in the morning and it was I was high on coffee. So it was like a 10 paragraph email asking how I can contribute because you know he wanted GitHub and you know, that's just not my area of expertise. So I just emailed him and we've been emailing each other back and forth over the past couple of months or so, and I was able to collaborate on
079 deep dive speaker raw (03:50)
DSDS and I raised my hand and I emailed him. And I know he thought it was crazy because it was like three in the morning and it was I was high on coffee. So it was like a 10 paragraph email asking how I can contribute because you know he wanted GitHub and you know, that's just not my area of expertise. So I just emailed him and we've been emailing each other back and forth over the past couple of months or so, and I was able to collaborate on
Ben Callahan (04:17)
doc origin and and just a a a lot of other really fun things with him. yeah. And so that's been my journey through documentation. It is a passion. It's what I want to do and continue to do. I want to grow it. I think now that AI is in play, there there should be much more importance played on documentation. I think it's far more
079 deep dive screen raw (04:17)
doc origin and and just a a a lot of other really fun things with him. yeah. And so that's been my journey through documentation. It is a passion. It's what I want to do and continue to do. I want to grow it. I think now that AI is in play, there there should be much more importance played on documentation. I think it's far more
079 deep dive speaker raw (04:17)
doc origin and and just a a a lot of other really fun things with him. yeah. And so that's been my journey through documentation. It is a passion. It's what I want to do and continue to do. I want to grow it. I think now that AI is in play, there there should be much more importance played on documentation. I think it's far more
Ben Callahan (04:41)
it's it has always been foundational. It just hasn't been giving the the rightful attention that it needs. And now's the time. Yeah. Afyia One of the things that really stands out about this role is the you said this it when you were describing it at the beginning. You were like, of course, you you came from design. So you had probably had a strong relationship or a natural affin affiliation with the designers, but it also gave you the opportunity to interact really closely with engineering.
079 deep dive screen raw (04:41)
it's it has always been foundational. It just hasn't been giving the the rightful attention that it needs. And now's the time. Yeah. Effia. One of the things that really stands out about this role is the you said this it when you were describing it at the beginning. You were like, of course, you you came from design. So you had probably had a strong relationship or a natural affin affiliation with the designers, but it also gave you the opportunity to interact really closely with engineering.
079 deep dive speaker raw (04:41)
it's it has always been foundational. It just hasn't been giving the the rightful attention that it needs. And now's the time. Yeah. Effia. One of the things that really stands out about this role is the you said this it when you were describing it at the beginning. You were like, of course, you you came from design. So you had probably had a strong relationship or a natural affin affiliation with the designers, but it also gave you the opportunity to interact really closely with engineering.
Ben Callahan (05:10)
And I could see that role as being almost like a bridge between these roles, you know, occasionally, because you you have to sort of bring the the things that are needed for both, you know. So that's really that's really cool. I hadn't thought of it quite like that, but that's awesome. you also mentioned PJ and some of the work that he's gotten started. There is a new design system documentation community group over in the W3C. So I dropped a link in the resource bar. You'll see that if you look at my shared screen. And if you're interested in that, you can join that group.
079 deep dive screen raw (05:10)
And I could see that role as being almost like a bridge between these roles, you know, occasionally, because you you have to sort of bring the the things that are needed for both, you know. So that's really that's really cool. I hadn't thought of it quite like that, but that's awesome. you also mentioned PJ and some of the work that he's gotten started. There is a new design system documentation community group over in the W3C. So I dropped a link in the resource bar. You'll see that if you look at my shared screen. And if you're interested in that, you can join that group.
079 deep dive speaker raw (05:10)
And I could see that role as being almost like a bridge between these roles, you know, occasionally, because you you have to sort of bring the the things that are needed for both, you know. So that's really that's really cool. I hadn't thought of it quite like that, but that's awesome. you also mentioned PJ and some of the work that he's gotten started. There is a new design system documentation community group over in the W3C. So I dropped a link in the resource bar. You'll see that if you look at my shared screen. And if you're interested in that, you can join that group.
Ben Callahan (05:39)
it, you know, of course, we want to get folks who care about design systems in that into that space. so jump in if that's a thing you're interested in doing. It's a slow process, of course, as you probably know if you've fought along with other standards as they get formed, but it's pretty cool to see that happening.
079 deep dive screen raw (05:39)
it, you know, of course, we want to get folks who care about design systems in that into that space. so jump in if that's a thing you're interested in doing. It's a slow process, of course, as you probably know if you've fought along with other standards as they get formed, but it's pretty cool to see that happening.
079 deep dive speaker raw (05:39)
it, you know, of course, we want to get folks who care about design systems in that into that space. so jump in if that's a thing you're interested in doing. It's a slow process, of course, as you probably know if you've fought along with other standards as they get formed, but it's pretty cool to see that happening.
Ben Callahan (05:55)
What else? I was gonna ask you something else, but let's jump in. Let's jump in. no, I wanted to say this. I know Afyia well enough now to know that she would love to connect with you. So if you're not connected on LinkedIn, please check out her website, check out her LinkedIn. Those are those links are both right here down in the bio section. And I'm the same way. you can you can reach out to me over here. Please stay connected. I it's it's really important to me. This is like a my favorite part of my day is when I get.
079 deep dive screen raw (05:55)
What else? I was gonna ask you something else, but let's jump in. Let's jump in. no, I wanted to say this. I know Afyia well enough now to know that she would love to connect with you. So if you're not connected on LinkedIn, please check out her website, check out her LinkedIn. Those are those links are both right here down in the bio section. And I'm the same way. you can you can reach out to me over here. Please stay connected. I it's it's really important to me. This is like a my favorite part of my day is when I get.
079 deep dive speaker raw (05:55)
What else? I was gonna ask you something else, but let's jump in. Let's jump in. no, I wanted to say this. I know Afyia well enough now to know that she would love to connect with you. So if you're not connected on LinkedIn, please check out her website, check out her LinkedIn. Those are those links are both right here down in the bio section. And I'm the same way. you can you can reach out to me over here. Please stay connected. I it's it's really important to me. This is like a my favorite part of my day is when I get.
Ben Callahan (06:21)
introduced to new people who are systems thinkers or out in the space doing interesting stuff. That's what I want to do. And so please take a minute to get connected out there. we sent this week an extra question. I think we had five questions this week, which is a little bit more than we normally send, but we just had a lot of we were just very curious. So let's just level set again and I'll remind you kind of what we asked you. The first question was who owns design system docs for your org? And you had to choose from one of a several of several options.
079 deep dive screen raw (06:21)
introduced to new people who are systems thinkers or out in the space doing interesting stuff. That's what I want to do. And so please take a minute to get connected out there. we sent this week an extra question. I think we had five questions this week, which is a little bit more than we normally send, but we just had a lot of we were just very curious. So let's just level set again and I'll remind you kind of what we asked you. The first question was who owns design system docs for your org? And you had to choose from one of a several of several options.
079 deep dive speaker raw (06:21)
introduced to new people who are systems thinkers or out in the space doing interesting stuff. That's what I want to do. And so please take a minute to get connected out there. we sent this week an extra question. I think we had five questions this week, which is a little bit more than we normally send, but we just had a lot of we were just very curious. So let's just level set again and I'll remind you kind of what we asked you. The first question was who owns design system docs for your org? And you had to choose from one of a several of several options.
Ben Callahan (06:50)
The second one, who owns product building agentic workflows for your org? and you'll see that there were several options for that as well. the third, how would you describe the success of product building agentic workflows at your org? So not only who owns those things, but how successful do you feel they are? and then we gave you kind of a a a few options to choose from. And then the last two, in the last 12 months, what have you changed about your documentation? And you could choose all that applied in that scenario.
079 deep dive screen raw (06:50)
The second one, who owns product building agentic workflows for your org? and you'll see that there were several options for that as well. the third, how would you describe the success of product building agenc workflows at your org? So not only who owns those things, but how successful do you feel they are? and then we gave you kind of a a a few options to choose from. And then the last two, in the last 12 months, what have you changed about your documentation? And you could choose all that applied in that scenario.
079 deep dive speaker raw (06:50)
The second one, who owns product building agentic workflows for your org? and you'll see that there were several options for that as well. the third, how would you describe the success of product building agenc workflows at your org? So not only who owns those things, but how successful do you feel they are? and then we gave you kind of a a a few options to choose from. And then the last two, in the last 12 months, what have you changed about your documentation? And you could choose all that applied in that scenario.
Ben Callahan (07:20)
And then the final one: what does the future of design system docs look like for you or your organization? And that was just open text, so you could just tell us kind of what you were thinking. And so lots of questions this week. we sent this out to over 1100 design system practitioners. We got 61 responses. As always, you can click this link to see the raw data yourself and do your own analysis. And then just a reminder: none of the data that we collect is statistically significant. We've got a very small
079 deep dive screen raw (07:20)
And then the final one: what does the future of design system docs look like for you or your organization? And that was just open text, so you could just tell us kind of what you were thinking. And so lots of questions this week. we sent this out to over 1100 design system practitioners. We got 61 responses. As always, you can click this link to see the raw data yourself and do your own analysis. And then just a reminder: none of the data that we collect is statistically significant. We've got a very small
079 deep dive speaker raw (07:20)
And then the final one: what does the future of design system docs look like for you or your organization? And that was just open text, so you could just tell us kind of what you were thinking. And so lots of questions this week. we sent this out to over 1100 design system practitioners. We got 61 responses. As always, you can click this link to see the raw data yourself and do your own analysis. And then just a reminder: none of the data that we collect is statistically significant. We've got a very small
Ben Callahan (07:49)
Subset of very opinionated folks who who join us here. So we're really just trying to collect, I think of it as like kindling for conversation is kind of what we're trying to gather here. So we can't draw super strong conclusions from this, but it helps us facilitate a conversation like this. the first question was really interesting. Afyia I don't know if you what what your expect expectation was given your back background, but like we were asking about who owns this. What did you think we would see in these in this data?
079 deep dive screen raw (07:49)
Subset of very opinionated folks who who join us here. So we're really just trying to collect, I think of it as like kindling for conversation is kind of what we're trying to gather here. So we can't draw super strong conclusions from this, but it helps us facilitate a conversation like this. the first question was really interesting. Afyia, I don't know if you what what your expect expectation was given your back background, but like we were asking about who owns this. What did you think we would see in these in this data?
079 deep dive speaker raw (07:49)
Subset of very opinionated folks who who join us here. So we're really just trying to collect, I think of it as like kindling for conversation is kind of what we're trying to gather here. So we can't draw super strong conclusions from this, but it helps us facilitate a conversation like this. the first question was really interesting. Afyia, I don't know if you what what your expect expectation was given your back background, but like we were asking about who owns this. What did you think we would see in these in this data?
Ben Callahan (08:18)
Yeah, well, I went in expecting no clear owner. That was just like my thesis, right? It's in the teams that I've worked with, it's hardly ever someone that owns this, owns documentation. So happily surprised to see it's close for, you know, between everyone contributes no clear owner and design system lead or manager and other. Yeah. for the design system lead or manager, I wonder if that's by default or design. Yeah.
079 deep dive screen raw (08:18)
Yeah, well, I went in expecting no clear owner. That was just like my thesis, right? It's in the teams that I've worked with, it's hardly ever someone that owns this, owns documentation. So happily surprised to see it's close for, you know, between everyone contributes no clear owner and design system lead or manager and other. Yeah. for the design system lead or manager, I wonder if that's by default or design. Yeah.
079 deep dive speaker raw (08:18)
Yeah, well, I went in expecting no clear owner. That was just like my thesis, right? It's in the teams that I've worked with, it's hardly ever someone that owns this, owns documentation. So happily surprised to see it's close for, you know, between everyone contributes no clear owner and design system lead or manager and other. Yeah. for the design system lead or manager, I wonder if that's by default or design. Yeah.
Ben Callahan (08:48)
Right. I wonder if that's just yeah, absorbed into, you know, the things that I already do, or if I'm actually just the focus owner for this thing. Yeah. I that's a that's actually a really good, really good point because I'm guessing people know it needs to be done and who else is gonna do it? Probably the lead. Yeah. I we were talking about this beforehand too, Afyia as folks were getting signed in, that I always I'm always like, I want to try to ask better questions when I see this many others in it, you know, people choosing other.
079 deep dive screen raw (08:48)
Right. I wonder if that's just yeah, absorbed into, you know, the things that I already do, or if I'm actually just the focus owner for this thing. Yeah. I that's a that's actually a really good, really good point because I'm guessing people know it needs to be done and who else is gonna do it? Probably the lead. Yeah. I we were talking about this beforehand too, Afyia, as folks were getting signed in, that I always I'm always like, I want to try to ask better questions when I see this many others in it, you know, people choosing other.
079 deep dive speaker raw (08:48)
Right. I wonder if that's just yeah, absorbed into, you know, the things that I already do, or if I'm actually just the focus owner for this thing. Yeah. I that's a that's actually a really good, really good point because I'm guessing people know it needs to be done and who else is gonna do it? Probably the lead. Yeah. I we were talking about this beforehand too, Afyia, as folks were getting signed in, that I always I'm always like, I want to try to ask better questions when I see this many others in it, you know, people choosing other.
Ben Callahan (09:18)
And so I went I dug in and kind of broke down the responses to other. And sixty percent of those kind of fall into a category that I would say it's everyone contributes, but we actually have real accountability. Like that's kind of by design, you know? And so that one might serve as like a counter to everyone contributes with no clear ownership. Instead, that's saying, no, the system team owns the docs, but we don't have an individual, it's our whole team, and we're and that's what we have done intentionally. So that's kind of a cool answer.
079 deep dive screen raw (09:18)
And so I went I dug in and kind of broke down the responses to other. And sixty percent of those kind of fall into a category that I would say it's everyone contributes, but we actually have real accountability. Like that's kind of by design, you know? And so that one might serve as like a counter to everyone contributes with no clear ownership. Instead, that's saying, no, the system team owns the docs, but we don't have an individual, it's our whole team, and we're and that's what we have done intentionally. So that's kind of a cool answer.
079 deep dive speaker raw (09:18)
And so I went I dug in and kind of broke down the responses to other. And sixty percent of those kind of fall into a category that I would say it's everyone contributes, but we actually have real accountability. Like that's kind of by design, you know? And so that one might serve as like a counter to everyone contributes with no clear ownership. Instead, that's saying, no, the system team owns the docs, but we don't have an individual, it's our whole team, and we're and that's what we have done intentionally. So that's kind of a cool answer.
Ben Callahan (09:47)
And then 30% of those answers were something around like ownership being split. In other words, it's not a single person, but it's more like somebody owns the design side of docs, somebody owns the engineering side of docs, you know, maybe somebody owns the process. So that was I thought I I've actually seen that a number of times inside of system teams. So that of course we should have just thought to put that one in as an option. So yeah. the next one was you know, the
079 deep dive screen raw (09:47)
And then 30% of those answers were something around like ownership being split. In other words, it's not a single person, but it's more like somebody owns the design side of docs, somebody owns the engineering side of docs, you know, maybe somebody owns the process. So that was I thought I I've actually seen that a number of times inside of system teams. So that of course we should have just thought to put that one in as an option. So yeah. the next one was you know, the
079 deep dive speaker raw (09:47)
And then 30% of those answers were something around like ownership being split. In other words, it's not a single person, but it's more like somebody owns the design side of docs, somebody owns the engineering side of docs, you know, maybe somebody owns the process. So that was I thought I I've actually seen that a number of times inside of system teams. So that of course we should have just thought to put that one in as an option. So yeah. the next one was you know, the
Ben Callahan (10:15)
Ownership around the agentic workflows. And so this one was a little bit more evenly distributed. but the over the overarching winner here with almost 50% was everybody's just kind of figuring this out on their own. that's that is kind of what I was thinking. Did you have thoughts on this? Or were you hoping to see, you know, more about like system folks owning this? Or what was your thinking?
079 deep dive screen raw (10:15)
Ownership around the agentic workflows. And so this one was a little bit more evenly distributed. but the over the overarching winner here with almost 50% was everybody's just kind of figuring this out on their own. that's that is kind of what I was thinking. Did you have thoughts on this? Or were you hoping to see, you know, more about like system folks owning this? Or what was your thinking?
079 deep dive speaker raw (10:15)
Ownership around the agentic workflows. And so this one was a little bit more evenly distributed. but the over the overarching winner here with almost 50% was everybody's just kind of figuring this out on their own. that's that is kind of what I was thinking. Did you have thoughts on this? Or were you hoping to see, you know, more about like system folks owning this? Or what was your thinking?
Ben Callahan (10:37)
No, I went and if I'm thinking for question one that there might not be any clear owner, then the correlation for me would be that everyone is figuring it out on the agentic workflow side. Because I think the person that if there is a a focus in own own owner for documentation, then that person would probably be the one building out the agentic workflows. I thought that there sh should be a correlation with that. yeah. Yeah. And I think I was hoping for that too, Effia. I tried to
079 deep dive screen raw (10:37)
No, I went and if I'm thinking for question one that there might not be any clear owner, then the correlation for me would be that everyone is figuring it out on the agentic workflow side. Because I think the person that if there is a a focus in own own owner for documentation, then that person would probably be the one building out the agentic workflows. I thought that there sh should be a correlation with that. yeah. Yeah. And I think I was hoping for that too, Effia. I tried to
079 deep dive speaker raw (10:37)
No, I went and if I'm thinking for question one that there might not be any clear owner, then the correlation for me would be that everyone is figuring it out on the agentic workflow side. Because I think the person that if there is a a focus in own own owner for documentation, then that person would probably be the one building out the agentic workflows. I thought that there sh should be a correlation with that. yeah. Yeah. And I think I was hoping for that too, Effia. I tried to
Ben Callahan (11:04)
Look at that. and actually dove into the data on that in a couple different ways. And there is absolutely no correlation. Yeah. Wow. So yeah. So I just did not show up in the data. So who knows if I I don't I don't suspect that that will hold. Like I do think there is a strong, I think it like you just said, having somebody who really understands that sort of system as context for A AI also carry that out in terms of the AI.
079 deep dive screen raw (11:04)
Look at that. and actually dove into the data on that in a couple different ways. And there is absolutely no correlation. Yeah. Wow. So yeah. So I just did not show up in the data. So who knows if I I don't I don't suspect that that will hold. Like I do think there is a strong, I think it like you just said, having somebody who really understands that sort of system as context for A AI also carry that out in terms of the AI.
079 deep dive speaker raw (11:04)
Look at that. and actually dove into the data on that in a couple different ways. And there is absolutely no correlation. Yeah. Wow. So yeah. So I just did not show up in the data. So who knows if I I don't I don't suspect that that will hold. Like I do think there is a strong, I think it like you just said, having somebody who really understands that sort of system as context for A AI also carry that out in terms of the AI.
Ben Callahan (11:32)
workflows makes a ton of sense, right? So perhaps it's a partnership with folks across different disciplines who need those workflows in their specific tooling. But there's got to be some at some point in the future that will I think be be relevant. we don't have to spend a ton of time diving into the rest of the details here, but I want to because I want to get the conversation opened up. So this is the point in the conversation where if you have a thread
079 deep dive screen raw (11:32)
workflows makes a ton of sense, right? So perhaps it's a partnership with folks across different disciplines who need those workflows in their specific tooling. But there's got to be some at some point in the future that will I think be be relevant. we don't have to spend a ton of time diving into the rest of the details here, but I want to because I want to get the conversation opened up. So this is the point in the conversation where if you have a thread
079 deep dive speaker raw (11:32)
workflows makes a ton of sense, right? So perhaps it's a partnership with folks across different disciplines who need those workflows in their specific tooling. But there's got to be some at some point in the future that will I think be be relevant. we don't have to spend a ton of time diving into the rest of the details here, but I want to because I want to get the conversation opened up. So this is the point in the conversation where if you have a thread
Ben Callahan (11:57)
Show us a check mark if you want to jump in and share just kind of a continuation of something that's being said. Raise your hand in Zoom and we'll get you into the conversation. while we're waiting for folks to volunteer, this was kind of an interesting thing I was looking at, which was comparing the success of agentic workflows with the idea of ownership of those workflows. And so I kind of just have a plot here that shows the answers on the left versus the top.
079 deep dive screen raw (11:57)
Show us a check mark if you want to jump in and share just kind of a continuation of something that's being said. Raise your hand in Zoom and we'll get you into the conversation. while we're waiting for folks to volunteer, this was kind of an interesting thing I was looking at, which was comparing the success of agentic workflows with the idea of ownership of those workflows. And so I kind of just have a plot here that shows the answers on the left versus the top.
079 deep dive speaker raw (11:57)
Show us a check mark if you want to jump in and share just kind of a continuation of something that's being said. Raise your hand in Zoom and we'll get you into the conversation. while we're waiting for folks to volunteer, this was kind of an interesting thing I was looking at, which was comparing the success of agentic workflows with the idea of ownership of those workflows. And so I kind of just have a plot here that shows the answers on the left versus the top.
Ben Callahan (12:26)
So on the left side, we're looking at who owns the workflow, the agentic workflows, and then across the top, the way that they purport success. and so you can kind of see a little bit of a thread here going top left to bottom right. but what was interesting to me was how high this 14 was, which is the no way to know. So when folks are figuring it out on their own, we don't have visibility, is kind of how I read this. I redrew that over here another way, which was trying to indicate ownership.
079 deep dive screen raw (12:26)
So on the left side, we're looking at who owns the workflow, the agentic workflows, and then across the top, the way that they purport success. and so you can kind of see a little bit of a thread here going top left to bottom right. but what was interesting to me was how high this 14 was, which is the no way to know. So when folks are figuring it out on their own, we don't have visibility, is kind of how I read this. I redrew that over here another way, which was trying to indicate ownership.
079 deep dive speaker raw (12:26)
So on the left side, we're looking at who owns the workflow, the agentic workflows, and then across the top, the way that they purport success. and so you can kind of see a little bit of a thread here going top left to bottom right. but what was interesting to me was how high this 14 was, which is the no way to know. So when folks are figuring it out on their own, we don't have visibility, is kind of how I read this. I redrew that over here another way, which was trying to indicate ownership.
Ben Callahan (12:56)
maybe indicates more about visibility than it does about success itself. Right. so interesting to kind of see that sort of thing showing up in the data. Robin, what's up, dude? Yeah, actually I I have a question about what I would like to hear, Afyia, your take on this. because I've seen that there are many people now working on how to encode like specifications for components or like in a structured way. and I
079 deep dive screen raw (12:56)
maybe indicates more about visibility than it does about success itself. Right. so interesting to kind of see that sort of thing showing up in the data. Robin, what's up, dude? Yeah, actually I I have a question about what I would like to hear, Afyia, your take on this. because I've seen that there are many people now working on how to encode like specifications for components or like in a structured way. and I
079 deep dive speaker raw (12:56)
maybe indicates more about visibility than it does about success itself. Right. so interesting to kind of see that sort of thing showing up in the data. Robin, what's up, dude? Yeah, actually I I have a question about what I would like to hear, Afyia, your take on this. because I've seen that there are many people now working on how to encode like specifications for components or like in a structured way. and I
Ben Callahan (13:25)
believe there's gonna be like maybe a standard at some point where there is like a a way like the same way the DTCG emerged like a way to specify structured specifications for component. But what I wonder is because right now I've seen many people putting the do's and don'ts inside these specifications with like a sentence that says in this case like you know do this or don't do that for s components and it's usually like a sentence, something very simple.
079 deep dive screen raw (13:25)
believe there's gonna be like maybe a standard at some point where there is like a a way like the same way the DTCG emerged like a way to specify structured specifications for component. But what I wonder is because right now I've seen many people putting the do's and don'ts inside these specifications with like a sentence that says in this case like you know do this or don't do that for s components and it's usually like a sentence, something very simple.
079 deep dive speaker raw (13:25)
believe there's gonna be like maybe a standard at some point where there is like a a way like the same way the DTCG emerged like a way to specify structured specifications for component. But what I wonder is because right now I've seen many people putting the do's and don'ts inside these specifications with like a sentence that says in this case like you know do this or don't do that for s components and it's usually like a sentence, something very simple.
Ben Callahan (13:55)
But I wonder how if if you have seen examples or maybe you thought already about like how to encode like more complex do's and don'ts and untime patterns when it comes to maybe compositions and page templates or something that has that encompasses more components. It's not a rule about one component, but it's about an organism or a page template, or it's like company wide, or it's a flow. Like I wonder how how people are gonna approach that in if
079 deep dive screen raw (13:55)
But I wonder how if if you have seen examples or maybe you thought already about like how to encode like more complex do's and don'ts and untime patterns when it comes to maybe compositions and page templates or something that has that encompasses more components. It's not a rule about one component, but it's about an organism or a page template, or it's like company wide, or it's a flow. Like I wonder how how people are gonna approach that in if
079 deep dive speaker raw (13:55)
But I wonder how if if you have seen examples or maybe you thought already about like how to encode like more complex do's and don'ts and untime patterns when it comes to maybe compositions and page templates or something that has that encompasses more components. It's not a rule about one component, but it's about an organism or a page template, or it's like company wide, or it's a flow. Like I wonder how how people are gonna approach that in if
Ben Callahan (14:24)
Someone already talked to about. I've been thinking about it, but I haven't. That's something that I'm actually working with PJ on to put that into the spec because we've managed to create guidelines for individual components, but for things like patterns, templates, recipes, do's and don'ts might be more complicated. I'd have I'm, you know, it's just something that I need to work through and test on how that comes out on the other side. nothing for
079 deep dive screen raw (14:24)
Someone already talked to about. I've been thinking about it, but I haven't. That's something that I'm actually working with PJ on to put that into the spec because we've managed to create guidelines for individual components, but for things like patterns, templates, recipes, do's and don'ts might be more complicated. I'd have I'm, you know, it's just something that I need to work through and test on how that comes out on the other side. nothing for
079 deep dive speaker raw (14:24)
Someone already talked to about. I've been thinking about it, but I haven't. That's something that I'm actually working with PJ on to put that into the spec because we've managed to create guidelines for individual components, but for things like patterns, templates, recipes, do's and don'ts might be more complicated. I'd have I'm, you know, it's just something that I need to work through and test on how that comes out on the other side. nothing for
Ben Callahan (14:53)
formal right now that I can share. But thank you for bringing it up because it's something that I can add to the growing list of things I want to add to that DSDS spec. Yeah, there's this just came up recently in over on LinkedIn. I'll grab a link to Nathan's post where he kind of mentions some of these ideas and then I asked him specifically this question, Robin. So let me, I'll grab that too and drop it in It's interesting to see all the different
079 deep dive screen raw (14:53)
formal right now that I can share. But thank you for bringing it up because it's something that I can add to the growing list of things I want to add to that DSDS spec. Yeah, there's this just came up recently in over on LinkedIn. I'll grab a link to Nathan's post where he kind of mentions some of these ideas and then I asked him specifically this question, Robin. So let me, I'll grab that too and drop it in It's interesting to see all the different
079 deep dive speaker raw (14:53)
formal right now that I can share. But thank you for bringing it up because it's something that I can add to the growing list of things I want to add to that DSDS spec. Yeah, there's this just came up recently in over on LinkedIn. I'll grab a link to Nathan's post where he kind of mentions some of these ideas and then I asked him specifically this question, Robin. So let me, I'll grab that too and drop it in It's interesting to see all the different
Ben Callahan (15:18)
like schemas that are being spun up and I need to spend some time on Nathan's as well because I think he's got some good stuff going. Yeah, for sure. Jump in. So one of the observations I've made in my work around sort of like this do's and don'ts and and being more opinionated about like use cases and trying to figure out when to use a specific component. There's just a couple things here. Like one, I think
079 deep dive screen raw (15:18)
like schemas that are being spun up and I need to spend some time on Nathan's as well because I think he's got some good stuff going. Yeah, for sure. Jump in. So one of the observations I've made in my work around sort of like this do's and don'ts and and being more opinionated about like use cases and trying to figure out when to use a specific component. There's just a couple things here. Like one, I think
079 deep dive speaker raw (15:18)
like schemas that are being spun up and I need to spend some time on Nathan's as well because I think he's got some good stuff going. Yeah, for sure. Jump in. So one of the observations I've made in my work around sort of like this do's and don'ts and and being more opinionated about like use cases and trying to figure out when to use a specific component. There's just a couple things here. Like one, I think
Ben Callahan (15:47)
Everyone's starting to become more versed in semantics and accessibility and roles of specific components is going to be important. And then that that idea probably needs to sort of start to extend into like patterns as well. But trying to pick between which thing to use and then also inside of that, different flavors of that thing are.
079 deep dive screen raw (15:47)
Everyone's starting to become more versed in semantics and accessibility and roles of specific components is going to be important. And then that that idea probably needs to sort of start to extend into like patterns as well. But trying to pick between which thing to use and then also inside of that, different flavors of that thing are.
079 deep dive speaker raw (15:47)
Everyone's starting to become more versed in semantics and accessibility and roles of specific components is going to be important. And then that that idea probably needs to sort of start to extend into like patterns as well. But trying to pick between which thing to use and then also inside of that, different flavors of that thing are.
Ben Callahan (16:15)
You know, that's why it's so complicated because it's not just a when you look at do's and don'ts, we think of them kind of flat, but then there's there's complexity in it, and that's why it's so hard and so confusing. and that's something that I've been working through in my work is is first the hey, pick this component to do this thing, and then second, once you've picked it, what are the sort of different variety of use cases that are known?
079 deep dive screen raw (16:15)
You know, that's why it's so complicated because it's not just a when you look at do's and don'ts, we think of them kind of flat, but then there's there's complexity in it, and that's why it's so hard and so confusing. and that's something that I've been working through in my work is is first the hey, pick this component to do this thing, and then second, once you've picked it, what are the sort of different variety of use cases that are known?
079 deep dive speaker raw (16:15)
You know, that's why it's so complicated because it's not just a when you look at do's and don'ts, we think of them kind of flat, but then there's there's complexity in it, and that's why it's so hard and so confusing. and that's something that I've been working through in my work is is first the hey, pick this component to do this thing, and then second, once you've picked it, what are the sort of different variety of use cases that are known?
Ben Callahan (16:45)
we can document and then obviously over time we're gonna have to add more as we kind of discover more because that's where the the the low visibility might happen. especially in a in a really large org like the one I work in where the system supports unknown like an unknown number of of different products. there's certainly a a a short list of the topmost, but there's a lot. So
079 deep dive screen raw (16:45)
we can document and then obviously over time we're gonna have to add more as we kind of discover more because that's where the the the low visibility might happen. especially in a in a really large org like the one I work in where the system supports unknown like an unknown number of of different products. there's certainly a a a short list of the topmost, but there's a lot. So
079 deep dive speaker raw (16:45)
we can document and then obviously over time we're gonna have to add more as we kind of discover more because that's where the the the low visibility might happen. especially in a in a really large org like the one I work in where the system supports unknown like an unknown number of of different products. there's certainly a a a short list of the topmost, but there's a lot. So
Ben Callahan (17:11)
You know, people have invented all sorts of different ways that they're using things, which you know I kind of look at as, well, those are not unsupported or or don't do it that way. Those are just unknowns that we need to to get surfaced and then actually like kind of figure out what to do with them. so yeah, yeah. That whole idea of more sort of larger scale compositional
079 deep dive screen raw (17:11)
You know, people have invented all sorts of different ways that they're using things, which you know I kind of look at as, well, those are not unsupported or or don't do it that way. Those are just unknowns that we need to to get surfaced and then actually like kind of figure out what to do with them. so yeah, yeah. That whole idea of more sort of larger scale compositional
079 deep dive speaker raw (17:11)
You know, people have invented all sorts of different ways that they're using things, which you know I kind of look at as, well, those are not unsupported or or don't do it that way. Those are just unknowns that we need to to get surfaced and then actually like kind of figure out what to do with them. so yeah, yeah. That whole idea of more sort of larger scale compositional
Ben Callahan (17:40)
guidelines is gonna be huge for any organization that's trying to do, you know, more sort of, you know, ephemeral UI stuff where it's like building UI on the fly, you know, for for users in specific use cases, right? Because we I don't know how else we could test any of this stuff. Like that's the thing I keep getting to is like when you allow these models to build anything, like how do you validate what they're building is actually good? You know, it's it's a it's gonna be interesting. But Doug, what are you thinking? just
079 deep dive screen raw (17:40)
guidelines is gonna be huge for any organization that's trying to do, you know, more sort of, you know, ephemeral UI stuff where it's like building UI on the fly, you know, for for users in specific use cases, right? Because we I don't know how else we could test any of this stuff. Like that's the thing I keep getting to is like when you allow these models to build anything, like how do you validate what they're building is actually good? You know, it's it's a it's gonna be interesting. But Doug, what are you thinking? just
079 deep dive speaker raw (17:40)
guidelines is gonna be huge for any organization that's trying to do, you know, more sort of, you know, ephemeral UI stuff where it's like building UI on the fly, you know, for for users in specific use cases, right? Because we I don't know how else we could test any of this stuff. Like that's the thing I keep getting to is like when you allow these models to build anything, like how do you validate what they're building is actually good? You know, it's it's a it's gonna be interesting. But Doug, what are you thinking? just
Ben Callahan (18:08)
Back to Robin's question, what how we've handled it. We our our design system is deployed as I think we're up to like 13 packages. And so there'll be like the UI kit, which is the core components, form components, and so forth. And so we have a distributed skill that devs can install that includes, like kind of embeds some of this knowledge. And the way we handled the cross-cutting things, or the things that didn't necessarily fit in one spot, in each of those libraries, we included a recipes, which would include
079 deep dive screen raw (18:08)
Back to Robin's question, what how we've handled it. We our our design system is deployed as I think we're up to like 13 packages. And so there'll be like the UI kit, which is the core components, form components, and so forth. And so we have a distributed skill that devs can install that includes, like kind of embeds some of this knowledge. And the way we handled the cross-cutting things, or the things that didn't necessarily fit in one spot, in each of those libraries, we included a recipes, which would include
079 deep dive speaker raw (18:08)
Back to Robin's question, what how we've handled it. We our our design system is deployed as I think we're up to like 13 packages. And so there'll be like the UI kit, which is the core components, form components, and so forth. And so we have a distributed skill that devs can install that includes, like kind of embeds some of this knowledge. And the way we handled the cross-cutting things, or the things that didn't necessarily fit in one spot, in each of those libraries, we included a recipes, which would include
Ben Callahan (18:37)
include some common combinations of the components, which would instruct it. And then we had for in the inverse, we have a gotchas file, which is basically like things that often go wrong or things to not do. And so there's not a ton of don't there is some level of don't do this, but it's more of a like use this instead or look out for this that you didn't generate code that looks like this. And so we the positives are reflected in the recipes, which might answer the question of how the AI can assemble components together.
079 deep dive screen raw (18:37)
include some common combinations of the components, which would instruct it. And then we had for in the inverse, we have a gotchas file, which is basically like things that often go wrong or things to not do. And so there's not a ton of don't there is some level of don't do this, but it's more of a like use this instead or look out for this that you didn't generate code that looks like this. And so we the positives are reflected in the recipes, which might answer the question of how the AI can assemble components together.
079 deep dive speaker raw (18:37)
include some common combinations of the components, which would instruct it. And then we had for in the inverse, we have a gotchas file, which is basically like things that often go wrong or things to not do. And so there's not a ton of don't there is some level of don't do this, but it's more of a like use this instead or look out for this that you didn't generate code that looks like this. And so we the positives are reflected in the recipes, which might answer the question of how the AI can assemble components together.
Ben Callahan (19:07)
And the negatives handled by a gotchas kind of grouped by topic. Nice. That's cool.
079 deep dive screen raw (19:07)
And the negatives handled by a gotchas kind of grouped by topic. Nice. That's cool.
079 deep dive speaker raw (19:07)
And the negatives handled by a gotchas kind of grouped by topic. Nice. That's cool.
Ben Callahan (19:16)
Okay. What else folks? Afyia, where do you want to take this conversation? There's a there's a thousand places we could go here. I I I saw in the comments really briefly, Kelby mentioned decision trees. I thought that's very interesting. More and more the agentic workflows are are turning into like decision trees when to do this, if not, or something else. So that's that was that's pretty interesting. that feels kind of deterministic though, right? Like we wouldn't even need
079 deep dive screen raw (19:16)
Okay. What else folks? Afyia, where do you want to take this conversation? There's a there's a thousand places we could go here. I I I saw in the comments really briefly, Kelby mentioned decision trees. I thought that's very interesting. More and more the agentic workflows are are turning into like decision trees when to do this, if not, or something else. So that's that was that's pretty interesting. that feels kind of deterministic though, right? Like we wouldn't even need
079 deep dive speaker raw (19:16)
Okay. What else folks? Afyia, where do you want to take this conversation? There's a there's a thousand places we could go here. I I I saw in the comments really briefly, Kelby mentioned decision trees. I thought that's very interesting. More and more the agentic workflows are are turning into like decision trees when to do this, if not, or something else. So that's that was that's pretty interesting. that feels kind of deterministic though, right? Like we wouldn't even need
Ben Callahan (19:45)
A model to do that. maybe a model could help us build that tree, or something. But once it's created, then it's easy to follow, you know. So that's cool. Yeah. Yeah. Kelby, you got anything else to add on that? I was gonna add that yes, once it's created, it's easy to follow. So I've actually changed the way that I'm really working as a designer in the design system space where
079 deep dive screen raw (19:45)
A model to do that. maybe a model could help us build that tree, or something. But once it's created, then it's easy to follow, you know. So that's cool. Yeah. Yeah. Kelby, you got anything else to add on that? I was gonna add that yes, once it's created, it's easy to follow. So I've actually changed the way that I'm really working as a designer in the design system space where
079 deep dive speaker raw (19:45)
A model to do that. maybe a model could help us build that tree, or something. But once it's created, then it's easy to follow, you know. So that's cool. Yeah. Yeah. Kelby, you got anything else to add on that? I was gonna add that yes, once it's created, it's easy to follow. So I've actually changed the way that I'm really working as a designer in the design system space where
Ben Callahan (20:14)
I'm starting with documentation to teach the agents what we need, what's correct, what's incorrect, and then following that up with other types of like use cases or maybe edge cases, as well as any kind of actual UI updates and doing the UI updates afterward, because there's less of a churn that way.
079 deep dive screen raw (20:14)
I'm starting with documentation to teach the agents what we need, what's correct, what's incorrect, and then following that up with other types of like use cases or maybe edge cases, as well as any kind of actual UI updates and doing the UI updates afterward, because there's less of a churn that way.
079 deep dive speaker raw (20:14)
I'm starting with documentation to teach the agents what we need, what's correct, what's incorrect, and then following that up with other types of like use cases or maybe edge cases, as well as any kind of actual UI updates and doing the UI updates afterward, because there's less of a churn that way.
Ben Callahan (20:43)
Because the agents can do it correctly regardless of what the UI actually looks like.
079 deep dive screen raw (20:43)
Because the agents can do it correctly regardless of what the UI actually looks like.
079 deep dive speaker raw (20:43)
Because the agents can do it correctly regardless of what the UI actually looks like.
Ben Callahan (20:48)
So it can follow it essentially like it can follow the recipes. Yeah. Even though the ingredients might change. Yeah. Yeah. Huh. Okay. It feels like there would be at some point some sort of like that's like a cyclical thing, like a chicken or an egg thing too. At some point, like it if you change the ingredients too much, the recipe no longer works, essentially, right? Right, right. Yeah. That's why, well, with everything else, it's
079 deep dive screen raw (20:48)
So it can follow it essentially like it can follow the recipes. Yeah. Even though the ingredients might change. Yeah. Yeah. Huh. Okay. It feels like there would be at some point some sort of like that's like a cyclical thing, like a chicken or an egg thing too. At some point, like it if you change the ingredients too much, the recipe no longer works, essentially, right? Right, right. Yeah. That's why, well, with everything else, it's
079 deep dive speaker raw (20:48)
So it can follow it essentially like it can follow the recipes. Yeah. Even though the ingredients might change. Yeah. Yeah. Huh. Okay. It feels like there would be at some point some sort of like that's like a cyclical thing, like a chicken or an egg thing too. At some point, like it if you change the ingredients too much, the recipe no longer works, essentially, right? Right, right. Yeah. That's why, well, with everything else, it's
Ben Callahan (21:14)
Iterative. It's constant living document. Yeah, yeah, yeah. Very cool. Awesome. Thanks for that example. Lance, what's up? Hey. Yeah, I've been thinking about this a lot too. We're definitely exploring ways to help make the agents more successful, but something I'm thinking about is I think with some success, having agents being able to build design system components.
079 deep dive screen raw (21:14)
Iterative. It's constant living document. Yeah, yeah, yeah. Very cool. Awesome. Thanks for that example. Lance, what's up? Hey. Yeah, I've been thinking about this a lot too. We're definitely exploring ways to help make the agents more successful, but something I'm thinking about is I think with some success, having agents being able to build design system components.
079 deep dive speaker raw (21:14)
Iterative. It's constant living document. Yeah, yeah, yeah. Very cool. Awesome. Thanks for that example. Lance, what's up? Hey. Yeah, I've been thinking about this a lot too. We're definitely exploring ways to help make the agents more successful, but something I'm thinking about is I think with some success, having agents being able to build design system components.
Ben Callahan (21:45)
I think like we need designers to have a better awareness of what the components do. And I think that the traditional docsite structure is kind of a little intimidating when it's page by page or tab by tab. So I'm thinking like design system adoption is probably going to go up, and expertise is probably going to be required to go up for all designers. And how do we meet that?
079 deep dive screen raw (21:45)
I think like we need designers to have a better awareness of what the components do. And I think that the traditional dock site structure is kind of a little intimidating when it's page by page or tab by tab. So I'm thinking like design system adoption is probably going to go up, and expertise is probably going to be required to go up for all designers. And how do we meet that?
079 deep dive speaker raw (21:45)
I think like we need designers to have a better awareness of what the components do. And I think that the traditional dock site structure is kind of a little intimidating when it's page by page or tab by tab. So I'm thinking like design system adoption is probably going to go up, and expertise is probably going to be required to go up for all designers. And how do we meet that?
Ben Callahan (22:14)
So they have a better understanding of the components that are available. And I think it needs to be more visual. Like I think of Mobbin and Pinterest, where you can see lots of things at once. so a different way to like catch agents from building the wrong thing or people from building the wrong thing in code is maybe teaching them what exists in a in a different way. Yeah, that's a that's a good.
079 deep dive screen raw (22:14)
So they have a better understanding of the components that are available. And I think it needs to be more visual. Like I think of Mobbin and Pinterest, where you can see lots of things at once. so a different way to like catch agents from building the wrong thing or people from building the wrong thing in code is maybe teaching them what exists in a in a different way. Yeah, that's a that's a good.
079 deep dive speaker raw (22:14)
So they have a better understanding of the components that are available. And I think it needs to be more visual. Like I think of Mobbin and Pinterest, where you can see lots of things at once. so a different way to like catch agents from building the wrong thing or people from building the wrong thing in code is maybe teaching them what exists in a in a different way. Yeah, that's a that's a good.
Ben Callahan (22:43)
Call. I think it's interesting because models are largely trained now on like everything that's publicly available to them, which is a massive internet of a lot of terrible experiences. You know, like in a sense, like most of the web's not accessible. Most of the web, you know, there are like dark patterns encoded in a lot of this stuff, right? So so that stuff kind of comes out in, you know, in the way these models sometimes create too. So being more intentional and like a more curated
079 deep dive screen raw (22:43)
Call. I think it's interesting because models are largely trained now on like everything that's publicly available to them, which is a massive internet of a lot of terrible experiences. You know, like in a sense, like most of the web's not accessible. Most of the web, you know, there are like dark patterns encoded in a lot of this stuff, right? So so that stuff kind of comes out in, you know, in the way these models sometimes create too. So being more intentional and like a more curated
079 deep dive speaker raw (22:43)
Call. I think it's interesting because models are largely trained now on like everything that's publicly available to them, which is a massive internet of a lot of terrible experiences. You know, like in a sense, like most of the web's not accessible. Most of the web, you know, there are like dark patterns encoded in a lot of this stuff, right? So so that stuff kind of comes out in, you know, in the way these models sometimes create too. So being more intentional and like a more curated
Ben Callahan (23:12)
set of things like you're describing, you know, the mobbins or or what a Pinterest or whatever, maybe is a way to counteract some of that. what else, folks?
079 deep dive screen raw (23:12)
set of things like you're describing, you know, the mobbins or or what a Pinterest or whatever, maybe is a way to counteract some of that. what else, folks?
079 deep dive speaker raw (23:12)
set of things like you're describing, you know, the mobbins or or what a Pinterest or whatever, maybe is a way to counteract some of that. what else, folks?
Ben Callahan (23:23)
Afyia, I'd I'd love to get your take on this. we asked folks in the last of the quantitative questions. So that question was in the last year, what have you changed? What have you actually changed about your docs? we gave them a handful of options to choose from. Restructured our content into a schema, added machine readable formats. And actually, Kevin, can I put you on the spot for this really quick? Cause you had a really good comment around this option. Would you just share this? Sure. Yeah.
079 deep dive screen raw (23:23)
Afyia, I'd I'd love to get your take on this. we asked folks in the last of the quantitative questions. So that question was in the last year, what have you changed? What have you actually changed about your docs? we gave them a handful of options to choose from. Restructured our content into a schema, added machine readable formats. And actually, Kevin, can I put you on the spot for this really quick? Cause you had a really good comment around this option. Would you just share this? Sure. Yeah.
079 deep dive speaker raw (23:23)
Afyia, I'd I'd love to get your take on this. we asked folks in the last of the quantitative questions. So that question was in the last year, what have you changed? What have you actually changed about your docs? we gave them a handful of options to choose from. Restructured our content into a schema, added machine readable formats. And actually, Kevin, can I put you on the spot for this really quick? Cause you had a really good comment around this option. Would you just share this? Sure. Yeah.
Ben Callahan (23:52)
So there's a little bit of a difference nowadays, never used to be, between machine readable and machine passable. So an LLM can quite easily read YAML, JSON, or prose So it'd be interesting to get a distinction between that and what are the benefits now if an LLM is going to be reading this content in using a syntax like JSON that, you know, has a lot of
079 deep dive screen raw (23:52)
So there's a little bit of a difference nowadays, never used to be, between machine readable and machine passable. So an LLM can quite easily read YAML, JSON, or Pros. So it'd be interesting to get a distinction between that and what are the benefits now if an LLM is going to be reading this content in using a syntax like JSON that, you know, has a lot of
079 deep dive speaker raw (23:52)
So there's a little bit of a difference nowadays, never used to be, between machine readable and machine passable. So an LLM can quite easily read YAML, JSON, or Pros. So it'd be interesting to get a distinction between that and what are the benefits now if an LLM is going to be reading this content in using a syntax like JSON that, you know, has a lot of
Ben Callahan (24:21)
features in it that are designed for machine parsable language in consider in in ad in contrast to machine readable. Yeah. When we were talking about this earlier, like I was just trying to think like what is it that why are we doing this? Like why are we taking, you know, large language models which have been trained largely on the way that we talk, you know, language, human language, and
079 deep dive screen raw (24:21)
features in it that are designed for machine passable language in consider in in ad in contrast to machine readable. Yeah. When we were talking about this earlier, like I was just trying to think like what is it that why are we doing this? Like why are we taking, you know, large language models which have been trained largely on the way that we talk, you know, language, human language, and
079 deep dive speaker raw (24:21)
features in it that are designed for machine passable language in consider in in ad in contrast to machine readable. Yeah. When we were talking about this earlier, like I was just trying to think like what is it that why are we doing this? Like why are we taking, you know, large language models which have been trained largely on the way that we talk, you know, language, human language, and
Ben Callahan (24:46)
And creating like recreating our docs in what we get we're calling machine readable, but just like much more clear is what we're doing. Really, we're clarifying for the model, right? We're saying like there's a pair of like you know, attribute value kind of things, right? And so we might have YAML or JSON or whatever, that's like it's very clear what we mean when we say these things. There, there's a definitive connection. but why are we doing that? And I I think probably efficiency of token usage, right? Long term, right? Where
079 deep dive screen raw (24:46)
And creating like recreating our docs in what we get we're calling machine readable, but just like much more clear is what we're doing. Really, we're clarifying for the model, right? We're saying like there's a pair of like you know, attribute value kind of things, right? And so we might have YAML or JSON or whatever, that's like it's very clear what we mean when we say these things. There, there's a definitive connection. but why are we doing that? And I I think probably efficiency of token usage, right? Long term, right? Where
079 deep dive speaker raw (24:46)
And creating like recreating our docs in what we get we're calling machine readable, but just like much more clear is what we're doing. Really, we're clarifying for the model, right? We're saying like there's a pair of like you know, attribute value kind of things, right? And so we might have YAML or JSON or whatever, that's like it's very clear what we mean when we say these things. There, there's a definitive connection. but why are we doing that? And I I think probably efficiency of token usage, right? Long term, right? Where
Ben Callahan (25:15)
If we're relying on these models to sort of parse all the English into something that's easy for them to use repeatedly, then that's gonna cost more tokens to do as opposed to giving them a much clearer articulated set of of instructions. So I don't know if that's it or not, but you're right, there's a big difference, right, between these. So 12 was 12 months, what have you done? Restructured, added these formats.
079 deep dive screen raw (25:15)
If we're relying on these models to sort of parse all the English into something that's easy for them to use repeatedly, then that's gonna cost more tokens to do as opposed to giving them a much clearer articulated set of of instructions. So I don't know if that's it or not, but you're right, there's a big difference, right, between these. So 12 was 12 months, what have you done? Restructured, added these formats.
079 deep dive speaker raw (25:15)
If we're relying on these models to sort of parse all the English into something that's easy for them to use repeatedly, then that's gonna cost more tokens to do as opposed to giving them a much clearer articulated set of of instructions. So I don't know if that's it or not, but you're right, there's a big difference, right, between these. So 12 was 12 months, what have you done? Restructured, added these formats.
Ben Callahan (25:40)
Written context files or agent instructions made language more explicit or ambiguous, nothing. And so I kind of grouped those a little bit into saying like some of them are like sort of on how we deliver it, right? Like the context files and agent instructions or making the format a little bit more readable versus foundational things like a schema or actually changing the content, you know. So I don't know if you agree with that split in those in that list, Afyia or if you've, you know.
079 deep dive screen raw (25:40)
Written context files or agent instructions made language more explicit or ambiguous, nothing. And so I kind of grouped those a little bit into saying like some of them are like sort of on how we deliver it, right? Like the context files and agent instructions or making the format a little bit more readable versus foundational things like a schema or actually changing the content, you know. So I don't know if you agree with that split in those in that list, AFIA, or if you've, you know.
079 deep dive speaker raw (25:40)
Written context files or agent instructions made language more explicit or ambiguous, nothing. And so I kind of grouped those a little bit into saying like some of them are like sort of on how we deliver it, right? Like the context files and agent instructions or making the format a little bit more readable versus foundational things like a schema or actually changing the content, you know. So I don't know if you agree with that split in those in that list, AFIA, or if you've, you know.
Ben Callahan (26:09)
What what we're what this shows is that people are doing making more changes to these sort of like surface level things. Yeah. And that's kind of the joke of like, well, we're we're serving bad content in more accessible ways to our models. You know, is do you agree with this or Yeah, that's kind of what what I I mentioned. yeah, though these top three look like they're doing look like we're doing real work, but are we if the content
079 deep dive screen raw (26:09)
What what we're what this shows is that people are doing making more changes to these sort of like surface level things. Yeah. And that's kind of the joke of like, well, we're we're serving bad content in more accessible ways to our models. You know, is do you agree with this or Yeah, that's kind of what what I I mentioned. yeah, though these top three look like they're doing look like we're doing real work, but are we if the content
079 deep dive speaker raw (26:09)
What what we're what this shows is that people are doing making more changes to these sort of like surface level things. Yeah. And that's kind of the joke of like, well, we're we're serving bad content in more accessible ways to our models. You know, is do you agree with this or Yeah, that's kind of what what I I mentioned. yeah, though these top three look like they're doing look like we're doing real work, but are we if the content
Ben Callahan (26:35)
isn't as in as explicit as it needs to be. If it's just bad content, then now we're going to serve up to the masses through the tools. Yeah. So, you know, I I say the focus should really be first and forem foremost on the content. Yeah. and then, you know, however we surface it, we surface it. But we'll we'll have confidence that the content is cr is is in good condition. Yeah. That's that's a that's a good call. I missed somebody. I know I saw a hand go up and I missed it because we were in the middle of something. So yeah, jump in, Stevie.
079 deep dive screen raw (26:35)
isn't as in as explicit as it needs to be. If it's just bad content, then now we're going to serve up to the masses through the tools. Yeah. So, you know, I I say the focus should really be first and forem foremost on the content. Yeah. and then, you know, however we surface it, we surface it. But we'll we'll have confidence that the content is cr is is in good condition. Yeah. That's that's a that's a good call. I missed somebody. I know I saw a hand go up and I missed it because we were in the middle of something. So yeah, jump in, Stevie.
079 deep dive speaker raw (26:35)
isn't as in as explicit as it needs to be. If it's just bad content, then now we're going to serve up to the masses through the tools. Yeah. So, you know, I I say the focus should really be first and forem foremost on the content. Yeah. and then, you know, however we surface it, we surface it. But we'll we'll have confidence that the content is cr is is in good condition. Yeah. That's that's a that's a good call. I missed somebody. I know I saw a hand go up and I missed it because we were in the middle of something. So yeah, jump in, Stevie.
Ben Callahan (27:04)
Yeah, it was me. I think you you kinda answered what I originally was gonna was gonna say, but this is like what you're talking about now is exactly where we're at in our documentation journey where we've kind of you know, we started out writing everything in Storybook and Figma and having all this stuff. And then, you know, as AI started to progress, we kind of converted everything into Markdown and our doc site. And now we've hit a point where we're like, well, you know, you know, we're we're starting to look at JSON and YAML and creating more specific specs.
079 deep dive screen raw (27:04)
Yeah, it was me. I think you you kinda answered what I originally was gonna was gonna say, but this is like what you're talking about now is exactly where we're at in our documentation journey where we've kind of you know, we started out writing everything in Storybook and Figma and having all this stuff. And then, you know, as AI started to progress, we kind of converted everything into Markdown and our doc site. And now we've hit a point where we're like, well, you know, you know, we're we're starting to look at JSON and YAML and creating more specific specs.
079 deep dive speaker raw (27:04)
Yeah, it was me. I think you you kinda answered what I originally was gonna was gonna say, but this is like what you're talking about now is exactly where we're at in our documentation journey where we've kind of you know, we started out writing everything in Storybook and Figma and having all this stuff. And then, you know, as AI started to progress, we kind of converted everything into Markdown and our doc site. And now we've hit a point where we're like, well, you know, you know, we're we're starting to look at JSON and YAML and creating more specific specs.
Ben Callahan (27:34)
Just because what we've found is a lot of the time the agent is just looking for an answer. And, you know, documentation can also be a negative when you have too much and it can dilute what the agent is looking for and ultimately it might go down the wrong path. So, you know, we found that the JSON and YAML was more effective at allowing the agent to just really hone in on exactly what it's looking for. and therefore getting our intention
079 deep dive screen raw (27:34)
Just because what we've found is a lot of the time the agent is just looking for an answer. And, you know, documentation can also be a negative when you have too much and it can dilute what the agent is looking for and ultimately it might go down the wrong path. So, you know, we found that the JSON and YAML was more effective at allowing the agent to just really hone in on exactly what it's looking for. and therefore getting our intention
079 deep dive speaker raw (27:34)
Just because what we've found is a lot of the time the agent is just looking for an answer. And, you know, documentation can also be a negative when you have too much and it can dilute what the agent is looking for and ultimately it might go down the wrong path. So, you know, we found that the JSON and YAML was more effective at allowing the agent to just really hone in on exactly what it's looking for. and therefore getting our intention
Ben Callahan (28:03)
To the agent in a more effective way. you know, so I think this is, you know, we we don't want to have to keep rewriting everything, right? I think laying the proper foundation and then figuring out how to distribute it out and to, you know, we can present that data in many different ways if we really need to. Yeah, that's great. jump in, Lance. Hey, yeah, we are also trying not to put
079 deep dive screen raw (28:03)
To the agent in a more effective way. you know, so I think this is, you know, we we don't want to have to keep rewriting everything, right? I think laying the proper foundation and then figuring out how to distribute it out and to, you know, we can present that data in many different ways if we really need to. Yeah, that's great. jump in, Lance. Hey, yeah, we are also trying not to put
079 deep dive speaker raw (28:03)
To the agent in a more effective way. you know, so I think this is, you know, we we don't want to have to keep rewriting everything, right? I think laying the proper foundation and then figuring out how to distribute it out and to, you know, we can present that data in many different ways if we really need to. Yeah, that's great. jump in, Lance. Hey, yeah, we are also trying not to put
Ben Callahan (28:32)
Too many values into MD files. We're taking a slightly different approach. We already had a doc site. so we've just made a an MCP for agents to be able to reference the doc site, which has our tokens and stuff. And then we do have like an agent's UI MD file, but it just
079 deep dive screen raw (28:32)
Too many values into MD files. We're taking a slightly different approach. We already had a doc site. so we've just made a an MCP for agents to be able to reference the doc site, which has our tokens and stuff. And then we do have like an agent's UI MD file, but it just
079 deep dive speaker raw (28:32)
Too many values into MD files. We're taking a slightly different approach. We already had a doc site. so we've just made a an MCP for agents to be able to reference the doc site, which has our tokens and stuff. And then we do have like an agent's UI MD file, but it just
Ben Callahan (28:56)
Tell it basically just tells it to build with our real components. It doesn't contain any of the actual values. so far that's been pretty successful. It seems to be able to do it. it will grab for the doc site MCP when it's like missing something or if there's a question. sometimes that's costly if you're trying to read too much of it, but it's been pretty good so far. Nice. Robin?
079 deep dive screen raw (28:56)
Tell it basically just tells it to build with our real components. It doesn't contain any of the actual values. so far that's been pretty successful. It seems to be able to do it. it will grab for the doc site MCP when it's like missing something or if there's a question. sometimes that's costly if you're trying to read too much of it, but it's been pretty good so far. Nice. Robin?
079 deep dive speaker raw (28:56)
Tell it basically just tells it to build with our real components. It doesn't contain any of the actual values. so far that's been pretty successful. It seems to be able to do it. it will grab for the doc site MCP when it's like missing something or if there's a question. sometimes that's costly if you're trying to read too much of it, but it's been pretty good so far. Nice. Robin?
Ben Callahan (29:27)
yeah, I actu I have another question where I would like to hear opinions. It's about like markdowns and and JSON slash YAML files. Because like one is prose, one is structured. Yeah. And I when when I heard everybody else talking, I thought that I probably I I can see maybe that there is value of still having both because and even to to actually craft both, because the prose one is very human readable, right? Like as a human I it's it's hard. I mean, I guess developers
079 deep dive screen raw (29:27)
yeah, I actu I have another question where I would like to hear opinions. It's about like markdowns and and JSON slash YAML files. Because like one is prose, one is structured. Yeah. And I when when I heard everybody else talking, I thought that I probably I I can see maybe that there is value of still having both because and even to to actually craft both, because the prose one is very human readable, right? Like as a human I it's it's hard. I mean, I guess developers
079 deep dive speaker raw (29:27)
yeah, I actu I have another question where I would like to hear opinions. It's about like markdowns and and JSON slash YAML files. Because like one is prose, one is structured. Yeah. And I when when I heard everybody else talking, I thought that I probably I I can see maybe that there is value of still having both because and even to to actually craft both, because the prose one is very human readable, right? Like as a human I it's it's hard. I mean, I guess developers
Ben Callahan (29:57)
probably different beasts, like they might prefer a JSON to Markdown. but like to to explain why did was something structured that way, like prose still might be valuable. So I was wondering if it's is if it's something you see also in the future too that we will both have a markdown file and YAML file, meaning structured data of of decisions, but also the prose that maybe that is then showed also on a nice website that for people to just go and read nicely instead of parsing through a JSON.
079 deep dive screen raw (29:57)
probably different beasts, like they might prefer a JSON to Markdown. but like to to explain why did was something structured that way, like prose still might be valuable. So I was wondering if it's is if it's something you see also in the future too that we will both have a markdown file and YAML file, meaning structured data of of decisions, but also the pro that maybe that is then showed also on a nice website that for people to just go and read nicely instead of parsing through a JSON.
079 deep dive speaker raw (29:57)
probably different beasts, like they might prefer a JSON to Markdown. but like to to explain why did was something structured that way, like prose still might be valuable. So I was wondering if it's is if it's something you see also in the future too that we will both have a markdown file and YAML file, meaning structured data of of decisions, but also the pro that maybe that is then showed also on a nice website that for people to just go and read nicely instead of parsing through a JSON.
Ben Callahan (30:26)
But also doing that exercise of taking prose and actually structure it, like s like taking what you've wrote in paragraphs and really converting it into structure. That also is a n nice exercise, I guess. It helps you distill decisions. Like do you see having both of those things still or or completely replace prose with with YAML, like in the most utopic case? Yeah, this is a good one. This is a great one. lots of thoughts here. Steven, do you have a response to Robin? Yeah, just a really quick one. I
079 deep dive screen raw (30:26)
But also doing that exercise of taking prose and actually structure it, like s like taking what you've wrote in paragraphs and really converting it into structure. That also is a n nice exercise, I guess. It helps you distill decisions. Like do you see having both of those things still or or completely replace prose with with YAML, like in the most utopic case? Yeah, this is a good one. This is a great one. lots of thoughts here. Steven, do you have a response to Robin? Yeah, just a really quick one. I
079 deep dive speaker raw (30:26)
But also doing that exercise of taking prose and actually structure it, like s like taking what you've wrote in paragraphs and really converting it into structure. That also is a n nice exercise, I guess. It helps you distill decisions. Like do you see having both of those things still or or completely replace prose with with YAML, like in the most utopic case? Yeah, this is a good one. This is a great one. lots of thoughts here. Steven, do you have a response to Robin? Yeah, just a really quick one. I
Ben Callahan (30:56)
I kind of I I definitely like agree with everything Robin just said there. And I think there is a place for both. And I think this is something that hasn't really been mentioned yet. But I think there is a difference between reference documentation and like guidance documentation, right? Whereas I think the reference documentation is the button has these variants and these sizes. This is like reference information that an agent needs. The guidance is how do I use these variants correctly in the scope of an entire page or an entire
079 deep dive screen raw (30:56)
I kind of I I definitely like agree with everything Robin just said there. And I think there is a place for both. And I think this is something that hasn't really been mentioned yet. But I think there is a difference between reference documentation and like guidance documentation, right? Whereas I think the reference documentation is the button has these variants and these sizes. This is like reference information that an agent needs. The guidance is how do I use these variants correctly in the scope of an entire page or an entire
079 deep dive speaker raw (30:56)
I kind of I I definitely like agree with everything Robin just said there. And I think there is a place for both. And I think this is something that hasn't really been mentioned yet. But I think there is a difference between reference documentation and like guidance documentation, right? Whereas I think the reference documentation is the button has these variants and these sizes. This is like reference information that an agent needs. The guidance is how do I use these variants correctly in the scope of an entire page or an entire
Ben Callahan (31:25)
feature within my app or product. I think guidance is obviously I think a lot more prone to being positive on the on the prose side, where you can actually maybe give more substance. whereas the reference data is more, I think, you know, better on the JSON or YAML side we've mixed it a little bit where we're starting to create YAML for things like do's and don'ts and guidance that we then transform into Markdown.
079 deep dive screen raw (31:25)
feature within my app or product. I think guidance is obviously I think a lot more prone to being positive on the on the prose side, where you can actually maybe give more substance. whereas the reference data is more, I think, you know, better on the JSON or YAML size. we've mixed it a little bit where we're starting to create YAML for things like do's and don'ts and guidance that we then transform into Markdown.
079 deep dive speaker raw (31:25)
feature within my app or product. I think guidance is obviously I think a lot more prone to being positive on the on the prose side, where you can actually maybe give more substance. whereas the reference data is more, I think, you know, better on the JSON or YAML size. we've mixed it a little bit where we're starting to create YAML for things like do's and don'ts and guidance that we then transform into Markdown.
Ben Callahan (31:54)
And then augment that with with more human readable stuff. But yeah. Yeah, I think it's a good, very good thought that Robin had there. Nice. Afyia do you have thoughts on this? No, this is an interesting conversation, I think. you know, from the point of view of a of a documentation specialist, I I write prose, of course, I write guidance.
079 deep dive screen raw (31:54)
And then augment that with with more human readable stuff. But yeah. Yeah, I think it's a good, very good thought that Robin had there. Nice. Afyia, do you have thoughts on this? No, this is an interesting conversation, I think. you know, from the point of view of a of a documentation specialist, I I write prose, of course, I write guidance.
079 deep dive speaker raw (31:54)
And then augment that with with more human readable stuff. But yeah. Yeah, I think it's a good, very good thought that Robin had there. Nice. Afyia, do you have thoughts on this? No, this is an interesting conversation, I think. you know, from the point of view of a of a documentation specialist, I I write prose, of course, I write guidance.
Ben Callahan (32:13)
I honestly wasn't quite sure what the difference was. And we worked through it. And I I even I think I even reached out to Kevin to kind of understand well why would what would be different in the formatting of those two, you know, human readable versus machine readable. Is are folks prone to using one more than the other or both? And I I'm seeing a mix of that, I guess. Yeah, I think I've been thinking about this. Kevin and I were actually talking about this the other day in
079 deep dive screen raw (32:13)
I honestly wasn't quite sure what the difference was. And we worked through it. And I I even I think I even reached out to Kevin to kind of understand well why would what would be different in the formatting of those two, you know, human readable versus machine readable. Is are folks prone to using one more than the other or both? And I I'm seeing a mix of that, I guess. Yeah, I think I've been thinking about this. Kevin and I were actually talking about this the other day in
079 deep dive speaker raw (32:13)
I honestly wasn't quite sure what the difference was. And we worked through it. And I I even I think I even reached out to Kevin to kind of understand well why would what would be different in the formatting of those two, you know, human readable versus machine readable. Is are folks prone to using one more than the other or both? And I I'm seeing a mix of that, I guess. Yeah, I think I've been thinking about this. Kevin and I were actually talking about this the other day in
Ben Callahan (32:41)
And Slack. I I was thinking about more like if you imagine what we do with tokens now, where we've sort of said, okay, you can sort of see your design decisions in a design, but you can codify them as like, you know, name value pairs as a token, right? So just a color is a name and a value. Well, I I I know people are working towards that in a much more complex scenario with like a component.
079 deep dive screen raw (32:41)
And Slack. I I was thinking about more like if you imagine what we do with tokens now, where we've sort of said, okay, you can sort of see your design decisions in a design, but you can codify them as like, you know, name value pairs as a token, right? So just a color is a name and a value. Well, I I I know people are working towards that in a much more complex scenario with like a component.
079 deep dive speaker raw (32:41)
And Slack. I I was thinking about more like if you imagine what we do with tokens now, where we've sort of said, okay, you can sort of see your design decisions in a design, but you can codify them as like, you know, name value pairs as a token, right? So just a color is a name and a value. Well, I I I know people are working towards that in a much more complex scenario with like a component.
Ben Callahan (33:06)
That's a lot of Nathan's work. You know, it's like, how do we articulate a component as structured content, structured data, right? And so if we can do that, then can we also just say that is really the source of truth? And from that, we can generate a Figma thing. From that, we can generate any code thing, any design component, actually, and documentation, you know, like, and so is there like a source of truth here that is really the documentation, the specification of the thing? And from that, it
079 deep dive screen raw (33:06)
That's a lot of Nathan's work. You know, it's like, how do we articulate a component as structured content, structured data, right? And so if we can do that, then can we also just say that is really the source of truth? And from that, we can generate a Figma thing. From that, we can generate any code thing, any design component, actually, and documentation, you know, like, and so is there like a source of truth here that is really the documentation, the specification of the thing? And from that, it
079 deep dive speaker raw (33:06)
That's a lot of Nathan's work. You know, it's like, how do we articulate a component as structured content, structured data, right? And so if we can do that, then can we also just say that is really the source of truth? And from that, we can generate a Figma thing. From that, we can generate any code thing, any design component, actually, and documentation, you know, like, and so is there like a source of truth here that is really the documentation, the specification of the thing? And from that, it
Ben Callahan (33:35)
being sort of focused on machine readable or like truly structured data, we could generate a prose version using like what AI is good at, you know, sort of like mimicking human voice and tone and stuff, you know, to kind of generate the human readable side. I don't know if that makes sense or not, but I'd love there to be like a single place where this stuff actually lives and from it we can get the the various artifacts that we need, you know. Yeah. Who knows? other thoughts on this?
079 deep dive screen raw (33:35)
being sort of focused on machine readable or like truly structured data, we could generate a prose version using like what AI is good at, you know, sort of like mimicking human voice and tone and stuff, you know, to kind of generate the human readable side. I don't know if that makes sense or not, but I'd love there to be like a single place where this stuff actually lives and from it we can get the the various artifacts that we need, you know. Yeah. Who knows? other thoughts on this?
079 deep dive speaker raw (33:35)
being sort of focused on machine readable or like truly structured data, we could generate a prose version using like what AI is good at, you know, sort of like mimicking human voice and tone and stuff, you know, to kind of generate the human readable side. I don't know if that makes sense or not, but I'd love there to be like a single place where this stuff actually lives and from it we can get the the various artifacts that we need, you know. Yeah. Who knows? other thoughts on this?
Ben Callahan (34:05)
There's another topic we can jump to up here, which might be a little controversial. All right. Quiet group today. So one of the things that we looked at in question five, which was the open text question, what does the future of system docs look like for you and your org? I kind of went through and was able to split it in a different couple different ways. So one of those ways was the type of change, like
079 deep dive screen raw (34:05)
There's another topic we can jump to up here, which might be a little controversial. All right. Quiet group today. So one of the things that we looked at in question five, which was the open text question, what does the future of system docs look like for you and your org? I kind of went through and was able to split it in a different couple different ways. So one of those ways was the type of change, like
079 deep dive speaker raw (34:05)
There's another topic we can jump to up here, which might be a little controversial. All right. Quiet group today. So one of the things that we looked at in question five, which was the open text question, what does the future of system docs look like for you and your org? I kind of went through and was able to split it in a different couple different ways. So one of those ways was the type of change, like
Ben Callahan (34:30)
16% of responses were like talking about delivery changes. In other words, we're going to make our docs accessible via MCP, chat, skills, playgrounds, that kind of thing, instead of just like go to the doc site, you know. 16% were consolidating, moving towards a single source of truth, kind of like what we were just talking about. 14% in authoring. So they were saying things like, our docs will be written by AI, and 10% disconnected from AI. So
079 deep dive screen raw (34:30)
16% of responses were like talking about delivery changes. In other words, we're going to make our docs accessible via MCP, chat, skills, playgrounds, that kind of thing, instead of just like go to the doc site, you know. 16% were consolidating, moving towards a single source of truth, kind of like what we were just talking about. 14% in authoring. So they were saying things like, our docs will be written by AI, and 10% disconnected from AI. So
079 deep dive speaker raw (34:30)
16% of responses were like talking about delivery changes. In other words, we're going to make our docs accessible via MCP, chat, skills, playgrounds, that kind of thing, instead of just like go to the doc site, you know. 16% were consolidating, moving towards a single source of truth, kind of like what we were just talking about. 14% in authoring. So they were saying things like, our docs will be written by AI, and 10% disconnected from AI. So
Ben Callahan (34:58)
this this problem space they were saying predates any any AI stuff. Documentation is a a thread that's been a problem for us for a long time, right? or a challenge for us. the other split that I thought was really interesting was on how folks talked about the future of their docs as it relates to human readers, human consumers of the docs and and machine consumers of the docs. So I sketched it out over here where you can see that like only five folks
079 deep dive screen raw (34:58)
this this problem space they were saying predates any any AI stuff. Documentation is a a thread that's been a problem for us for a long time, right? or a challenge for us. the other split that I thought was really interesting was on how folks talked about the future of their docs as it relates to human readers, human consumers of the docs and and machine consumers of the docs. So I sketched it out over here where you can see that like only five folks
079 deep dive speaker raw (34:58)
this this problem space they were saying predates any any AI stuff. Documentation is a a thread that's been a problem for us for a long time, right? or a challenge for us. the other split that I thought was really interesting was on how folks talked about the future of their docs as it relates to human readers, human consumers of the docs and and machine consumers of the docs. So I sketched it out over here where you can see that like only five folks
Ben Callahan (35:28)
in the in this answer referred to humans as the primary consumers of their documentation in the future. fourteen said only machines will be the consumers of the docs in the future. And then we have this primarily machines, which was like they acknowledge humans will still need them, but that was five people and then somewhere in the between primarily of these two was a balanced response. So I just thought this was fascinating to hear. It wasn't like the question didn't say,
079 deep dive screen raw (35:28)
in the in this answer referred to humans as the primary consumers of their documentation in the future. fourteen said only machines will be the consumers of the docs in the future. And then we have this primarily machines, which was like they acknowledge humans will still need them, but that was five people and then somewhere in the between primarily of these two was a balanced response. So I just thought this was fascinating to hear. It wasn't like the question didn't say,
079 deep dive speaker raw (35:28)
in the in this answer referred to humans as the primary consumers of their documentation in the future. fourteen said only machines will be the consumers of the docs in the future. And then we have this primarily machines, which was like they acknowledge humans will still need them, but that was five people and then somewhere in the between primarily of these two was a balanced response. So I just thought this was fascinating to hear. It wasn't like the question didn't say,
Ben Callahan (35:54)
Is this human or computer in the future that's reading these things? But I don't know if folks have thoughts on this. Like how does this sit with you? Does this make you feel nervous about your future or or hopeful? Doug? So okay. So I the my opinion on this is the same reason I read certain books on my Kindle and other books I buy physical. So I think the only machines is the Kindle where you want to find the information quickly. You need a machine to find it and read it.
079 deep dive screen raw (35:54)
Is this human or computer in the future that's reading these things? But I don't know if folks have thoughts on this. Like how does this sit with you? Does this make you feel nervous about your future or or hopeful? Doug? So okay. So I the my opinion on this is the same reason I read certain books on my Kindle and other books I buy physical. So I think the only machines is the Kindle where you want to find the information quickly. You need a machine to find it and read it.
079 deep dive speaker raw (35:54)
Is this human or computer in the future that's reading these things? But I don't know if folks have thoughts on this. Like how does this sit with you? Does this make you feel nervous about your future or or hopeful? Doug? So okay. So I the my opinion on this is the same reason I read certain books on my Kindle and other books I buy physical. So I think the only machines is the Kindle where you want to find the information quickly. You need a machine to find it and read it.
Ben Callahan (36:22)
But when I want to see what's available, if I'm looking at a recipe book or something, I want to flip through, I want to have something catch my eye and learn about that. And even in tech in things, I do the same thing. I'll go to a website for Apollo or some some technical thing. And I'll go page by page through the docs and looking at stuff to make myself aware of questions I didn't even know I should be asking or I should be prompting.
079 deep dive screen raw (36:22)
But when I want to see what's available, if I'm looking at a recipe book or something, I want to flip through, I want to have something catch my eye and learn about that. And even in tech in things, I do the same thing. I'll go to a website for Apollo or some some technical thing. And I'll go page by page through the docs and looking at stuff to make myself aware of questions I didn't even know I should be asking or I should be prompting.
079 deep dive speaker raw (36:22)
But when I want to see what's available, if I'm looking at a recipe book or something, I want to flip through, I want to have something catch my eye and learn about that. And even in tech in things, I do the same thing. I'll go to a website for Apollo or some some technical thing. And I'll go page by page through the docs and looking at stuff to make myself aware of questions I didn't even know I should be asking or I should be prompting.
Ben Callahan (36:45)
So I still think there's going to be a human readable side of this stuff for a while. So there's that that discovery of questions you didn't even know you should be asking, as opposed to I don't think that precludes the AIs from being able to get the questions to what they already know to ask or what you know to prompt them. But I still think we're going to need this discovery process that is visible in the future. Yeah. Totally. Sean, jump in. That was great. That was great, Doc. The analogy was awesome. I I work with with two different
079 deep dive screen raw (36:45)
So I still think there's going to be a human readable side of this stuff for a while. So there's that that discovery of questions you didn't even know you should be asking, as opposed to I don't think that precludes the AIs from being able to get the questions to what they already know to ask or what you know to prompt them. But I still think we're going to need this discovery process that is visible in the future. Yeah. Totally. Sean, jump in. That was great. That was great, Doc. The analogy was awesome. I I work with with two different
079 deep dive speaker raw (36:45)
So I still think there's going to be a human readable side of this stuff for a while. So there's that that discovery of questions you didn't even know you should be asking, as opposed to I don't think that precludes the AIs from being able to get the questions to what they already know to ask or what you know to prompt them. But I still think we're going to need this discovery process that is visible in the future. Yeah. Totally. Sean, jump in. That was great. That was great, Doc. The analogy was awesome. I I work with with two different
Ben Callahan (37:14)
two different clients on their design systems and they could not be polar opposites in in in in how they're how they're working and that has a real impact on the on the docs. You know, I work with with a government organization who is terrified of using AI and is nowhere and and so everything that we're producing is still it it's it's done for humans. We're still drawing things in Figma and and handwriting all of the code.
079 deep dive screen raw (37:14)
two different clients on their design systems and they could not be polar opposites in in in in how they're how they're working and that has a real impact on the on the docs. You know, I work with with a government organization who is terrified of using AI and is nowhere and and so everything that we're producing is still it it's it's done for humans. We're still drawing things in Figma and and handwriting all of the code.
079 deep dive speaker raw (37:14)
two different clients on their design systems and they could not be polar opposites in in in in how they're how they're working and that has a real impact on the on the docs. You know, I work with with a government organization who is terrified of using AI and is nowhere and and so everything that we're producing is still it it's it's done for humans. We're still drawing things in Figma and and handwriting all of the code.
Ben Callahan (37:42)
Whereas on the other side, I work with a with an AI startup and every line, 100% of their code is written by AI. Yeah. So we almost have no choice but to that that the machines are the primary consumers of of the documentation. But even in that situation, like there still needs to be there's still a step in the process where humans need to look at some docs. They they need to look at what are the list of components that I've got available in order to guide the machines when they write their prompts.
079 deep dive screen raw (37:42)
Whereas on the other side, I work with a with an AI startup and every line, 100% of their code is written by AI. Yeah. So we almost have no choice but to that that the machines are the primary consumers of of the documentation. But even in that situation, like there still needs to be there's still a step in the process where humans need to look at some docs. They they need to look at what are the list of components that I've got available in order to guide the machines when they write their prompts.
079 deep dive speaker raw (37:42)
Whereas on the other side, I work with a with an AI startup and every line, 100% of their code is written by AI. Yeah. So we almost have no choice but to that that the machines are the primary consumers of of the documentation. But even in that situation, like there still needs to be there's still a step in the process where humans need to look at some docs. They they need to look at what are the list of components that I've got available in order to guide the machines when they write their prompts.
Ben Callahan (38:13)
but the majority of the content over there, it's it's primarily focused on on being consumed by by machines. So yeah, just kind of really, yeah, really interesting conversation, especially on especially on this point, just like knowing the contrast between those those two folks that I work with. Yeah. Context is king there, right? That's it, that's almost like a cultural thing. It's like these two cultures or these orgs are so different. You know, what's what's palatable in one is not in the other, you know? Yes.
079 deep dive screen raw (38:13)
but the majority of the content over there, it's it's primarily focused on on being consumed by by machines. So yeah, just kind of really, yeah, really interesting conversation, especially on especially on this point, just like knowing the contrast between those those two folks that I work with. Yeah. Context is king there, right? That's it, that's almost like a cultural thing. It's like these two cultures or these orgs are so different. You know, what's what's palatable in one is not in the other, you know? Yes.
079 deep dive speaker raw (38:13)
but the majority of the content over there, it's it's primarily focused on on being consumed by by machines. So yeah, just kind of really, yeah, really interesting conversation, especially on especially on this point, just like knowing the contrast between those those two folks that I work with. Yeah. Context is king there, right? That's it, that's almost like a cultural thing. It's like these two cultures or these orgs are so different. You know, what's what's palatable in one is not in the other, you know? Yes.
Ben Callahan (38:41)
how does that sit with you, Afyia Are you is that aligned with your experience or what were you what was the culture like in your past experiences? Was it open to this or yeah, we so my past experience well, I'll just say here's what I will say about about this, about what I think. I I think we're in a great place where we don't really need to necessarily pick a side, right, human or AI. I think the goal,
079 deep dive screen raw (38:41)
how does that sit with you, Afyia? Are you is that aligned with your experience or what were you what was the culture like in your past experiences? Was it open to this or yeah, we so my past experience well, I'll just say here's what I will say about about this, about what I think. I I think we're in a great place where we don't really need to necessarily pick a side, right, human or AI. I think the goal,
079 deep dive speaker raw (38:41)
how does that sit with you, Afyia? Are you is that aligned with your experience or what were you what was the culture like in your past experiences? Was it open to this or yeah, we so my past experience well, I'll just say here's what I will say about about this, about what I think. I I think we're in a great place where we don't really need to necessarily pick a side, right, human or AI. I think the goal,
Ben Callahan (39:08)
is to be able to create a standard or create a a source of truth that is is structured well enough that it can feed or be delivered to any audience. So I'm I'm not necessarily for human or for AI. I'm my goal has always been to serve multiple audiences and to serve them what they need where they need it. Yeah. So and I've always pushed that and where wherever I was, whether that was at, you know, Epic Games or, you know, consulting or doing
079 deep dive screen raw (39:08)
is to be able to create a standard or create a a source of truth that is is structured well enough that it can feed or be delivered to any audience. So I'm I'm not necessarily for human or for AI. I'm my goal has always been to serve multiple audiences and to serve them what they need where they need it. Yeah. So and I've always pushed that and where wherever I was, whether that was at, you know, Epic Games or, you know, consulting or doing
079 deep dive speaker raw (39:08)
is to be able to create a standard or create a a source of truth that is is structured well enough that it can feed or be delivered to any audience. So I'm I'm not necessarily for human or for AI. I'm my goal has always been to serve multiple audiences and to serve them what they need where they need it. Yeah. So and I've always pushed that and where wherever I was, whether that was at, you know, Epic Games or, you know, consulting or doing
Ben Callahan (39:38)
Freelance work with clients. Yeah. Yeah. Steven, jump in. I think that is that's like a great thought. Like it's it's not really about where we think the documentation should be. It's where people are expecting it to be. And people are going to use the tool that they want. Like if people like and go to go and read the documentation, that's what they're going to do. If they want to ask Claude or whatever, what are all the components available in the design system? They're going to expect that an answer comes back. Whether we like it or not.
079 deep dive screen raw (39:38)
Freelance work with clients. Yeah. Yeah. Steven, jump in. I think that is that's like a great thought. Like it's it's not really about where we think the documentation should be. It's where people are expecting it to be. And people are going to use the tool that they want. Like if people like and go to go and read the documentation, that's what they're going to do. If they want to ask Claude or whatever, what are all the components available in the design system? They're going to expect that an answer comes back. Whether we like it or not.
079 deep dive speaker raw (39:38)
Freelance work with clients. Yeah. Yeah. Steven, jump in. I think that is that's like a great thought. Like it's it's not really about where we think the documentation should be. It's where people are expecting it to be. And people are going to use the tool that they want. Like if people like and go to go and read the documentation, that's what they're going to do. If they want to ask Claude or whatever, what are all the components available in the design system? They're going to expect that an answer comes back. Whether we like it or not.
Ben Callahan (40:06)
A lot of the time those AI tools will just hallucinate and just make something up if it doesn't know. And then they just think that that's the truth, right? I mean, that's what I've seen in our org is a lot of techno functional folks, product owners, people that are not necessarily technical that are asking questions, they just assume whatever the AI tells them is the truth. So our job has become like, well, if they're gonna use these tools, I want the information to be as correct as possible because they're gonna use them no matter what.
079 deep dive screen raw (40:06)
A lot of the time those AI tools will just hallucinate and just make something up if it doesn't know. And then they just think that that's the truth, right? I mean, that's what I've seen in our org is a lot of techno functional folks, product owners, people that are not necessarily technical that are asking questions, they just assume whatever the AI tells them is the truth. So our job has become like, well, if they're gonna use these tools, I want the information to be as correct as possible because they're gonna use them no matter what.
079 deep dive speaker raw (40:06)
A lot of the time those AI tools will just hallucinate and just make something up if it doesn't know. And then they just think that that's the truth, right? I mean, that's what I've seen in our org is a lot of techno functional folks, product owners, people that are not necessarily technical that are asking questions, they just assume whatever the AI tells them is the truth. So our job has become like, well, if they're gonna use these tools, I want the information to be as correct as possible because they're gonna use them no matter what.
Ben Callahan (40:35)
They expect to be able to ask these tools these questions in the same way that they can ask about their emails or documents in Confluence or Microsoft 365 or whatever it may be. Yeah. They expect the same thing. So, you know, and you can see this in people posting on LinkedIn all the time, right? Documentation sites like Tailwind and other places. I mean, their visitors are cratering to the point where it's, you know, they're losing millions and millions of of visitors a month because.
079 deep dive screen raw (40:35)
They expect to be able to ask these tools these questions in the same way that they can ask about their emails or documents in Confluence or Microsoft 365 or whatever it may be. Yeah. They expect the same thing. So, you know, and you can see this in people posting on LinkedIn all the time, right? Documentation sites like Tailwind and other places. I mean, their visitors are cratering to the point where it's, you know, they're losing millions and millions of of visitors a month because.
079 deep dive speaker raw (40:35)
They expect to be able to ask these tools these questions in the same way that they can ask about their emails or documents in Confluence or Microsoft 365 or whatever it may be. Yeah. They expect the same thing. So, you know, and you can see this in people posting on LinkedIn all the time, right? Documentation sites like Tailwind and other places. I mean, their visitors are cratering to the point where it's, you know, they're losing millions and millions of of visitors a month because.
Ben Callahan (41:04)
People just don't go and look at that stuff anymore. So Yeah. That's I think this aligns with, and maybe this is your your response, Stephen, but this quote here from the question five responses that says, our docs are already being consumed less and less by people. And we've found that most of our users are just trusting their preferred AI tools. So our goal is that our documentation and guidance is just injected into people's workflows as the default experience without them ever really knowing. Was that you, Stephen?
079 deep dive screen raw (41:04)
People just don't go and look at that stuff anymore. So Yeah. That's I think this aligns with, and maybe this is your your response, Stephen, but this quote here from the question five responses that says, our docs are already being consumed less and less by people. And we've found that most of our users are just trusting their preferred AI tools. So our goal is that our documentation and guidance is just injected into people's workflows as the default experience without them ever really knowing. Was that you, Stephen?
079 deep dive speaker raw (41:04)
People just don't go and look at that stuff anymore. So Yeah. That's I think this aligns with, and maybe this is your your response, Stephen, but this quote here from the question five responses that says, our docs are already being consumed less and less by people. And we've found that most of our users are just trusting their preferred AI tools. So our goal is that our documentation and guidance is just injected into people's workflows as the default experience without them ever really knowing. Was that you, Stephen?
Ben Callahan (41:32)
I'm pretty sure that was probably me. Yeah. Okay. Yeah. Perfect. Yeah. So you restated that. I think that's that's interesting. Yeah. And that's our that's our orgs goal is we don't want people to ever ask like build this with the design system. I want it to just if they're prototyping or building anything, it should just use these things and know this stuff without anybody ever even having to know that it's doing it. Yeah. so yeah. Yeah, it's it's like I agree with that. I also just
079 deep dive screen raw (41:32)
I'm pretty sure that was probably me. Yeah. Okay. Yeah. Perfect. Yeah. So you restated that. I think that's that's interesting. Yeah. And that's our that's our orgs goal is we don't want people to ever ask like build this with the design system. I want it to just if they're prototyping or building anything, it should just use these things and know this stuff without anybody ever even having to know that it's doing it. Yeah. so yeah. Yeah, it's it's like I agree with that. I also just
079 deep dive speaker raw (41:32)
I'm pretty sure that was probably me. Yeah. Okay. Yeah. Perfect. Yeah. So you restated that. I think that's that's interesting. Yeah. And that's our that's our orgs goal is we don't want people to ever ask like build this with the design system. I want it to just if they're prototyping or building anything, it should just use these things and know this stuff without anybody ever even having to know that it's doing it. Yeah. so yeah. Yeah, it's it's like I agree with that. I also just
Ben Callahan (42:02)
Acknowledge what the cost of that is, right? It's like removing the friction removes the requirement for somebody to go actually understand it themselves. And I think there's like obviously we have to kind of move in that direction, but I also just part of me is always like hesitant about like what that really caught what the real cost of that is. you know, for us as as humans, like what are we, what are we automating away from our own learning? Hattie.
079 deep dive screen raw (42:02)
Acknowledge what the cost of that is, right? It's like removing the friction removes the requirement for somebody to go actually understand it themselves. And I think there's like obviously we have to kind of move in that direction, but I also just part of me is always like hesitant about like what that really caught what the real cost of that is. you know, for us as as humans, like what are we, what are we automating away from our own learning? Hattie.
079 deep dive speaker raw (42:02)
Acknowledge what the cost of that is, right? It's like removing the friction removes the requirement for somebody to go actually understand it themselves. And I think there's like obviously we have to kind of move in that direction, but I also just part of me is always like hesitant about like what that really caught what the real cost of that is. you know, for us as as humans, like what are we, what are we automating away from our own learning? Hattie.
Ben Callahan (42:30)
I see your comment about meet people where they are and I've heard Afyia say that twice now about about docs. So will one of you or both of you say what you mean by that? Because I think people mean different things by that.
079 deep dive screen raw (42:30)
I see your comment about meet people where they are and I've heard Afyia say that twice now about about docs. So will one of you or both of you say what you mean by that? Because I think people mean different things by that.
079 deep dive speaker raw (42:30)
I see your comment about meet people where they are and I've heard Afyia say that twice now about about docs. So will one of you or both of you say what you mean by that? Because I think people mean different things by that.
Ben Callahan (42:43)
Does Hattie want to jump in first? Yeah. I mean, I answer questions all day in Microsoft Teams that is asking about how do I do this thing? Okay. If you're in Teams and asking me this, I need to make sure that some sort of documentation can be called or respond to you right away because that's where you already are. So if there's someone that wants to
079 deep dive screen raw (42:43)
Does Hattie want to jump in first? Yeah. I mean, I answer questions all day in Microsoft Teams that is asking about how do I do this thing? Okay. If you're in Teams and asking me this, I need to make sure that some sort of documentation can be called or respond to you right away because that's where you already are. So if there's someone that wants to
079 deep dive speaker raw (42:43)
Does Hattie want to jump in first? Yeah. I mean, I answer questions all day in Microsoft Teams that is asking about how do I do this thing? Okay. If you're in Teams and asking me this, I need to make sure that some sort of documentation can be called or respond to you right away because that's where you already are. So if there's someone that wants to
Ben Callahan (43:12)
Just ask questions. Great. Have we got an AI chatbot for you to read out those things? If it's someone that wants to go figure it out and like have the answer and take it to them with their product team or something like that, amazing. We have got a documentation site you can link to, right? So it's just having.
079 deep dive screen raw (43:12)
Just ask questions. Great. Have we got an AI chatbot for you to read out those things? If it's someone that wants to go figure it out and like have the answer and take it to them with their product team or something like that, amazing. We have got a documentation site you can link to, right? So it's just having.
079 deep dive speaker raw (43:12)
Just ask questions. Great. Have we got an AI chatbot for you to read out those things? If it's someone that wants to go figure it out and like have the answer and take it to them with their product team or something like that, amazing. We have got a documentation site you can link to, right? So it's just having.
Ben Callahan (43:35)
the breadth of things available to like however they're working, right? And I think that's a little bit of our like, I would say anybody's challenge because we've always had a problem with like people using components in Figma. And then there isn't really any place to like have documentation with it. Like use it this way, right? Yeah. Yeah. This this
079 deep dive screen raw (43:35)
the breadth of things available to like however they're working, right? And I think that's a little bit of our like, I would say anybody's challenge because we've always had a problem with like people using components in Figma. And then there isn't really any place to like have documentation with it. Like use it this way, right? Yeah. Yeah. This this
079 deep dive speaker raw (43:35)
the breadth of things available to like however they're working, right? And I think that's a little bit of our like, I would say anybody's challenge because we've always had a problem with like people using components in Figma. And then there isn't really any place to like have documentation with it. Like use it this way, right? Yeah. Yeah. This this
Ben Callahan (44:02)
To me is the future of documentation or what I thought it was when we when once we started experimenting with AI tools. I think way back when, and it's it's actually a phrase that you probably already know. It's called just in time docs or just in time documentation. It's the right docs in the right place where people need it for the right folks. and that has always been my goal whenever I was writing documentation. I had an understanding that people weren't reading it, they weren't going to
079 deep dive screen raw (44:02)
To me is the future of documentation or what I thought it was when we when once we started experimenting with AI tools. I think way back when, and it's it's actually a phrase that you probably already know. It's called just in time docs or just in time documentation. It's the right docs in the right place where people need it for the right folks. and that has always been my goal whenever I was writing documentation. I had an understanding that people weren't reading it, they weren't going to
079 deep dive speaker raw (44:02)
To me is the future of documentation or what I thought it was when we when once we started experimenting with AI tools. I think way back when, and it's it's actually a phrase that you probably already know. It's called just in time docs or just in time documentation. It's the right docs in the right place where people need it for the right folks. and that has always been my goal whenever I was writing documentation. I had an understanding that people weren't reading it, they weren't going to
Ben Callahan (44:31)
Some static reference site, they weren't contributing to it because it just wasn't where they were in the tools that they needed it. And I just I, you know, bang my head trying to figure out plugins of s of of some sort that could kind of surface up component guidance, maybe in Figmore or wherever the folks were. And, you know, it just wasn't possible at the time. And I think now, you know, we're in a place where we can do that, and we should. So one single source of truth that
079 deep dive screen raw (44:31)
Some static reference site, they weren't contributing to it because it just wasn't where they were in the tools that they needed it. And I just I, you know, bang my head trying to figure out plugins of s of of some sort that could kind of surface up component guidance, maybe in Figma or wherever the folks were. And, you know, it just wasn't possible at the time. And I think now, you know, we're in a place where we can do that, and we should. So one single source of truth that
079 deep dive speaker raw (44:31)
Some static reference site, they weren't contributing to it because it just wasn't where they were in the tools that they needed it. And I just I, you know, bang my head trying to figure out plugins of s of of some sort that could kind of surface up component guidance, maybe in Figmore or wherever the folks were. And, you know, it just wasn't possible at the time. And I think now, you know, we're in a place where we can do that, and we should. So one single source of truth that
Ben Callahan (45:00)
serves multiple audiences and can be served in multiple places however they need it in whatever format that is. prose and the code and storybook and Figma, just that's the ultimate goal. And that if you were to ask me right now what the future is for documentation, that to me would be the ideal. Yeah. Love that. Okay. Just a few minutes left. Doug, I see your check mark, my friend. What do you got for us? Okay. So
079 deep dive screen raw (45:00)
serves multiple audiences and can be served in multiple places however they need it in whatever format that is. prose and the code and storybook and Figma, just that's the ultimate goal. And that if you were to ask me right now what the future is for documentation, that to me would be the ideal. Yeah. Love that. Okay. Just a few minutes left. Doug, I see your check mark, my friend. What do you got for us? Okay. So
079 deep dive speaker raw (45:00)
serves multiple audiences and can be served in multiple places however they need it in whatever format that is. prose and the code and storybook and Figma, just that's the ultimate goal. And that if you were to ask me right now what the future is for documentation, that to me would be the ideal. Yeah. Love that. Okay. Just a few minutes left. Doug, I see your check mark, my friend. What do you got for us? Okay. So
Ben Callahan (45:28)
The only thing I was going to mention that one neg very negative, and this isn't necessarily in our own design systems, but is using the AI to write the documentation, and which is fine when an AI is consuming the documentation, but I'm just seeing this increase, I'm seeing internally, where people are just so productive and they're generating these gobs of things that I they're expecting me to read through. And then I'm seeing it now on actual like tech sites where I'm I'm reading tooling information, and you just see the same thought repeated three.
079 deep dive screen raw (45:28)
The only thing I was going to mention that one neg very negative, and this isn't necessarily in our own design systems, but is using the AI to write the documentation, and which is fine when an AI is consuming the documentation, but I'm just seeing this increase, I'm seeing internally, where people are just so productive and they're generating these gobs of things that I they're expecting me to read through. And then I'm seeing it now on actual like tech sites where I'm I'm reading tooling information, and you just see the same thought repeated three.
079 deep dive speaker raw (45:28)
The only thing I was going to mention that one neg very negative, and this isn't necessarily in our own design systems, but is using the AI to write the documentation, and which is fine when an AI is consuming the documentation, but I'm just seeing this increase, I'm seeing internally, where people are just so productive and they're generating these gobs of things that I they're expecting me to read through. And then I'm seeing it now on actual like tech sites where I'm I'm reading tooling information, and you just see the same thought repeated three.
Ben Callahan (45:57)
Four times, reword in different ways, rephrasing yourself. You're like, this is just the AI has written this, which again, when it's consuming it, doesn't matter. It it dedupes it and it just gives you the thought. But that's I don't know. It didn't, it was tangential. That's why I put the checkbox. But it's something that I think we have to fight with everything we can in our own systems to just bite the false productivity that AI gives you and and and really.
079 deep dive screen raw (45:57)
Four times, reword in different ways, rephrasing yourself. You're like, this is just the AI has written this, which again, when it's consuming it, doesn't matter. It it dedupes it and it just gives you the thought. But that's I don't know. It didn't, it was tangential. That's why I put the checkbox. But it's something that I think we have to fight with everything we can in our own systems to just bite the false productivity that AI gives you and and and really.
079 deep dive speaker raw (45:57)
Four times, reword in different ways, rephrasing yourself. You're like, this is just the AI has written this, which again, when it's consuming it, doesn't matter. It it dedupes it and it just gives you the thought. But that's I don't know. It didn't, it was tangential. That's why I put the checkbox. But it's something that I think we have to fight with everything we can in our own systems to just bite the false productivity that AI gives you and and and really.
Ben Callahan (46:22)
Put yourself in your user's shoes when you're reading the documentation. Is this sentence needed? Is this too wordy? Is it adding to something, you know? I don't know. But if it's something I think we're gonna it's gonna get worse before it gets better. For sure. Afyia thoughts on that? yeah. I whole of course, I wholeheartedly agree. there should you we should not be using AI to write the docs. That's all I have to say. Just don't do it, please. my gosh. I love it that you stated it so clearly.
079 deep dive screen raw (46:22)
Put yourself in your user's shoes when you're reading the documentation. Is this sentence needed? Is this too wordy? Is it adding to something, you know? I don't know. But if it's something I think we're gonna it's gonna get worse before it gets better. For sure. Afyia thoughts on that? yeah. I whole of course, I wholeheartedly agree. there should you we should not be using AI to write the docs. That's all I have to say. Just don't do it, please. my gosh. I love it that you stated it so clearly.
079 deep dive speaker raw (46:22)
Put yourself in your user's shoes when you're reading the documentation. Is this sentence needed? Is this too wordy? Is it adding to something, you know? I don't know. But if it's something I think we're gonna it's gonna get worse before it gets better. For sure. Afyia thoughts on that? yeah. I whole of course, I wholeheartedly agree. there should you we should not be using AI to write the docs. That's all I have to say. Just don't do it, please. my gosh. I love it that you stated it so clearly.
Ben Callahan (46:53)
And does anybody so does anybody disagree with that? Like I I've I feel that way because I think the most in in in one sense, one of the most valuable things about the docs is the alignment it requires you to find in order to put the put it in writing. So it's like the process of writing, it requires you to get everybody together, you know? So if you just hand it to an AI to write, you're not you're missing that. So who else thinks that is okay. All right, we got some people. Hattie, jump in.
079 deep dive screen raw (46:53)
And does anybody so does anybody disagree with that? Like I I've I feel that way because I think the most in in in one sense, one of the most valuable things about the docs is the alignment it requires you to find in order to put the put it in writing. So it's like the process of writing, it requires you to get everybody together, you know? So if you just hand it to an AI to write, you're not you're missing that. So who else thinks that is okay. All right, we got some people. Hattie, jump in.
079 deep dive speaker raw (46:53)
And does anybody so does anybody disagree with that? Like I I've I feel that way because I think the most in in in one sense, one of the most valuable things about the docs is the alignment it requires you to find in order to put the put it in writing. So it's like the process of writing, it requires you to get everybody together, you know? So if you just hand it to an AI to write, you're not you're missing that. So who else thinks that is okay. All right, we got some people. Hattie, jump in.
Ben Callahan (47:19)
Maybe maybe this is just my experience, but as the person that's kind of worked in the shadows until I have now an AI enablement role, I had to use AI to help me get some of this stuff together because I don't know as much about the I didn't I don't know heck of YAML languages or any of that type of stuff. So I would what I started off with is like we have documentation in Figma, we have documentation in Storybook, GitHub, we have a content system. I just
079 deep dive screen raw (47:19)
Maybe maybe this is just my experience, but as the person that's kind of worked in the shadows until I have now an AI enablement role, I had to use AI to help me get some of this stuff together because I don't know as much about the I didn't I don't know heck of YAML languages or any of that type of stuff. So I would what I started off with is like we have documentation in Figma, we have documentation in Storybook, GitHub, we have a content system. I just
079 deep dive speaker raw (47:19)
Maybe maybe this is just my experience, but as the person that's kind of worked in the shadows until I have now an AI enablement role, I had to use AI to help me get some of this stuff together because I don't know as much about the I didn't I don't know heck of YAML languages or any of that type of stuff. So I would what I started off with is like we have documentation in Figma, we have documentation in Storybook, GitHub, we have a content system. I just
Ben Callahan (47:49)
Brought all the links together. And I said, hey, let's start with this. And yeah, I'll get it reviewed by people, but to get me off of like a blank page and to have something to then be able to be reviewed. AI has really helped me not make those meetings working sessions and more let me have something available. We can talk through this asynchronously, giving a direction before, you know, you're kind of just sitting there staring at each other and you're like, how do we want to put this together?
079 deep dive screen raw (47:49)
Brought all the links together. And I said, hey, let's start with this. And yeah, I'll get it reviewed by people, but to get me off of like a blank page and to have something to then be able to be reviewed. AI has really helped me not make those meetings working sessions and more let me have something available. We can talk through this asynchronously, giving a direction before, you know, you're kind of just sitting there staring at each other and you're like, how do we want to put this together?
079 deep dive speaker raw (47:49)
Brought all the links together. And I said, hey, let's start with this. And yeah, I'll get it reviewed by people, but to get me off of like a blank page and to have something to then be able to be reviewed. AI has really helped me not make those meetings working sessions and more let me have something available. We can talk through this asynchronously, giving a direction before, you know, you're kind of just sitting there staring at each other and you're like, how do we want to put this together?
Ben Callahan (48:19)
Yeah. Can you just jump in on that? I think there's nothing wrong with that. AI assisted is fine for folks, non writing folks, moving from that dreaded blank page to some kind of you know, maybe skeleton draft. That's fine. It's the AI gen complete AI generation that I push back on. Yeah.
079 deep dive screen raw (48:19)
Yeah. Can you just jump in on that? I think there's nothing wrong with that. AI assisted is fine for folks, non writing folks, moving from that dreaded blank page to some kind of you know, maybe skeleton draft. That's fine. It's the AI gen complete AI generation that I push back on. Yeah.
079 deep dive speaker raw (48:19)
Yeah. Can you just jump in on that? I think there's nothing wrong with that. AI assisted is fine for folks, non writing folks, moving from that dreaded blank page to some kind of you know, memeskeleton draft. That's fine. It's the AI gen complete AI generation that I push back on. Yeah.
Ben Callahan (48:40)
Yes. Hattie, you good about you good you feel good about yeah, I ag agree. It was more just, you know, I like, yeah, I don't want AI taking all our jobs because it doesn't know any better than you know what it can read everywhere else. And yeah, you know. Lance, jump in. Yeah, I I agree with both of you. human writing is definitely the goal. I will say I have made a lot of naive mistakes in documentation.
079 deep dive screen raw (48:40)
Yes. Hattie, you good about you good you feel good about yeah, I ag agree. It was more just, you know, I like, yeah, I don't want AI taking all our jobs because it doesn't know any better than you know what it can read everywhere else. And yeah, you know. Lance, jump in. Yeah, I I agree with both of you. human writing is definitely the goal. I will say I have made a lot of naive mistakes in documentation.
079 deep dive speaker raw (48:40)
Yes. Hattie, you good about you good you feel good about yeah, I ag agree. It was more just, you know, I like, yeah, I don't want AI taking all our jobs because it doesn't know any better than you know what it can read everywhere else. And yeah, you know. Lance, jump in. Yeah, I I agree with both of you. human writing is definitely the goal. I will say I have made a lot of naive mistakes in documentation.
Ben Callahan (49:10)
writing too much and doing things that are too require too much human expertise like making imagery and stuff like that. So we're trying to find ways to like leverage AI to move quicker. And a lot of that is like having less human content that we can manage. yeah. Yeah. Yeah. Kevin.
079 deep dive screen raw (49:10)
writing too much and doing things that are too require too much human expertise like making imagery and stuff like that. So we're trying to find ways to like leverage AI to move quicker. And a lot of that is like having less human content that we can manage. yeah. Yeah. Yeah. Kevin.
079 deep dive speaker raw (49:10)
writing too much and doing things that are too require too much human expertise like making imagery and stuff like that. So we're trying to find ways to like leverage AI to move quicker. And a lot of that is like having less human content that we can manage. yeah. Yeah. Yeah. Kevin.
Ben Callahan (49:39)
So definitely agree with you. I I've had to deal with this and generated a lot of really crappy docs from AI. and had to read it as well. the one thing that it is good at is taking the schema that we already have that might be presented in one particular way and reformatting it. almost deterministically, but not quite. It is good at that and presenting it maybe on a different format, like into an HTML page or something like that.
079 deep dive screen raw (49:39)
So definitely agree with you. I I've had to deal with this and generated a lot of really crappy docs from AI. and had to read it as well. the one thing that it is good at is taking the schema that we already have that might be presented in one particular way and reformatting it. almost deterministically, but not quite. It is good at that and presenting it maybe on a different format, like into an HTML page or something like that.
079 deep dive speaker raw (49:39)
So definitely agree with you. I I've had to deal with this and generated a lot of really crappy docs from AI. and had to read it as well. the one thing that it is good at is taking the schema that we already have that might be presented in one particular way and reformatting it. almost deterministically, but not quite. It is good at that and presenting it maybe on a different format, like into an HTML page or something like that.
Ben Callahan (50:09)
that that you can visually see it. But as soon as it starts writing prose, it's cringe and it's really hard to actually want to stomach through reading it. Cause you just know that that there's there's no there's no experience behind it. And and that's the feeling that that I get from it. but for just writing bullet points that come from data, that's cool. Okay. Yeah.
079 deep dive screen raw (50:09)
that that you can visually see it. But as soon as it starts writing prose, it's cringe and it's really hard to actually want to stomach through reading it. Cause you just know that that there's there's no there's no experience behind it. And and that's the feeling that that I get from it. but for just writing bullet points that come from data, that's cool. Okay. Yeah.
079 deep dive speaker raw (50:09)
that that you can visually see it. But as soon as it starts writing prose, it's cringe and it's really hard to actually want to stomach through reading it. Cause you just know that that there's there's no there's no experience behind it. And and that's the feeling that that I get from it. but for just writing bullet points that come from data, that's cool. Okay. Yeah.
Ben Callahan (50:38)
nice. We've got just a minute left. Greg, what you got? yeah, just kind of adding to that, I think one of the things that I'm I want to experiment with myself. I don't know if this will work, but it seems like it could is you know with AI we have the ability to add the context to help determine like standards of like say it's translating from you know deterministic or or
079 deep dive screen raw (50:38)
nice. We've got just a minute left. Greg, what you got? yeah, just kind of adding to that, I think one of the things that I'm I want to experiment with myself. I don't know if this will work, but it seems like it could is you know with AI we have the ability to add the context to help determine like standards of like say it's translating from you know deterministic or or
079 deep dive speaker raw (50:38)
nice. We've got just a minute left. Greg, what you got? yeah, just kind of adding to that, I think one of the things that I'm I want to experiment with myself. I don't know if this will work, but it seems like it could is you know with AI we have the ability to add the context to help determine like standards of like say it's translating from you know deterministic or or
Ben Callahan (51:07)
structured data, how it transforms that into prose. I think there's some things there that that are seem possible to help get it to be better. Not going to be as good as someone that's good at it, but better than maybe some of the cringe default stuff that that we tend to experience, you know, getting it to
079 deep dive screen raw (51:07)
structured data, how it transforms that into prose. I think there's some things there that that are seem possible to help get it to be better. Not going to be as good as someone that's good at it, but better than maybe some of the cringe default stuff that that we tend to experience, you know, getting it to
079 deep dive speaker raw (51:07)
structured data, how it transforms that into prose. I think there's some things there that that are seem possible to help get it to be better. Not going to be as good as someone that's good at it, but better than maybe some of the cringe default stuff that that we tend to experience, you know, getting it to
Ben Callahan (51:34)
make sure it's deduping its stuff and not repeating things. Yeah. All that kind of base level stuff that we experience when we're just having a chat and iterating on stuff with with you know a chat agent. but yeah, like there's you know the there's there's context everywhere. So you know we can add we can infuse context in different places and that's I think the the meta ness of all of this that as we kind of dig deeper and try different things that those are
079 deep dive screen raw (51:34)
make sure it's deduping its stuff and not repeating things. Yeah. All that kind of base level stuff that we experience when we're just having a chat and iterating on stuff with with you know a chat agent. but yeah, like there's you know the there's there's context everywhere. So you know we can add we can infuse context in different places and that's I think the the meta ness of all of this that as we kind of dig deeper and try different things that those are
079 deep dive speaker raw (51:34)
make sure it's deduping its stuff and not repeating things. Yeah. All that kind of base level stuff that we experience when we're just having a chat and iterating on stuff with with you know a chat agent. but yeah, like there's you know the there's there's context everywhere. So you know we can add we can infuse context in different places and that's I think the the meta ness of all of this that as we kind of dig deeper and try different things that those are
Ben Callahan (52:04)
interesting places to play with this and see if we can get better outputs, you know. but you know it's gotta come from the human has to define what is that structured information, what are those decisions that the AI doesn't know for us. So yeah. Thank you, Greg. Afyia, you're awesome. Thank you for bringing this conversation to us. It's one that's well past due for us as a community. So thanks for for bringing it. just calling out if you haven't yet given some suggestions.
079 deep dive screen raw (52:04)
interesting places to play with this and see if we can get better outputs, you know. but you know it's gotta come from the human has to define what is that structured information, what are those decisions that the AI doesn't know for us. So yeah. Thank you, Greg. Afyia, you're awesome. Thank you for bringing this conversation to us. It's one that's well past due for us as a community. So thanks for for bringing it. just calling out if you haven't yet given some suggestions.
079 deep dive speaker raw (52:04)
interesting places to play with this and see if we can get better outputs, you know. but you know it's gotta come from the human has to define what is that structured information, what are those decisions that the AI doesn't know for us. So yeah. Thank you, Greg. Afyia, you're awesome. Thank you for bringing this conversation to us. It's one that's well past due for us as a community. So thanks for for bringing it. just calling out if you haven't yet given some suggestions.
Ben Callahan (52:33)
To the name for Annette's RV. I'll leave this big gym open for a little while so you can get those in. she's a beauty. lots of fun stuff happening in the space. Again, get connected to Afyia and I out on LinkedIn. Of course, listen to some more Annie Lennox this week. Redwoods is open. Come check it out there if you like this kind of conversation. We're we're always having this stuff. All the deep dive and recaps are released for the last episode 78 with Kacey
079 deep dive screen raw (52:33)
To the name for Annette's RV. I'll leave this big gym open for a little while so you can get those in. she's a beauty. lots of fun stuff happening in the space. Again, get connected to Afyia and I out on LinkedIn. Of course, listen to some more Annie Lennox this week. Redwoods is open. Come check it out there if you like this kind of conversation. We're we're always having this stuff. All the deep dive and recaps are released for the last episode 78 with Casey.
079 deep dive speaker raw (52:33)
To the name for Annette's RV. I'll leave this big gym open for a little while so you can get those in. she's a beauty. lots of fun stuff happening in the space. Again, get connected to Afyia and I out on LinkedIn. Of course, listen to some more Annie Lennox this week. Redwoods is open. Come check it out there if you like this kind of conversation. We're we're always having this stuff. All the deep dive and recaps are released for the last episode 78 with Casey.
Ben Callahan (53:01)
Couple articles here. Kaelig just released this incredible state of AI design systems tool. Please go check this out. really cool stuff. And gosh, just lots happening in the space. As always, some really good stuff happening in the community here too. So check all of it out. It's great to see you all. Afyia, thank you. Thank you. Yeah, we'll see you all in a week. Thanks, everybody.
079 deep dive screen raw (53:01)
Couple articles here. Kayleigh just released this incredible state of AI design systems tool. Please go check this out. really cool stuff. And gosh, just lots happening in the space. As always, some really good stuff happening in the community here too. So check all of it out. It's great to see you all. Afyia, thank you. Thank you. Yeah, we'll see you all in a week. Thanks, everybody.
079 deep dive speaker raw (53:01)
Couple articles here. Kayleigh just released this incredible state of AI design systems tool. Please go check this out. really cool stuff. And gosh, just lots happening in the space. As always, some really good stuff happening in the community here too. So check all of it out. It's great to see you all. Afyia, thank you. Thank you. Yeah, we'll see you all in a week. Thanks, everybody.
Ben Callahan (53:25)
Thank you so much for joining us on this episode of The Question. Remember, you can get access to the raw data, the collaborative fig jam, and all of the recordings for this episode and so many more on my website, bencallahan.com. If you or your team could use an outside perspective in your design system efforts, I'd be honored to support you in that way. There's more information about my coaching and consulting offerings over on the site. And if you need a hand building a sustainable design system program,
My studio Sparkbox is standing by ready to partner. Hiring Sparkbox is an investment in your design system and your team. Reach out and let's see if there's a good fit. Thank you so much for being here and remember, stay in learning mode.