Slightly Caffeinated

Chris Gmyr and TJ Miller catch up before Thanksgiving, from TJ's Iris caching rework and Sonari soft launch to Arc shutting down for good. They also talk Hive catching on at work, sharing insights on shipping side projects and keeping AI costs in check.

Links
References Mentioned
  • Luma Brighter Learning - where TJ is building features on OpenAI's and Anthropic's managed agents
  • Sonari - TJ's local transcription app, soft launching the day of recording at $29
  • Prism - TJ has some Prism work on the books
  • Great Wolf Lodge Concord - the indoor water park near Charlotte where Chris's family is spending Thanksgiving
  • Kalahari Resorts - TJ's family's water park pick, four trips and counting
  • Iris - TJ's personal AI assistant, cheaper to run after the caching rework and still on intro pricing
  • Wikipedia: Signs of AI writing - the page TJ pointed Opus at to build Iris's humanization layer
  • Polar - handles Sonari's license keys and checkout
  • Avocet - Chris's meeting recorder, in the same "built it for me" boat as Sonari
  • Wispr Flow - one of the competitors Opus wanted a Sonari comparison page against
  • Product Hunt - where TJ is thinking about launching Sonari on Tuesday
  • Hive - Chris's agent orchestration project, now with a 40-person Slack channel at work
  • Halcyon - TJ's notes app, which he is building with Hive
  • Arc - shutting down in December, which sent Chris looking for a new browser
  • Dia - The Browser Company's AI browser that Arc users are being pushed toward
  • Brave - Chris's pick to replace Arc: Chromium, no built-in AI, and familiar extensions and dev tools
  • Zen - TJ's Firefox-based browser since Dia was announced

Creators and Guests

Host
Chris Gmyr
Husband, dad, & grilling aficionado. Loves Laravel & coffee. Staff Engineer @ Rula | TrianglePHP Co-Organizer
Host
TJ Miller
Dreamer ⋅ ADHD advocate ⋅ Laravel astronaut ⋅ Building Prism ⋅ Principal at Geocodio ⋅ Thoughts are mine!

What is Slightly Caffeinated?

Join Chris Gmyr and TJ Miller as they dive into the world of PHP, Laravel, and all things programming, while also sharing insights on family life and other musings.

Chris Gmyr (00:00)
Welcome back to the slightly keffeinated podcast. I'm Chris Gmyr.

TJ (00:03)
I'm TJ Miller.

Chris Gmyr (00:05)
Hey TJ, so what's new in your world the last few weeks?

TJ (00:07)
Oh

man, I don't even know. Like the last few weeks, just everything feels like such a blur. Um, I guess on like updates from previous episodes, my wife's recovery is going really well. She's you know, as far as we understand, like the infection's gone. She's like got her energy back, which is great. So um there's that. rough news, like my

My grandfather's ninety-one and he's like not doing well. so that's that's a bit of a bummer. So um trying to find opportunities to go spend time with him. Cause my my sister and I are the only family he has that are still like in the state. Everyone else has moved away. So like he had uh a couple appointments over the last like two weeks that he needed a to.

So I was I was, happy to have an excuse to hang out with them. so that's that's nice. but other than that, man, just heads down jamming at Luma, getting to work with some new really cool stuff. been building features using open AI and anthropic's managed agents. So like their infrastructure, their sandbox,

really, really cool stuff actually. Um, because like what you can imagine it is like a a remote clawed code session. And so you can give it context, you can give it skills, tools, all sorts of stuff. And then uh yeah, it's super powerful. And I ended up using it for one thing and now I'm using it for another thing. So it's just like now that I'm like acclimated to it, I'm using it a lot.

So that's been cool. And then actually have had some time to work on scenari. You know, I've got some Prism stuff on the books I'd like to get to. So yeah. Just busy busy. How about you, man?

Chris Gmyr (02:08)
Yeah, yeah.

Cool. yeah, just kinda getting ready for holiday stuff. last week we

TJ (02:15)
Yeah.

Chris Gmyr (02:16)
were uh I know I mentioned we were gonna go on like a family camping trip. That unfortunately did not happen. It was like way too cold here and didn't wanna subject ourselves and the kids to that 'cause it was like gonna go down in

the low thirties, like overnight and in the morning. So

TJ (02:33)
Yeah.

Chris Gmyr (02:35)
I don't wanna sign up for this and have it be my daughter's like first camping trip. So we pivoted. Um I still had the day off. we picked up my son early from school and did some stuff with him. I spent the day with my daughter and then we set up the tent locally a fit in the living room. So we set that up on Saturday so they got to like play in it and

put the fireplace on and they brought their sleeping bags down and played and read books and listened to their yodos and stuff like that in there. Uh we did like, you know, our hot dogs and chips for dinner. That was gonna like our super easy like camping meal. So we kinda kept like everything the same. We just did it at home. did some like s'mores outside Saturday night. and just kinda hung out and kept warm, luckily. so

TJ (03:20)
Yeah.

Chris Gmyr (03:21)
Still had fun. It wasn't the full, experience, but we'll try again in the springtime when it gets a little bit warmer.

TJ (03:28)
Yeah, we've been getting into the the 30s here too. I think a couple days this week I've had to scrape a little little ice off the windshield. So um like taking taking the kiddo to school. So yeah, it's we're getting the same thing here, but I'm glad you guys were able to to do something and have some fun with it.

Chris Gmyr (03:45)
Yeah, yeah. but yeah, very similar geminet work and lots of projects at home, which we'll get into. But yeah. anything new for coffee for you the last couple weeks?

TJ (03:57)
Yeah, I unfortunately forgot to grab the brand before I sat down. Um, but we I still have this this damn espresso blend I gotta get through. But my wife ordered like we just had like a we ordered like refill of like stock stuff for around the house and kind of like prepping for the holiday and everything. Um so she ended up ordering like two like bags of

some brand that Costco it's not a it's not one of the Costco brands, but it's just like a brand they're selling. Uh and it's it's like a medium roast. It's super good. so I think anything to avoid this this espresso blend that I have open.

Chris Gmyr (04:38)
Mm-hmm.

TJ (04:39)
Um but it's it's good stuff. So other than that, really nothing new.

Chris Gmyr (04:43)
Yeah, similar for me. Thought I was gonna share some camp coffee, French press stories, but alas, we just made our regular coffee here last weekend since we didn't go camping. I did go crazy with trying to make camp coffee in the house, but yeah, pretty pretty boring over here for coffee. But maybe with travel coming up we'll try some different things out and about, but we'll see.

TJ (05:07)
Yeah, sick, man.

Chris Gmyr (05:10)
Yeah. yeah, what are you guys doing for Thanksgiving plans coming up next week?

TJ (05:15)
Yeah, so with I mean we the last two or three years we have done um both for Christmas and Thanksgiving, we've done meals at this restaurant. it's the food is mid at best.

not not great, nothing to write home about for sure. But you know, because it's just my sister, her husband, the three of us, and my grandfather, like it it it's really convenient because like we're I don't know, like half an hour, forty minutes from his house, and this is kind of like halfway point. So it's easy for him to get to, it's easy for us to get to, and then on top of it, like no one has to clean up or deal with anything.

I can't host because my sister is crazy allergic to cats and dogs, and we have a ton of those. So she even hopped up

Chris Gmyr (06:11)
Yeah.

TJ (06:13)
on allergy meds. She can only last like 15-20 minutes in our house before she's like, I got I gotta get the hell out of here. So

Chris Gmyr (06:20)
Ha ha ha.

TJ (06:22)
we can't host. and

She's even a little bit further south of us. So it's definitely it would be a haul for my grandpa. So we just go there. It's easy, no cleanup. But the downside is like outside of the food being not super great, uh, the downside is especially around Thanksgiving, we don't have any leftovers. So that always sucks because like I

I love having Thanksgiving leftovers. I just I do. So uh this year my wife found uh that Walmart had some like deal for a Thanksgiving kit. It was 35 bucks, but it came with a turkey, like green bean casserole supplies, some like canned corn, stuff for gravy, uh

Potatoes, rolls, it everything, everything you'd need for a setup. And uh

Chris Gmyr (07:20)
Yeah.

TJ (07:21)
it ended up like we went and picked it up the other day, and it was like a nice butterball turkey, like actually, so like overall, a pretty solid deal. And the nice thing is like it was like everything all together. You know, we pick up the order, now we're set to go. So we're gonna, I don't know if we're gonna do it day of or day after, but we're gonna cook.

our own meal so that we've got plenty of leftovers for us to do. Um

Chris Gmyr (07:47)
Yeah, nice.

TJ (07:48)
so yeah that'll be that'll be pretty chill. Um but yeah just you know especially with everything that's been going on with my grandpa like just the the focus is spend time with him. So yeah dude look he's

Chris Gmyr (07:59)
Yeah, a hundred percent. Yeah, spend time where you can, especially once those medical issues start piling up and just inevitable.

TJ (08:08)
he's 91

He coached baseball for years and years and years and years. And um with that, he's developed skin cancer and uh he just went through like six weeks of radiation that didn't work. So it's it's a rough situation. So like we're you know

That that's all all of Thanksgiving. He's like, he's not even sure if he's gonna be able to make it to the reservation um at the restaurant. But we're like, well, no matter what, we'll we'll figure out something and we'll come hang out with you at least for the day. So yep.

Chris Gmyr (08:45)
Thank you.

TJ (08:46)
Yeah. How about you, man? What are you what are you looking at for your Thanksgiving plans?

Chris Gmyr (08:50)
Yeah, we're changing it up a little bit this year. Usually we're the travelers kinda going around to all the different places. but this year we just wanted to try and do something as just our little family four and do something kind of fun. So we are going to go over to the Great Wolf Lodge in Charlotte, which is about two hours away, because they're actually open on holidays. So we're

Uh Great Wolf Lodge is like an indoor and outdoor water like theme park and it has like arcades and laser tag and all this like putt putt and fun stuff and I don't know, like 10 restaurants on site. So we're going there from Wednesday morning to Friday morning. Um, and then they have like a big like Thanksgiving buffet type of thing on Thursday. So like kind of very similar is like baked in entertainment food.

Don't have to do dishes, don't have to do anything. Uh we got a pretty cool room that has like like a themed like forest, like bunk room for the kids to be in. so and then your ticket is just for like the whole thing. So as much as you want to go to the like the water park, putt-putt, all the different activities, laser tag, stuff like that. So yeah, just getting out of the house and doing something a little bit different with just us and hopefully everyone will have a blast.

ended up taking the full week off again. I usually try and do that. So have a couple days leading up to where my son is still in school and do some things around the house, do some prep, maybe do some like early Christmas shopping or something like that. and then yeah, go to the Great Wolf Lodge and come back on Friday and have the day and the weekend to kind of do some stuff around the house again and relax a little bit. recoup and yeah, back to work after that.

TJ (10:33)
Yeah, dude,

that's sick. I've we've not done Great Wolf. We've we've thought about it. we've always just done the Kalahari. Uh and there's like a Great Wolf like down not too far from the Kalahari. And there's also another Great Wolf, I think, up here in Michigan too. But we've just we've talked about it, never, never got around to it. So that sounds really cool, man. That'll be a total blast.

Chris Gmyr (10:57)
Yeah, I'll report back

'cause we haven't been there yet. We've heard a bunch about it. and a lot of people really like it. So we'll see how it goes and see how their uh holiday spread is. And coffee.

TJ (11:06)
Yeah, yeah. we've

done we've done Kalahari, I think four times. And my son's been five times because my sister took him to like a Comic-Con convention um down there. So that he's he's been a time extra from us. But we've every time we've gone, we've had a blast. We've done it in the winter, so like you have all the indoor stuff.

We've done it in the summer, so you have all the outdoor stuff too. yeah, just the water park, those those water parks are super fun, especially for I think like the the younger kids because they're like themed. I think the theme room is gonna be a blast, so that'll be

Chris Gmyr (11:44)
Yep.

TJ (11:44)
that'll be really cool, man.

Chris Gmyr (11:46)
Yeah, yeah. Looking forward to it. So yeah, wanna chat and give us some updates on Sonari and Iris and some of the other things that you've been working on behind the scenes.

TJ (11:55)
Yeah, so I fell off using Iris like a while ago. It just it got expensive and um you know it really that's that's what happened. Like it just got really expensive, especially with some of the experiments I was running around like giving Iris its own internal journal to like have like its own.

thoughts and and things and like goals that it wanted to have. So it just like those background processes like got super expensive. Um it was really, really interesting experiments and I think was really promising, you know, in in an attempt to give Iris more agency. Um so it was really interesting, but like, yeah, it just got really expensive. Like I think when I kind of fell off

My wife got upset with me because I think I burned through basically $300 in tokens in like a week and a half. And so she was like, dude, you can't be doing this. So Yeah.

Chris Gmyr (13:00)
Yeah. Which is really easy to do with that system or uh really any mid to large size system of any kind using API costs like a a

crazy amount of money.

TJ (13:12)
Yeah, and it was, it was still using Sonnet for six too, so it wasn't even using Opus or anything, but just it was like a lot of background processes and then on top of chat and yeah, it yeah, it just got super expensive. So at some point I didn't realize this, but at some point I realized with my ultra plan, I also get

two hundred bucks a month in API credits. So once I saw that, like I was like, shit, this is like an opportunity desires again. Like I just paying for it. So like this works out great.

Chris Gmyr (13:52)
Yeah, I think they put it out probably like a month or two or six weeks ago or something like that. It like wasn't too long ago, but I mean it was it was a little bit. And I don't think it was I know they sent out some like emails about it, but if I don't know, you had to kind of like read through the lines, you had to go in and click a couple buttons to connect your like consumer.

org to the API uh platform org and then that's how you got the the free credits that matched your subscription uh amount each month. but yeah it was probably a little while ago but that was super cool that they did that and I mean hopefully

TJ (14:26)
Yeah. Yeah.

Chris Gmyr (14:31)
two hundred dollars goes a little bit further.

TJ (14:33)
And I had I had no idea. Like I don't even I don't even remember how I stumbled. I it was I stumbled across it. So like I don't even know how I did. Like but I stumbled across it and I was like, this is great. Now I like

I I have the means to get back into using Iris. But with that, like I knew that I was probably leaving some sort of like token efficiency on the table. Like

I felt like that I could be more efficient with token use, especially on the caching side of things. So I sat down with Opus, we did some digging into the code base. And um the way I had the cache breakpoint set up for the system prompt, basically like the the system prompts and everything, there was like stacking based on what pieces were dynamic. So like it went the the cache breakpoints went from the most static to

least static so that you could have as much stacking of these breakpoints, uh these caching breakpoints as you could. But that meant that all of the message context that you're sending was uncached because like all your cache breakpoints were used in the system prompt and like the messages come after that. So what we ended up doing is like flipping it all around.

So that, the system prompt stuff is mostly static. Um, there's like maybe I think like pinned skills and pinned prompts could break the cache in system prompting, but other than that, it should be fairly static. and then all of the messages now are able to get cached. So like your message history, except for the most recent two messages.

Are now part of that cash stream. And so now you're building up like this cache, you're getting super heavy cash reuse across the system prompts and all of the message history. And what we ended up doing is whenever you send a message to Iris, on send, like when it goes to Claude, like Anthropic, through the API, uh, we end up having a like context block.

That includes all of the dynamic stuff. Since you can't cache your input message, anyways, we just send all the dynamic stuff along with that. So any matched memories, any matched truths, any sort of like other dynamic things just get sent along in your message content inside of like some XML tags to like kind of signify that like this is additional context, not user message.

And that means that everything before the message you send is able to get cashed. So we had some massive cash wins there. So not only did we get the credits, but I also paired that with some big cashing changes. So we're way more token efficient now. Uh I also went through and like did some other tweaks on things to to make some improvements. So no

No new features, but lots of fixes for that. And then um did some additional like I found some bugs with memories, so I ironed those out. And what was the other big thing? Um

Yeah, we handled some memory and and bugs and memory improvement. and I also did some like humanization research. So Iris's responses feel way more human now. it they're it was falling into the same trap that like Cloud Code responses had for a while, where it's just it screams AI, it's like

resummarizing things all the time and yeah it just it felt it didn't feel super natural. So I pointed Opus at the Wikipedia uh entry for like a signs of AI writing and I had it basically go into the prompting and like add a humanization layer to it. So the responses feel way better now.

But I was also using Sonnet four five, and that's end of life now ish. Um,

Chris Gmyr (18:57)
Mm-hmm.

TJ (18:58)
so I also um also upgraded to Sonnet five five for my models and the default models for iris. So lots of lots of work has gone into iris lately. I'm using it all day every day. Um, so I I'm getting a ton of use out of it, which is where all these like

improvements are coming from. So yeah, Iris is still available for sale. And like that was a funny thing. I was setting things up for Sonori and found like five sales that I didn't realize had happened. So yeah, that was that was a fun find. Um so that's still available. Like I'm still doing intro pricing on it. So if anyone's interested, like you can definitely check that out. but yeah I've been since I've been using it a lot, I've been putting a lot of work back into it. I don't know if I have any ideas for

for net new features, but I'm definitely keeping an eye out for room, like room for improvement in the system. but yeah, it's in a I think a really strong state at the moment. And then

Chris Gmyr (19:55)
Nice.

TJ (19:56)
yeah, on the Sonori front, like I'm now that things with my wife's health has stabilized up until her next surgery, which we don't know when that's gonna be yet. That's probably like a month away, but

Now that she's been stabilized, my workload has been easing up around the house and caretaker duties. So I've had some time to put into Sonori. Um, I've made a few improvements on things, but it's in a really good state. I've got everything with Polar set up. actually, like today is we're recording on a Thursday. Um, I'm probably gonna have it.

available as a soft launch later today. So like I got the license key stuff all figured out, got the update pipeline figured out. Um all the updates on the website. So like it'll probably be like soft launch today. So people want to go buy it, they should be able to buy it today. And then I think I'm gonna go like I'll email the list so anybody that signed up can get in on it now and then

I think Tuesday I'm gonna go public with the launch. So that'll be that'll be pretty exciting. yeah, um

Chris Gmyr (21:05)
Nice.

TJ (21:07)
even with that, like I don't know if I have like new features in mind. Um, but I'm I'm pretty stoked to get it out there. It works great as is. Like I use it I use it every day at work. Um so

Chris Gmyr (21:21)
Yeah. Is it a bunch of.

TJ (21:23)
yeah.

So it's it's in a really good spot. I'm super excited to get it out there. with launching anything, especially something paid for, there's some baseline anxiety, works well for me. I know you've been using it. but you know, I just like I don't know, I'm anxious and nervous to put myself out there and 'cause I want to go a little bit wider with this than Iris. Like Iris I is very niche, I think. and it was

It is very like developer focused. You need to know how to set it up. You can put it on cloud if you wanted to, but and everything, but like it's still like it's it's a developer kind of thing. Um, whereas scenarios definitely a little more consumer, so I want to go a little wider with it. Like I was talking to my therapist about it, and she's like, man, I would love to have this because like I would love to transcribe my session notes, like but

She's like a hippa and all this other stuff. And I'm like, yeah, it doesn't leave your system. And so she's like, it's great. It's perfect. Like, so I think there's like a bit of a broader market for it. so I'd like to I th I I think this will go on I think I might put it on product hunt as well as posting on socials, like to just go a little broader with it, which makes me nervous, but um

We'll see. We'll see how it pans out. I think I'm not expecting to have a huge windfall from it or anything. Um, but I'd I'd love to have some sales. You know, I'd love to have people use it and be useful. Um, but I'm not I'm not expecting anything crazy out of it.

Chris Gmyr (22:57)
Yeah. Same thoughts over here with like AvaSet and some of the other projects. It's like built it for me and if other people get some use and utility out of it, that's great. If make a couple bucks off of it, that's also great. But I mean

TJ (23:12)
Yeah.

Chris Gmyr (23:12)
like I'm just using it, you know, daily, you know, where I can, where it makes sense. And I don't know, just putting it out there. it's more of like a I don't know, practice to do it.

TJ (23:22)
Yep.

Chris Gmyr (23:23)
Some other skills and things like that. I

don't feel like I'm doing a great job on the marketing side right now because I tried doing um a few tweets like as I was kind of like building things out, like kind of trial and error, you know, kind of things that I was learning uh as I was going. But it's like, I don't know. It's so hard to keep up with the social, you know, aspect of it or like social aspect of it,

TJ (23:45)
Yeah.

Chris Gmyr (23:47)
you know, just posting online. Um

'Cause it seems like I don't know, the the algorithms are completely different where it doesn't really care as much about, those, uh, versus mm, just like a a random, you know, message about something else. I don't know. It's hard to put my finger on, but because like I'm not that like active on it, I think it also adjusts the algorithm, you know, as well. And just like my first thought is not to always just post the social if I'm like working through something or to

share or promote, I'm just I don't know. I just put my head down and keep on going. You know? So it's it's like a a different muscle to build cause like on the other side I'm like, d would anyone really care about this? Like I don't know. It seems silly to me to to post and share like that, but I guess that's how you get it out too.

TJ (24:37)
Yeah. Yeah. I've

I've noticed the same thing as I was like ramping up to like promote Sonari early on. Like I did some you know, I I only know the like developer way of doing things, right? So I was like posting I had a couple like a blog post or two about Sonari and like building it out that really didn't get much traffic. and then like I don't know, my reach

My my social reach doesn't feel like what it used to be, even though I was like being pretty dedicated about it. It just it wasn't really going anywhere. So I I kind of fell off. But I've been I've been kind of stingy with social stuff. I just I've been I've been buried with life. So I've not been super active on anything anyways. but we'll see. Like I I'd I'd love to have some sales.

I've priced it very aggressively. So I think I think it should be like an easy it's a low barrier to entry. I th I'm doing like twenty-nine dollars for it. So like, you know,

Chris Gmyr (25:36)
Yeah, stoler.

TJ (25:38)
I I think that's like a an easy amount to spend on it. It works really well. so we'll see where it goes. like I said, I'd like to have some sales, but I'm not I'm keeping my expectations low.

Chris Gmyr (25:48)
Yeah. Yeah. yeah, this might be like a good thing to continue to circle back on 'cause like as you were talking about other like avenues and even like your therapist possibly like interested in it. It's like, well, very similar to Avicet and meeting recordings. And it's like,

TJ (26:02)
Yeah.

Chris Gmyr (26:04)
Well, okay, I really need to do a lot more research on like what the current state of like online marketing is and the best avenues to go. Is it

Do we still set up dedicated pages for therapists, for doctors, for counselors, for like is that, you know, the best way to go? Is there a different avenue to go? Do we do paid, you know, Google ads or Facebook or Meta or whatever? You know, is that still the best way to spend, you know, that sort of money? I'm sure a lot depends on the app and the

TJ (26:35)
So Chris lagged out a little bit here. I think we were were we left off talking about like marketing with this stuff. And I think it is I really don't know how to do product marketing. I've only ever done like developer centric stuff like.

I don't think I did a good job with Iris either, but like Prism did a lot of marketing on its own and that's open source anyways. So yeah, and I don't know, like I, you know, you had mentioned like AdWords and stuff. I don't even know. That world is so foreign to me. And talking with Opus about marketing for Scenari, it was adding, like it wanted to add product comparison pages.

So it wanted to add it with, against whisper flow with a couple other ones. And, um, so I think it was still like reaching for that, but I don't know how valid that is. And I think that was interesting too, about like putting together marketing specific pages for like physicians or therapists or stuff like that, that I think that's like SEO. I think that's like SEO marketing, right? Of like having those pages that

that rank for those things. at least AI is thinking that those are still valid routes to go. just have, it's so foreign to me, man. I have no idea.

Chris Gmyr (27:54)
Yeah, I used to do it a bunch for other clients, back in the day when I was like building out websites and uh apps and services for people. But I mean, that has been years, years now. So like I it'd be basically like starting fresh, see what's out there and see what's working. And like we know, you know, because of the models are are trained on, you know, older, you know, data timelines, technology, things like that, the newest

latest thing is probably not baked into the models unless it does like active like live research on things. but even that's probably not a hundred percent accurate either. It's probably a good place to start, but yeah, just have to do some like exploratory research into it and see, you know, what makes sense and also not break the bank for something that might not pay off either if you do go like paid ads.

TJ (28:49)
Yeah, yeah, and so I yeah, I don't even know man. We'll we'll see how it goes. I think I'll I'll definitely like on Tuesday when I go loud with it, I'll post on like LinkedIn. I'll post on Twitter and

no, uh, like the, like the new Dev Null, social network. I like, I'll post on those places. I'll, I've been considering posting on Product Hunt. I've never done that before, so we'll see, we'll see if and how that pans out. and then we'll just kind of see what happens. Like I, I really don't know next steps beyond that. but yeah.

I don't know. We'll see. Yeah. Yeah.

Chris Gmyr (29:31)
Well, we'll circle back and hopefully we can do a little bit of marketing on our respective projects and get

the a little bit of income on the side. At least pay for like hosting and some of the tokens being burned and, you know, kinda go from there.

TJ (29:44)
Yeah, at the very least, I built scenario because I wanted it, right? There's other, there's other local solutions out there. There's even open source options, but none of them really had the features that I wanted it to have. And so people out there are like bound to have similar opinions on things or looking for similar features. So we'll see. You know, I think that's why I priced it so aggressively and we'll see from there, but.

I know you had had some rollout with Hive, I think, at work, and I would love to hear how that all panned out. Okay.

Chris Gmyr (30:21)
Yeah, so this has been pretty exciting the last couple of weeks. Um, so I think I mentioned before that um I was doing a tech talk. So we usually have like Tuesday afternoon tech talks, um, either like full talks or like lightning talks, things like that. So a couple of weeks ago I did a full tech talk on

Hive and just kind of like a show and tell of like what I built, the

some of the top features in it, my day to day workflow, at work and also some comparison to what I use at home. And it went over really well. Everyone was like super stoked about it. I ended up setting up a separate Slack channel at work so people could like ask questions or get the scripts that I used to like install it. Cause installing it on work, we have like a whole bunch of like AI

guardrails and settings and all this other stuff. So the install path is a little bit different than what's actually on the GitHub repo. So everyone can stay in sync over there. So now there's like 40 people in the Slack channel, which is pretty crazy. Um, and like asking questions and actually using it. Um I got a couple new ideas for like tweaks and changes and little

TJ (31:25)
Yeah.

Chris Gmyr (31:33)
features and stuff like that that I've already put into latest versions of Hive.

and it's just really cool to talk to people at work about it and everyone's like, this is this thing is amazing. It's like so cool. It just kind of works, you know. And it's like, yeah, that's the whole point of you know, building it and just wanted to keep it easy and just something that works. and I think once people kind of grasp like kind of where you start and then what's a possible path in front of you, because the tech talk was a little bit more of like a a feature.

dump and just a a really quick overview of everything, which is a little overwhelming. But if you like just use it, even the simple profile and using uh like a a run book and a board pad and to-dos for like a a simple project, like that's gonna get you pretty far, you know, as it is. And then as you want a little bit more, as you expand to more projects and different types of projects and eventually like going faster and

working with your own profiles and things like that. Then you can spread off into everything else that's in IVE. But it's been like awesome to see it kind of bloom, you know, at work and even get like feature requests and you know, a bunch of questions and stuff like that. So yeah, it's been pretty cool. And yeah, just unofficially supporting it at work and just tinkering with it, you know, still like on the side when I'm done. so yeah, just

Super excited about it and uh just was awesome to get that much interest in it in the work setting too.

TJ (33:07)
Yeah, I've been trying out some of the new stuff and I've been really happy with it. It's I've been also doing a little bit of work on Halcyon stuff. And so I had it I have it spinned up like running on that. And that's been fantastic. I I didn't spin it up at work and I really wish I would have because

It would have been really useful, but like I had a basic prompt for a little skill I created of just like lead agent where it kind of just gives it some instructions around spinning up like sub agents to do the work and like how to verify and validate things and who's in charge of doing what. And so I use that, but I really wish like it's that's great for like basic stuff, but Hive really shines at the level of work that I was doing there. And so I wish

I wish I would have had high of two on that first, but yeah, I think the project's going great. you'd sent me a couple screenshots of the slides. The deck looked fantastic. So I'm so stoked to hear that it went over well and there's been, it sounds like pretty solid adoption. I think it's great. I've loved it. think it, you know, there's, I've seen a lot of discussion across everywhere.

about using sub agents and all of this stuff. And I think Anthropic recently released something around advisory mode with Fable, which is kind of, kind of what Hive does. But Hive does it

Chris Gmyr (34:28)
Mm-hmm.

TJ (34:29)
so much better. And so I think there's just, even though there's lots of features and lots of things with Hive, I think the mental model is pretty simple.

You have the lead agent and then you have sub agents that are doing all the work. And like, as long as you go into it with that mental model. Yeah. I think all the basic presets and everything is it gives you a huge bang for your buck right off the rip. And then as you start wanting to customize things, it's very customizable. You know, it works. It works on small stuff, works on big stuff, but like

The mental model is so easy to fit in your head about how the basics work that, yeah, it's super cool project, man. I'm so tickled that you're getting like adoption and people are engaged.

Chris Gmyr (35:18)
Thanks. Yeah. I'm super excited about it. So I'm just been yeah, just really like amped up to use it. Like I use it every day at work, at home. So like I'm trying to, you know, surface like any little you know, issue or tweak that I see. Like I'm just trying to go in and hive and fix it and, you know, push that out as soon as I can. and especially working with like the different, you know, systems and a little bit different setup gives me like a little bit more breath of

experience and potential for issues and tweaks that I can fix automatically where if I can push them out and basically fix things before anyone else really notices them. That's fantastic as well. so yeah just trying to focus on that uh

TJ (36:01)
you

Chris Gmyr (36:01)
at least with Hive and then you know using it on all the other projects across the board. So it's been super nice. So yeah, it's just a fantastic project. I'm super happy with it where it is right now.

TJ (36:11)
Yeah, dude, that's great. Like I said, I love it. I'm excited to see where it continues ahead. yeah, just good work,

Chris Gmyr (36:20)
Thanks.

And let's see, one little quick topic to wrap up for today is that I don't know if you saw the announcement, but maybe like a month ago, ARK sent out an email blast to everyone saying that ARK is gonna be no more in December. I don't remember the exact date, but everything is gonna be turned off. Like Arksync is gonna go away, you can't use it, you can't install it.

I think the the basic browser functionality might still be live, but basically like all the back end arc stuff is just not gonna be available anymore. And they wanted everyone to go over to their right, the Diab browser, their I like

TJ (37:00)
Yeah.

Chris Gmyr (37:01)
arc like thing. And so brought up this whole, you know, decision about moving browsers again, which is a pain. Uh don't love it. but did a bunch of research. I definitely didn't want

Anything with AI built into it because I don't know, as much as we like and talk about AI, I don't want it built into everything. I don't want to just magically coming up with things. Like I just want to go to websites and do my thing and not have to worry about AI stuff or privacy or stuff,

TJ (37:31)
Mm-hmm.

Chris Gmyr (37:32)
you know, fill in in the gaps. I just kind of wanna use it like as is, the old fashioned way, you know, and I'll put AI in it or bring

certain resources or whatever I need to do into the AI systems that I already have. so I looked at a couple, ended up with Brave. So I've been using that for at least a month or so now. So I was it Arc also made it w like really hard to export anything like because it was just tabs. It wasn't using uh Chromium's bookmark feature.

I had to ask Claude to basically go into the the app library thing and pull that out and make JSON and HTML files that I could bring into uh Brave to import them. So once I got through that, set up all the different profiles again, um, for like work inside projects and you know things like that. So yeah, it's been working really well. I wanted to stick to Chromium so I could use all the same uh

extensions and plugins and you know, bookmarks work similarly and just the the Chrome like dev tools and all that. so I didn't want like a huge change to go over to uh Zen or you know some of the other ones. so yeah, pretty pretty happy with Brave so far and I don't know, it just feels like Chrome, but a little bit better anyways. so yeah and the syncing works pretty well. So I can sync, you know, bookmarks across

multiple systems and the only thing that I haven't loved is ARC and ARC Mobile. I was able to get to all of my like bookmarks and things and mobile. I used it like that often, but it's really nice to go into, you know, personal projects profile

TJ (39:03)
Mm-hmm.

Chris Gmyr (39:10)
and click, you know, that link instead of just like typing it in. So now Brave only allows you to sync one profile.

but it's just connected to my personal profile and that's been fine for now. So that's the only thing that I really miss, but you know, it's a small miss. so yeah. Bye Arc and hello brave.

TJ (39:29)
Yeah, I was sad to see it go because I thought Arc was massively transformational in the browser space. It introduced vertical tabs, which everyone's starting to do now.

Chris Gmyr (39:42)
Mm-hmm.

TJ (39:42)
I think even Safari is doing vertical tabs now.

I think it had a really big impact in the browser space and I loved it. But as soon as they spun up Dia, I saw this was going to happen. And so I switched

Chris Gmyr (39:55)
yeah.

TJ (39:55)
at that point. So I moved over to Zen. Zen early days was definitely rough around the edges.

are clunky pieces, things that didn't work. But since then, it's become really refined. It works great. So I've been all in on Zen. There's, I do miss Chrome DevTools. I think Chrome DevTools are better. So I do have like Chrome also still installed for when I need to load that. But I don't want AI in my browsers. I want the browser to be

easily controllable by AI, right? I want to bring my own agent to interact the browser with me so I can log into something and then say, all right, now operate on this. I don't necessarily want that to be part of my browser. I want MCP tools or whatever so that I can bring my own agents into the browser with me. So that's kind of my stance on AI with browsers, but I've used

Chris Gmyr (40:50)
Mm-hmm.

TJ (40:51)
Brave in the past.

I liked it. I'm so locked into Zen right now. I love it. It's got everything I needed. It's got like spaces and containers and you know, it's I've loved it, but I was never heavily invested in Chrome and like used all of that stuff. So, you know, it wasn't a huge

I didn't really lose anything going to Zen and the Firefox ecosystem. ah

Chris Gmyr (41:17)
Yeah.

TJ (41:18)
So yeah, I'm all in there. But Brave is a fantastic browser. Like I said, I've used it before. know, it's nice to be. There are nice things about Chrome, too. Like I said, especially the dev tools. So yeah, I think it's a great choice.

I'm loving Zen. think Ian Landsman was also on the hunt for browsers. I think I saw that on X2. And I had suggested Zen and he's like, I'm out using some like some small, some small browser and they having to deal with like security and stuff like that. So I think he was excited that about some of the updates in Safari. But yeah, I've loved it. But it had a rough start for sure.

Chris Gmyr (41:59)
Yeah. Yeah. Well yeah. I know yeah, we we talked about browsers a while ago and especially when LinkedIn came out. So and then seeing like the updates and just being like basically just back end Chromium updates for I don't know how long it's been a year, year and a half now. It's like, yeah, this is this is a limited timeline, you know, anyway. So kind of sad, but I tried to push it like as long as possible. I just didn't want to push

TJ (42:16)
Yeah.

Chris Gmyr (42:27)
it.

To the last possible day. So I think

TJ (42:29)
No, yeah.

Chris Gmyr (42:30)
maybe like a week after I got the announcement, you know, email blast. I'm like, okay, now it's time to dig in, try a couple, and then ended up on Brave, because

TJ (42:36)
Mm-hmm.

Chris Gmyr (42:39)
that's what I don't know, felt the the easiest to move over in the most familiar, you know, across a handful of you know features. So like I said, so far, so good. maybe I'll tinker with more, you know, in the future, but right now I just need something that works and that's not gonna

TJ (42:52)
Yeah.

Right. Yeah, no Brave is going to be around. I think like Zen's got kind of a similar thing. so same with underlying Firefox. There's a lot of security affordances, which feels really good, especially in this day and age where we're getting slapped with vulnerabilities everywhere, it seems. So picking a security first and

first browser is a smart choice. Cool.

Chris Gmyr (43:18)
Yeah. Yeah, totally. Sweet. So on that note, wanna wrap up?

TJ (43:22)
Yeah, let's wrap up.

Chris Gmyr (43:24)
Cool. Thanks for listening to the Slately Caffeinated podcasts. Show notes and all the links and social channels are down below and also available at Slately Caffeinated dot Fm. If you have any questions for us or a content suggestion, go to the ask a question page on our site and we'll feature it in an upcoming episode. Thank you all for listening and we'll catch you all in a few weeks.

TJ (43:43)
Yeah, thank you.