Subscribe
Share
Share
Embed
- Kin Lane - personal website
- API Evangelist
- Kin on GitHub
- Kin on Socials: Bluesky Mastodon LinkedIn
- Check out Kin's Podcast, API Evangelist Conversations (Spotify & Apple Podcasts)
Creators and Guests
Host
What is APIs You Won't Hate?
A no-nonsense (well, some-nonsense) podcast about API design & development, new features in the world of HTTP, service-orientated architecture, microservices, and probably bikes.
[00:00:00]
Introduction and Welcome
---
Phil Sturgeon: Hello everybody. Welcome to APOs You Won't Hate. Uh, I'm Phil Sturgeon and I'm really excited this time to be joined by the very kin Lane, uh, also known as the API Evangelist. How you doing? Kin?
You're in New York City. I'm so jealous. I used to live there and I don't now. It makes me mad. How is it?
Have you got ice? Have you got, have you got ice all off in New York City or are they staying outta there?
Yeah,
yeah. I just feel like they're, they're gonna get to like the Bronx and wish they didn't. It's gonna be real interesting when they get there.
Oops, let's go back to Portland. This was a bad choice.[00:01:00]
Yeah.
Catching Up with Kin Lane
---
Phil Sturgeon: Anyway, that's, that's not exactly what we're here to talk about, but um. Excellent. Uh, it's been a while since I've seen you in person. You're at all the API conferences and I don't do those anymore, but, uh, but hey, what, what are you up to at the moment? So like you were super involved in open API for a bit.
You were involved Postman for a bit. Like what's going on? What have you been up to?[00:02:00]
Ooh.
Huh.
That's very cool.
Bloomberg and Spectral APIs
---
Phil Sturgeon: Uh, firstly, what the heck is Bloomberg doing with Spectral? They just got big APIs and they wanted a style guide or what's going on there? I've not heard about that one.
I.
Yeah.[00:03:00]
That's, that is really interesting. 'cause usually it's just like, uh, this team is in Singapore, or some shit. Or it's just like, that team is run by Gary, and I fucking hate Gary, but like the, the lack of communication between teams is usually not like legally mandated wherever I've been, even, even at WeWork as much of a shit state as that was.
That's quite funny.
Mm.
[00:04:00] Yeah.
Hmm.
Yeah.
We gotta push some non-standard garbage to production right now that, that we'll be trapped with for a million years. But, uh, yeah, we couldn't possibly wait another day to fix that before it goes out. That would be bad. Yeah, that is, that is stressful.
[00:05:00] Yeah. Very nice.
Yeah.
Yeah, yeah, yeah. I mean, yeah. Most of my brain right now is focused on the fact that, uh, we have spent two years waiting for our next bit of land to come through, and as soon as like one thing needs to happen, we've got half a million pounds just sitting in a bank account waiting to be given to this person, and as soon as they like do one thing, it's literally like send one email.
We can just buy it. It's so fucking frustrating. We've been here for like three months and what I wanna do is go and like, work with the beavers that are already on that site and design like a [00:06:00] sweet habitat for them. They're gonna love it. I'm gonna love working with them. It's gonna be a nice time. Um, but instead, yeah, just like hitting refresh on my inbox consistently is pretty stressful.
But, um.
Frustrations with AI Pitches
---
Phil Sturgeon: When I'm not hitting refresh on my inbox, I, or when I am hitting refresh on my inbox, I'm seeing emails hitting the protector, uh, inbox. Um, such as Rick from current company. I won't mention. Just absolutely, uh, convincing me that I need this new AI startup, this AI startup is, um, going to read all of the emails that hit my inbox in my own voice.
Uh, I know y'all think I love the sound of my own voice, but god dammit, I actually do not wanna hear it all that much. Not, not, not like reading an email from an asshole in my voice as well is a weird pitch. So I get pretty frustrated with people sending these AI pitches sometimes to APOs you won't hate.
Fair enough. You want us to talk about it, sending it to Protect Earth. However, does my head in, so I got this email the other day. I responded with, I'm not investing [00:07:00] any time or effort thinking about the AI bubble, especially not on this email address you scraped online somewhere. If you thought about the address you were contacting, you would've noticed that email is, uh, the email address is a climate and nature charity, which is being absolutely devastated by compute, burning slot machines like the one you're working on.
I'd recommend you look for another job. Opportunity promptly and consider getting into Green Tech. Linked to my article about Green Tech, what is that and how to get into it if you'd like to do something constructive instead. Warm regards. Right.
AI in the API World
---
Phil Sturgeon: Um, so I am not necessarily the most reasonable, uh, person when it comes to talking about ai.
Now, I, I come back all this up with a long chat. We can talk for fucking six hours about why this is stupid idea. But I did really appreciate your blog post the other day where you had a few more useful things to say compared to that nonsense that I just wrote to that poor person who's just trying to make a wage, um.
You wrote, I entered into these conversations in the early part of the summer with much empathy for people doing artificial intelligence without, sorry, without much empathy for people doing artificial intelligence. You are the same as me, [00:08:00] and I've come out the other side with a lot more empathy for what's happening and where we are headed.
I still feel strongly that AI is over hyped and overblown and obfuscates some very dangerous realities around labor, climate change, and copyright. As I've learned more, I've softened the edge on how I talk with folks about this moment that we are in, and specifically regarding how we are going to address what is next without jeopardizing what we've already built.
Now that is pretty well worded. Do you wanna expand on that whole bunch? Like wha what have you learned that makes you not think that everyone working on AI stuff is just being a silly money grabbing nut tea?
Yeah.
That's funny. Yeah. I mean, obviously the, um, in most companies there's the suits who, who don't really know anything about anything, who have just heard that like this is the future. Um, 'cause there are, there, there's, there's two types of suits. There's the ones that, no, that this is bullshit and are pushing it to make money.
And there's the ones that. Don't know that this is bullshit and are desperately trying to keep up so that they can make some money. I mean, that's just, that's just capitalism right there. Um, but I feel like, yeah, there's a lot of people, there's a lot of people in the C-suite just being like, wow, AI is gonna be the thing we all gonna do.
Ai, shove it in [00:11:00] everywhere, blah, make some money. And, um, they're all British. Even if it's in Silicon Valley, that's just the money. That's the money guy voice that I do. I don't have another one. Um. Yeah, so obviously the people being, being given their deliverables, like KPIs get set, product targets are given, sales are are being told.
You know, sales are always like defining the roadmap of just like we've been, we told everyone that we can do this and now you have to go figure it out. So there's a lot of layers of business, um, in between kind of the person who's writing the feature and I'm, I'm never usually mad at them. It is just, uh.
Challenges in API Development
---
Phil Sturgeon: A lot of, I feel like basically my, my thing with ai, especially in the API scene, we could talk about AI in general. That's huge. But like in the API scene, there was an API Days conference, I went to, I don't remember what it was. I think it was 2020 in Paris, and it was like before they started doing AI tracks.
And it was just like everyone was there that year talking about ai. Obviously that's why they got AI [00:12:00] tracks after that. And it was just loads to note of people being like. We are developing this tool, which means that you don't need. To, you don't ever need to integrate with an API ever, again, you don't ever need to think about any of the data that's being handed around.
You can just have a random API and, and you can generate a random front end for it based on vibes. And it will just be brilliant and absolutely fine. And you got the whole way through his 45 minute odd talk going on about how you never need to write integration code ever again. And it would all just work perfectly.
And, and one, the first question was like. So how does any of it work? Is it literally, can you just point it at any API like you've been saying, or like, do you have to. Build certain conventions into your API's, like, ah, yes, well you do need to like rewrite your API to follow these specific, very restrictive rules and guidance and everything else.
Right? And obviously this is what became the precursor for MCP. They wanted to add a layer in between a perfectly [00:13:00] good functioning API and whatever bullshit the slot machine required to function. And so that. It was frustrating for a long time that like my very first introduction to it in the API world was someone lying through their teeth and then crumbling under the very first follow up question they got.
Um, but since then, people just seem to have got better at hiding some of that, and, and it's never got better at functioning. It just seems to be they're better at selling you on the product and convincing you that you should try really hard, by which point you're kind of locked in. So. I haven't been delving into delving, uh, into the kind of products as much lately to see if they've got any better, and I'm sure people are working hard on solving it.
And I imagine that, that, that may well be what you are trying to do, right? Like, make this stuff work good. Because from from the early days, I kind of went and then I just cut and stopped paying attention. Like how, how do you see AI. Helping [00:14:00] nicely with integrations. Like what? What do you do to make it be useful beyond the demo in a talk that gets people signing up to your product and then wandering off after they've paid you for a few months and it hasn't worked, but you still had them paying for a few months.
Nice. Mm.[00:15:00] [00:16:00]
Yeah.
Yeah, absolutely.
The Evolution of API Tools
---
Phil Sturgeon: I, um, I've, I've been, I've been really enjoying kind of the, as much as I've kind of had this kind of, uh, uh, Luddite in the proper definition of it, uh, approach to ai, um, I just like. I don't hate technology. I hate when it completely replaces labor and, and makes everyone's lives worse, which is what the ludic were about.
Right? Um, I, yeah. Exactly that, that needs to be said more because people just toss that term around like, you ate [00:17:00] computers. I'm like, I've defined my entire fucking life by computers. How dare you. Um, like I even run like a reforestation charity now that uses like a whole bunch of tech. We've literally got an API in the background that shows you where our trees are and like it powers the entire company.
I don't hate tech. I hate assholes doing asshole shit for money with tech. Um, the, I do, i'd, I do occasionally find like a wonderful use of AI that actually brings me joy. So we were speaking on, I think it was the last episode, um, with Tom from Wire Mock and, um, they have a, a, a mocking system, which does a whole bunch of stuff, right?
Like mocking does a million things, but they were talking about how they actually kind of edge into testing and edge into fudge fuzz testing. And the, the cloud offering can like, use AI to create. Useless requests that you can see how your API handles. I was like, that's fucking genius. That's like create your request that tries to do this and like make a bunch of mistakes and like obviously that last part is redundant.
It was gonna do that anyway. That's what it does. [00:18:00] Um, and, and so to use AI slop to be slop. Um, and, and test how your API handles slop is brilliant because the requests that were gonna hit your API for years and years and years, were gonna be wholly constructed by someone that didn't read the docs or didn't know you had docs.
Or you didn't have docs, right. They were gonna be bad requests coming in. Um, but now is just used like this entire swath, this entire treasure trove of the entire internet of, of, of dumb mistakes to. Creatively cock up requests to your API and you can see how your API responds to those. Like, that's obviously brilliant and that's not melting the planet.
Um, so like I do like it when people find good things. I think a lot of the problem has. For me, it's just, it's obviously exactly the same people who were really excited about crypto and that didn't go anywhere and they're really excited about NFTs and that didn't go anywhere, and they got really excited about ai and that's not really, you know, the bubble [00:19:00] hasn't blown yet.
So they're all still really excited about it. But it, it, it is that it's seeing the exact same people, the exact same like snake oil salesman, just keep pushing bollocks. Um, so when someone can go. Well actually there is this very limited use case in which I think it can genuinely, really help. I'm like, brilliant, do it.
I actually quite liked, um, uh, optic, um, so before Optic got bought out by Postman, um, they were working on, oh, it's Atlassian, sorry. Yeah, yeah, yeah, yeah. It was uh, ah, postman bought a similar learning tool, which was Aita. Thank you very much. Yeah, I always confuse those two. Um, but yeah, postman bought Aita.
Atlassian bought optic. Yeah. And, uh, optic had that Lynch, GPT and um, there was a little minute where I was like, this is actually pretty good. Like, I've written some rules that Spectral couldn't handle and it was like, um. Uh, watch out for any fields that have PII in the [00:20:00] response or something like that.
And it would, it would it very inconsistently and, and non determinative, but quite well occasionally spots and fields that I wouldn't have thought of, that I wouldn't have put into a big if field name equals array. Um, in spectral. So there have been some things that I think are pretty cool. Um, have you got, have you found any other little bits like that that, that you think it's really helping with?
Hmm.
Hmm.[00:21:00]
Nice.
Yeah.
Yeah. Yeah.[00:22:00]
Okay. So where's the AI involved in that? So you're crawling, you're crawling through search engines, and then what are you doing? You find an API and you start like generating APIs, Jason, with ai. So it's like, is that, where does the AI get involved?
Hmm.
Yep. Yep.
Hmm.[00:23:00]
Right.
Mm.
Yeah.[00:24:00]
Yep.
Hmm.
Yeah. Okay. Fun. That's pretty nice. Um, I've not, yeah, I've not played around with too many of the specific providers. I mean, like, I think most of my AI interaction is like, I. Vs. Code has just wedged a bunch of it in there, and every now and then it, it makes recommendations at me that are just absolutely horrific.
And then I say, no, [00:25:00] go away. Um, it spends a lot of time trying to help me write. Articles and I'm like, no, my, this, I can, I can make money by typing words. Don't type words for me. I will type better words than you anyway. And, but then sometimes I feel like it's just vomiting entire paragraphs stuff I wrote.
'cause I feel like the open API writing about open API is a pretty niche section of the world. And like it will, it will recommend a paragraph. I'm like, that's my fucking paragraph, man. I wrote that.
Yeah, yeah, yeah, yeah, exactly. Yeah. It started like if you try and write about open API with, uh, copilot on it will, it will use British English and swear a lot. Um, anyway, that's probably enough about ai. I mean, I, I think we talked about it in general, but like what, specifically? What, what is your, uh, startup up to?
What, what are you. What's your plan? Or is it early days? You don't really know.[00:26:00]
Yeah. Okay.
Yeah.[00:27:00]
Yeah.
Hmm.
Yeah.[00:28:00]
Yeah.
Right.
Yeah, that makes a lot of sense. I mean, the. Yeah, there's definitely those companies that are like, yeah, still on soap and things that came before it. And, um, they seven, seven or eight years thinking about switching to a new thing whilst they work through that and, and really spec it out. Like big companies, slow roadmaps, like I think a lot of us forget that that is a thing that exists.
Like the, the thing for me that was always really annoying about GraphQL was that people. [00:29:00] People were like throwing away a rest API they built six months ago, a year ago, two years ago, because they thought they needed to go run off and do GraphQL to do a thing, which was perfectly possible and reasonable and common to do with rest API be it sparse failed sets or compound documents.
Um, and so I always got really frustrated about like, if, if your rest API. After a year or two is so bad, you want to throw it out completely. Redesigning everything in a brand new paradigm that you don't understand yet is not gonna make that better, right? Like it's, it's gonna be worse than what you currently have.
And so I spent a lot of time trying to just get people to pump the brakes a little bit and, and yeah, like there's. Just how, how do you build, how do you design and plan an API that's gonna be useful and usable for more than a year? 'cause I feel like half the people were WeWork's happening. A lot of people went from rest to GraphQL [00:30:00] running straight back to reco.
And actually it was fine over there. I'm just gonna make this one tweak. Or just like adding a Gz compression, uh, G GIP compression header will probably help. Or switching the JSON passer means that we don't need to completely switch to some other GRPC. There were these like small improvements that could be made, and I want people, I spent a lot of time focusing on helping people improve where they're at, but I was always trying to help them improve the rest so they didn't have to run off to something else or rewrite to another rest.
Um, but I, I do like the idea of like. You are there saying, Hey, you are on soap right now. Rest is gonna be a good step forward for you. If you do it like this, please don't run off and just AI generate some slop. Like you've spent eight years thinking about how to build this new rest a p that will solve all your problems.
Go in, make API hot into a prompt, isn't gonna, isn't gonna help. So, um, that's a really funny use case. You talk to different people to me.
Yeah.[00:31:00]
Yep.
Mm.[00:32:00]
Yeah, that is always the way.
The Cycle of API Startups
---
Phil Sturgeon: Um, I think there's also like something we've both been pretty annoyed about is that in all of these, like. Emerging markets. There's an immediate money grab of like cool hot startups with a brilliantly beautiful, uh, marketing page and a snazzy title that are just there to be like, we've solved every problem that you never know existed.
And they come out and they make a bunch of really cool stuff, and then they might settle down and like. By version three or four or five, they're like making some genuinely really useful stuff. And I've seen so many of these companies, like Optic, they, they were on like version nine, uh, within a, within two years or something.
They, they rewrote their stuff so many times. I have no idea how Aiden had time for this, but like, they were making some really useful stuff and, and stoplight, like when I joined, they're on version four of the SaaS platform at least. And they'd rewritten Prism three or four times from Golan to. Type Js, uh, type script.
And it [00:33:00] was getting pretty useful. And like spectral was the third version they'd written. And it was a ripoff of two versions of specky before that. Like you, you kind of get this like quick burst of just like, we're really excited about stuff. How are we gonna make loads of things? Loads of things, loads of things.
And it takes a little while to settle down. It becomes genuinely useful. And then you've got this golden little time where things are genuinely really useful. And then the massive corporates just swing by and, and snatch everything up like. You I, I was writing about API E closing down, apparently them putting a little banner on apiary.
If anyone's not familiar listening, uh, on the podcast. 'cause there are other people here, not just me and you. I keep forgetting this. Um, API E kind of built that a PA blueprint format. It was one of the three main rivals. Be, uh, you know, the open API Ramel, A P reprint. They made a P reprint pretty useful.
Loved it. It was my first introduction to API descriptions and I was a big fan. Um. They made that whole thing. And then that was like 10 years ago. I was working with that. 2016, um, they recently put a [00:34:00] banner up. It was apparently a bug. It wasn't intentional. It was a bug that said like, Hey, we're closing this down in September.
Y'all better shove off. Uh, apparently that was a bug, but it's like no one, no, no one's cat got on the keyboard and typed out that prompt, like they're clearly about to get rid of it, even if it wasn't meant to go live at that point. Um, so yeah, like obviously being bought out by Oracle, they've been closing down the, um, paid pricing.
They've been suggesting people move over to Oracle Cloud. They've been completely failing to write a new feature for almost a decade like these. An api API was amazing. And API Blueprint was amazing and the tool suite was amazing. Dread was amazing. All these things and, and that seems to just be on repeat throughout the API, especially open API world, which is.
Call new startup. They go for a while. They get a few rounds of funding. They get really big, they get really useful, and then some giant boring corporate buys them, merges them into a really unflattering, [00:35:00] uh, platform where you can barely find that functionality. And you've gotta pay five squillion dollars to get anywhere near anything above the free plan that doesn't do anything before you can even start trying to find the functionality that used to be reasonably priced.
How do you feel about that? Why does that keep happening in APIs? Is that just capitalism and we're a bunch of Moy little socialists on the internet or what's going on there?
Yeah. Mm-hmm.[00:36:00]
Hmm.
Yeah.[00:37:00]
Yeah.
Yeah.
Yeah,
I mean, it makes it hard to get excited about new players in the space sometimes, because I. I see new things come out when I'm really, and, and I, I'll, I'll share them around like, um, uh, you know, uh, speakeasy do an incredibly good SDK generation where you don't have to install Java and, and scale it [00:38:00] like recreating most of the kind of open source foundational layers of open API stuff that have just been completely unmaintained for so long.
Like swagger bars are being replaced. Brilliant. They, they've done loads of open source stuff and they're starting to roll out more and more products on top of that. Um, yeah, bump, just being like, here's one single CLI command you can run and your, your docs are all hosted, like these tooling providers are the underdog and they're scrappy, and, and some of them are more early days than others, but like they're going and they're doing really useful stuff.
And then, you know, I'm, I'm writing like, Hey, these things are great. You should go use these things. And then the more that you say, Hey, these are great, you should go use these things, the more the chance of them just getting acquired by an asshole company that fires everyone and like deletes the code.
Increases and it, it feels like, um, I'm just like the gentrification of the API tooling space of just like, it gets good for a little bit and then it's just, and it's completely unaffordable and terrible. [00:39:00] Um, so it's, it's hard to, to be in that cycle, I guess is is all I'm thinking because Stop, go on. I was just gonna say like, stoplight being the main example, like, you know, pouring that, that was, that was, I was an insider.
I, I was working there for a long time. Product manager, project manager. Really, really like making those tools what they were. And then they've just been like, some of it was copied and pasted into, into, uh, swagger, uh, swagger hub. And then a lot of it's just being kind of ditched, right? And so it's hard whether you are working on it, advocating for it.
Being just the person at your company that says, Hey, this thing's cool, we should use that, and then you are the person that's responsible for making this entire company reliant on something that's vanished six months later. It's really hard. I don't know how to feel about any of that. It just makes me angry.
Oh yeah. [00:40:00] Hell yeah.
Oh, let me tell you a story after we've recorded here.
Yeah.[00:41:00]
Yeah.
Yeah, that makes sense. Good things can't last forever, but they can be handy for a while. Um.
Conclusion and Farewell
---
Phil Sturgeon: Cool. I mean, on that note, we are probably about out of time, but I feel like I could talk to you about a million things forever. Um, uh, where, where can people find more of you and what is the name of the startup that you're working on right now?
Uh.
Yeah.[00:42:00]
Nice.
That's brilliant. I'll shove some links in the show notes for everyone listening and, and, uh. Yeah, they can follow along. Thank you so much for joining us on the podcast. It's nuts that we've had however many episodes we've had. I, it'd be too boring to count and I've definitely lost track, but we've been going for years We've never had you on.
That's nuts. So thank you for helping me rectify that.
Thank you very much. Alright. Cheers folks. None of this matters. The true stuff is important. Yeah, this is very [00:43:00] true. Thank you very much.