The Triple, Newsletters, and the Claude Rollercoaster
Chris Gmyr (00:00)
Hey, welcome back to the Slightly Caffeinated podcast. I'm Chris Gmyr. Hey TJ, so what is new in your world?
TJ (00:03)
I'm TJ Miller.
man, it has been just...
It has been a lot over the last week, to be honest. A lot of just like interpersonal, struggle-y stuff and then still struggling with getting Claude to do some stuff at work. And so I've had to go old school and manually write a bunch of code and shift a bunch of things around because it just cannot.
I don't know why it can't, but it just cannot do what I'm asking it to do. So I've tried doing things like breaking it down into smaller stuff, more discrete tasks, but it's still just like choking. So it's been back to writing manual code for the last week or so. And gosh, I feel so behind on Prism.
an overwhelming amount of like new issues and PRs and stuff. I've been meaning to get to you, but just I am in the midst of making lemonade out of all the lemons life is hurling at me at the moment.
Chris Gmyr (01:06)
Yeah,
yeah, totally. Can only do what you can do, right?
TJ (01:10)
Yeah. And you know, I posted something on Twitter the other day about it and the feedback and support has been just incredible with people like, you know what, take your time, get healthy, you know, and, you know, people can wait for their, their free project and stuff. So it's not like anyone's like demanding me to.
be doing stuff. I just know that I'm feeling very behind on things. we can talk about some updates a little bit later. But I did build something kind of cool with it. It's just kind of like an experiment for an upcoming thing. So I'm not doing nothing on it, but I'm not necessarily doing a great job with open source maintenance at
Chris Gmyr (01:48)
That's right, it ebbs and flows and like I said, you got a bunch of other stuff going on so people can wait for their free updates.
TJ (01:54)
Yeah, it's, it's, and
it's just me. Like it's just me working on Prism. So, you know, kind of the, the woe is of a single person project too. So.
But yeah, man, what's new with you?
Chris Gmyr (02:04)
yeah, not too much. that's crazy or anything. just been busy at work. ⁓ like I mentioned last week, ⁓ scouts started up. So, had our first den meeting last week, have a packed meeting tonight and, know, just more, more meetings, in future weeks. So, ⁓ I did do some planning with other, den leaders, cause I'm still the main one, but I got two, assistants this year. So.
TJ (02:30)
nice.
Chris Gmyr (02:30)
before
the school year started up, we got together and planned out the majority of the school year, because the Scout calendar basically follows the traditional school year, basically off for the majority of the summer. ⁓ So instead of me trying to figure out what we're doing a couple days out, it's a lot better that I have some extra helping hands with that, some extra brains and whatever people need to get for.
TJ (02:43)
Mm-hmm.
Chris Gmyr (02:55)
know, supplies or resources for those meetings. So a little bit off my plate, but still have to coordinate kind of all the things, ⁓ which isn't as bad as doing all the things. So yeah, not too bad, but just kind of tune my thing. Got a couple of new projects spinning up. So if we don't get a chance to talk about those today, can always chat about those next week, because I'll have more information on that as well as the next week or so goes on.
But yeah, nothing too crazy or out of the ordinary. Just kind of chugging along, doing my thing.
TJ (03:27)
Yeah, that's sick that you
were able to kind of get the year planned out. I'm sure that's like a big relief. I know if I was in that position, I'd be planning stuff like a week ahead of time. So yeah, heaven.
Chris Gmyr (03:40)
Yeah, and that's what I did
like the last couple of years and it's stressful. It's like, you know, Sunday afternoon, I'll add something to my to do list of like plan, then meeting and like when be able to like get to it, like can't do it Monday because I don't know, busy at work or whatever. And then it gets down to like Wednesday. It's like, I still need to figure out what we're doing tomorrow because we make usually Thursday nights or like over the weekend occasionally. So yeah, it was definitely like the stressful weeks of meetings. So hopefully it'll.
TJ (03:44)
Yeah.
Mm-hmm.
Yep.
Chris Gmyr (04:08)
just be a little bit easier this year.
TJ (04:11)
Yeah, that's sick, Cool, dude. You want to talk some coffee?
Chris Gmyr (04:14)
Yeah, let's talk about coffee. I heard that you're not having a great experience today.
TJ (04:17)
Hi
No. so I had my regular like French press this morning, ran out of the stuff I was drinking last week, which was the counterculture, hologram. And this week we were on some clearance blue bottle coffee. So we were, just walking past the like grocery store clearance and they had it on sale for like half off. And, it was.
Chris Gmyr (04:36)
Nice.
TJ (04:44)
No, it's like, you know, normally when you see that you go and look at the like, roast date and it's like getting pretty old and the roast date on it was like actually like fairly recent for store bought stuff. went ahead and got that and busted into that this morning for the French press and that was great. And so normally what I do, is I'll drink like French press, like hot coffee, and then I'll follow that with, instead of like making another French press.
I'll just do like some iced coffee. And since we're kind of out of summer season a little bit, I've stopped like making my own cold brew. so I went and bought some store bought stuff and it's not the brand I normally bought. I normally buy, but it was on sale. So I went with it and, ⁓ it is awful. This is, this is some pretty horrible cold brew.
Chris Gmyr (05:29)
That is unfortunate.
TJ (05:30)
Yeah, super unfortunate because I have a whole bottle of it and it needs to be gone through. So I'm going to be stuck drinking it for the next couple of days. and it is just so bad. and I feel like I've had this brand before and I think I just got like a bunk bottle or something because like, it wasn't, I don't remember it being this bad, but gosh, it's horrible. It's like,
brush your teeth and then drink it. Like it's like toothpaste level bad, like toothpaste aftertaste levels of like bad. Like it's just forcing myself to get through it because I want the caffeine and I need to get through the coffee, but.
Chris Gmyr (06:03)
Sounds terrible.
Yeah, are there like, is there anything that you can put in it? Like I know you talked about putting like cinnamon and some like hot apple cider. I wonder if like some cinnamon in there or like doing some sort of like homemade like cold foam or making a syrup or something like that. Just adding something to take the yuck out of there would be beneficial.
TJ (06:29)
no, I've got,
I think I'm down to like a third of the bottle. So I'm just going to muscle through it tomorrow morning and just finish it off. Just plug my nose and down the hatch, I guess.
Chris Gmyr (06:34)
Yeah, might as well just finish it off.
Yep, that's all right. Just be done with that. That's a good plan, too. Yep. That's unfortunate. Hopefully it was just a bad batch or bottle that you got and I don't know.
TJ (06:43)
Yeah, just be done with it and like never again.
Yeah, I got to stick with the store-bought brand I normally go with and just, yeah, forego any on-sale shenanigans and just stick with what I know is at least halfway decent.
Chris Gmyr (07:03)
Yep, yep, totally. Yeah, I'm still on the same beans. Just about done with the ones that I had mentioned last week. And then I a new bag that will probably be starting in the next couple of days. So hopefully I'll have an update for some new beans next week. But as far as like cold brew, we don't get them a lot. But usually when we like travel or have a trip or vacation or something like that for the road, we usually get these, the column.
cold brews usually in the small cans and I like the the mochas and the triples Super tasty they do have like a little bit of sugar and it's like Not great, but it's fine for The car ride because they are like more of like a mixed latte drink and the triple
Also, we'll give you a little bit more of a punch because it has three shots of espresso in there. So those are the ones that I prefer. But they're good, especially for like once in a while and they're pretty tasty and easy to drink and easy to throw in a cooler or the car.
TJ (08:04)
Yeah, definitely, I just want to look them up. cause I feel like I've had them before and I've definitely had them before. also in like road trip scenarios or, you know, like travel kind of situations where just having like a few cans or something makes, makes life pretty easy. And I definitely prefer the triple. and I've done the mocha before as well. I'm not a big vanilla fan, so I've not done that one, but,
definitely prefer the triple. And yeah, it's like these little like canned lattes. ⁓
Chris Gmyr (08:32)
Yep,
I like nine ounce, pretty small, super thin can. And usually we get the four pack. So we'll get like a four pack of the triple and a four pack of the mocha and drink a couple on the way there and either have some just in case like at the Airbnb or wherever we're staying and hopefully like one or two for the way back home between the two four packs between the two of us. So yeah, just pretty easy drinking nice little caffeine boost for when you need it and
TJ (08:50)
Yep.
Chris Gmyr (09:00)
Yeah, pretty good go-to and I think they've been pretty consistent for all the times that we've gotten them. so far so good.
TJ (09:07)
Yeah, I've not. I've enjoyed them and I've had them, you know, so definitely not as bad as this awful cold brew. This stuff is so bad. I'm like just finishing it up too, so I'm like suffering my way through it as we talk.
Chris Gmyr (09:14)
You
Yep, definitely. Well, hopefully you get some new, better coffee from here on out or maybe like tomorrow afternoon on out once you finish it off.
TJ (09:29)
Yeah, I gotta do some,
I gotta do some grocery shopping anyways. My 12 year old is eating this out of house and home. I just measured his height and he grew like a quarter of an inch in the last month. so I'm like grocery shopping two or three times a week at this point.
Chris Gmyr (09:40)
Yep.
Yep, they do that. You keep on feeding them. They keep growing.
TJ (09:48)
Yeah, I'm convinced he's like, maybe two or three inches away from being as tall as my wife. And that's a terrifying thing. we're like, now that he's starting to like sprout up, we're like pretty sure he's going to get taller than me. And I don't know how I feel about that. My wife's family's all over six foot. I'm five 11. So
Chris Gmyr (10:07)
wow.
TJ (10:10)
⁓ it's not that I'm like short or anything, but, my wife's family is all very tall. And so we have a feeling that he got, he got the tall jeans cause he's growing just like her brother did and her brother's like six, four. so we'll, we'll see what happens, but yeah, you keep feeding them. keep growing.
Chris Gmyr (10:15)
Mm-hmm.
⁓ yeah.
Yep.
Totally. Same thing with R2. It's just like thinking back to even like a year ago and more than that, it's just like the amount of food or like, you know, protein sources that we've had to cook before versus like now and like having more leftovers like in past years and months. And then now it's like, I made like a whole bunch and there's really either nothing leftover or very little.
And they're only like three and nine. And it's just like, what? What are you guys doing? Like have to make double and triple batches of everything now, like every night? just, yeah, it's a lot and nothing is getting cheaper either. So.
TJ (10:57)
Yep.
Nope. Yep.
Yep. Same was, man. Same was.
Chris Gmyr (11:13)
Yep, totally. So yeah, I know you had this topic on for a little while and we touched on it ⁓ about starting up the newsletter on your site. You promoted it at your talks and like, I don't know, I think it's been great so far. Been reading it on weekly basis and yeah, wanted to know how everything's been going. How you've been feeling about like the cadence, like writing the newsletter, keeping it up. Do you want to keep it up? ⁓
TJ (11:34)
Yeah.
Yeah, I'm just
realizing now that I like missed last week's newsletter. This weekend was just ⁓ a mental health struggle and it just completely slipped by me. Like timeframes I normally do it. just, I don't know. I blanked on it. So we're going to get a double update this week, but I've been enjoying doing it. It's not been like a ton of work. And honestly, what I ended up doing for most of it is I just like bullet point out.
like, you know, or maybe like a few bullet points, a couple of paragraphs. And then, um, I have a, uh, like writing style guide that I've put together a while ago for like having AI ghost write for me. And so I have it just kind of like fill in around all the content that I want. And then I go through and like read through and edit it, everything. Um, so like my workflow is like pretty locked in, which feels good. Um, and it's.
that workflow has really lowered the barrier for me to do a weekly thing. But I've been enjoying getting it out there. There's been decent feedback. The last one I sent out was just touching a little bit on some of my mental health stuff and had a few replies back of really supportive emails. And that's been great, actually getting some interaction out of it. So I think it's still pretty small. I think I just passed like
a hundred subscribers, so nothing, nothing like super significant, but it's, you know, it's kind of like the podcast. Like I don't, I don't do it for the stats. So, you know, and, and I think Aaron Francis nailed it when he was talking about his and he's just like, no, it's like a nice way for me to like, kind of like recap what I've done for the week personally. And sending it out is like not that big of a lift beyond.
beyond that, and I found that to be true too. So it's just been nice being able to recap things. And honestly, with work on Prism, it's been a bit motivating in that aspect too of I want to update people on what I've done on Prism. that means that I need to be doing something on it.
And so there's like that kind of like a little bit of motivating factor for like keeping work moving on Prism or like doing something with it every week to give me something to talk about. like that's been that's been kind of like a nice aspect of it as well.
Chris Gmyr (13:54)
Mm-hmm.
Yeah, totally. I like that it's easily digestible and readable. think the format and the amount of content is perfect, at least for me. I can read it pretty quickly and get the gist of what is going on. The Prism updates are solid. And then I also appreciate the call-outs to the podcast. So hopefully, I get some more listeners from there.
TJ (14:21)
Gaff for sure.
Chris Gmyr (14:25)
But yeah, just like it. And you kind of got me thinking of like, man, is this something that I can or should or want to do? Because I don't have a project like Prism that's publicly out there or anything like that. But with all the stuff going on with just learnings at work, general engineering leadership with
larger teams, developer productivity, a sprinkle of AI stuff. I don't know, could be useful, I don't know, to some people, or just like, I don't know, helpful for me, just like you said, wrap up the week, same thing with Aaron, and send something out, but yeah, I don't know, something that has definitely popped into my mind a couple times.
TJ (15:04)
Yeah, I think you could totally do a newsletter. It's honestly got me itching to write some blog posts again, which has been kind of like an interesting side effect of like, yeah, no, I kind of want to keep writing and do some more of that. So it's been nice. But part of it was like, gosh, am I going to have content to fill out a newsletter every week?
And as long as they keep on like picking away on things on Prism, I think there's like at least there's always at least that to talk about and then add some like personal anecdotes in there. so no, I'm, I'm really happy to hear that, like you're enjoying it and that like the, the attributes of it are like kind of hitting for you. so that's cool, man. I dig it. Like I've been enjoying it. I'm going to keep going.
Chris Gmyr (15:43)
Yeah, yeah.
TJ (15:46)
⁓ like the weekly cadence feels pretty good. You know, I'm going to miss a couple of weeks here and there. I'm sure like, kind of like I missed last week's, but, ⁓ yeah, I think that's just like, it's bound to happen here and there.
Chris Gmyr (15:54)
Yeah, understandable.
Totally. Yeah. Well, you know, keep tinkering with the idea and see where we're at. But you and the podcast will be somewhere that I can promote that if and when that I do pull the trigger. But yeah, I've been thinking about getting back to writing as well, because the blog posts have kind of waned for a while. So, yeah, I enjoy writing to it just takes a little bit more time. And we already got like a
content stream for doing the podcast too. So yeah, I'll think about it a little bit more, but it's definitely something I've been thinking about.
TJ (16:28)
Yeah, and it sounds like you've got some interesting upcoming projects at work too that might be worth blogging about. I think that's excellent content for a newsletter also. I think it's.
thing is I was like thinking about doing the newsletter, I was really, especially something weekly, I was really concerned about having enough content. ⁓ yeah, and I've, regardless of Prism, I think it's been
Chris Gmyr (16:47)
Mm-hmm. That's my concern as well.
TJ (16:56)
with like a little bit of momentum behind it, it's been not as hard as I expected to like come up with things.
Chris Gmyr (17:01)
Yeah. Yeah, and that's like with any habit, right? Like the more momentum that you get and the more days that you have in it in weeks, the easier it is to keep up with that momentum. So same thing with like just general writing, going to the gym, eating healthy, you know, doing whatever it is that you know you're interested or wanting to get into. It's always easier with a little bit of momentum. Yeah, but.
TJ (17:17)
Yeah.
Chris Gmyr (17:26)
thank for right now. Like keep up the good work and keep sending send on that newsletter. Love it.
TJ (17:31)
Yeah, yeah, yeah,
for sure. And I think like something that's been helpful is I've been like, I use bare notes. I've been, I've been trying to get back into like note taking and being on top of things. And I have just like a pinned open note for like newsletter and I can just like throughout the week add little bullet points to it. So like when I come down to write it, it's like, yeah, like I did do that and that and that.
Chris Gmyr (17:43)
Mm.
TJ (17:55)
Um, and like that's been helpful too, at the end of the week to like, feel like I've accomplished something because there are, there are weeks where just like I'm grinding, but like, I feel like I have done nothing and like I have nothing to show for it. But then I go back and like, I can look at this newsletter note and be like, oh yeah, like, no, I actually did do some stuff this week and like feel better about things. So there's, there's some like side effects that.
have been pretty beneficial outside of just having content for a newsletter, you know?
Chris Gmyr (18:24)
Yeah, totally. Yeah, I was going to ask about your process too. So I'm glad that you're taking those kind of throughout the week because going back to momentum like yeah trying to going back to momentum like that's your momentum kind of throughout the week. So it's not like hopefully as big of a push or amount of work when you actually sit down like on the weekend or whatever you write the newsletter to be like, OK, I got to think throughout the whole week.
TJ (18:33)
trying to.
Chris Gmyr (18:53)
what happened in work with Prism, outside, whatever extras. So just having that little tally throughout the week hopefully keeps that momentum into the actual writing. So I was curious about that too.
TJ (19:04)
Yeah.
Yeah. What sucks is like coming up to like forgetting last week's. so knowing that I've got like a little bit bigger of an update that I should like probably do for this week. I have like the last two weeks have completely fallen off the train of like putting anything in this note. So it'll be, it'll be interesting trying to like jog my memory of like what, what I've done because honestly, man, on like a day-to-day basis, like I forget what I had for lunch, like lunch yesterday. Like my short-term memory is just.
completely shot. So yeah, well, this next one will be interesting. I don't know. We'll see.
Chris Gmyr (19:39)
Yeah, totally looking forward to it.
TJ (19:41)
Gotta stay on the train of updating the note and keeping track of things, which is with severe ADHD like I have, is remembering to do those little things can be so challenging.
Chris Gmyr (19:53)
Yeah, yeah, for sure.
TJ (19:55)
Yeah, I wonder, I got into like daily journaling again a couple of months ago and like fell off the train. Uh, but I think that could be like helpful. It's just kind of like recapping my day in a journal entry or something and like being able to, you know, even like feed those to Claude and be like, yeah, what should I include in my newsletter? You know, like, um, here's all my journal entries for the week. Like figure it out.
Chris Gmyr (20:13)
Mm-hmm.
Yeah, yeah. I've been on and off with daily journaling off for quite a while now. And I have a work journal set up and a personal journal set up in Obsidian. And the work journal, the daily one also, I have all my meetings and meeting notes in there too. So basically it's just a meeting log at this point. Because very rarely at the end of the day do I have the time or
TJ (20:39)
Yeah.
Chris Gmyr (20:45)
the energy to actually go back and kind of synthesize what I've done for the day outside of meetings or thoughts or even linking to PRs. So mean, ideally, I'd be able to use those tools for that and then extract data out of there on a weekly basis. Same thing on the personal side. But yeah, it's just the level of effort and.
the energy level at end of the day and not wanting also to do that at night at some point or having even less time and energy at the end of the day. So I have to figure out a way that I can get back to that and make that more usable.
TJ (21:22)
Yeah, for sure. Habits are hard.
Chris Gmyr (21:26)
Yes,
totally.
TJ (21:28)
So something else that happened over the last week as I celebrated my one year at Geocodeo, which is wild to think about, that it's been a year since I started there and all the projects I've worked on and everything else. That's been really cool. But I think something that's pretty exciting
Chris Gmyr (21:35)
Yeah.
I
TJ (21:48)
that also took place at the one year mark is got a title update. So I've been the principal software engineer over the last year there. And now I am going to be the principal software engineer, comma AI. we, yeah, we're kind of getting a little bit more specific. And the exciting part about that is the goal moving forward.
is going to be for me to be spending like 60 to 70 % of my time working on AI initiatives. And so plenty of time to still be working on like other projects and helping out in other areas. But the bulk of my time moving forward is going to be focused on like different AI initiatives. I think we've got a few things, a few ideas for things we can do inside of our data pipeline because we have a pretty
extensive data pipeline to be able to do all the data appends and everything else that we do at Geocodeo on top of geocoding. And I mean, all the addresses and latitude, longitude and all of that kind of stuff. We have like an in-depth data pipeline. And I think there's some areas there that we can sprinkle in some AI. There's some like developer tool-y stuff I think that we could do too. So there's...
I think some cool stuff, but even beyond just offerings for our end customers and things inside of our data pipeline, there's things like our AI tool chain for development. So I've got Hack Friday tomorrow. We've kind of shifted to two-week chunks of work.
of planned work, and then at the end of the two weeks, we have a Hack Friday where you can put a cool project to hack on. And for this Hack Friday, I think I'm going to do Laravel Boost integration into our code base. But we have actually a very interesting challenge along with that, is that we're a mono repo of two Laravel applications. So trying to figure out how to get Boost to play nice with that scenario is going to be
Interesting. The nice thing is we keep the Laravel versions the same between the two. So if we shift one, the other one gets shifted too, so that we are working inside the same version of Laravel for both applications. But we have our dashboard and our API. And so part of the AI initiatives is just kind of driving forward AI stuff. So if we can get Boost integrated, I think that'd be great. Earlier this week, I decided to commit our
like my Claw.md file to the repo so that I went through and did like a fine tune pass on it. And that's got like a ton of rules on like, it's got a bunch of like coding rules, but it also has a bunch of like workflow rules in there as well, trying to like guide Claw towards like thinking closer to the way that we think as developers about how to approach problems and break things down. So that was like kind of, think my first.
My first AI initiative was like, let's just start tracking this. Because Cloud Code also supports local files. I think it's like cloud.local.md. So you can ignore that and add your own custom stuff in there, but all the project specifics and things like that in there. And Matthias had said in one of our weekly updates that he already noticed the difference of the added Cloud file and like,
it's like prompting to like, yeah, let me like think about this solution first and kind of like go through some of its like workflow stuff of like breaking down the problem and like how to approach it. So it was, it was cool to like kind of make that decision and then get feedback that it's already been effective. So I'm looking forward to it. Matthias has said that he would love for me to be able to like spend time working with Prism at Geocodeo, which is like
super cool because like honestly I haven't had a ton of time to build with Prism. It's been mostly building Prism and so getting an opportunity to like really use it more in depth I think is going to be a lot of fun for me. Great benefit to Geocodeo but I think it's also going to feed back a lot of cool stuff into Prism. You know the more I use it the more ideas I get to do things with it so
I think it'll help keep pushing that project forward.
Chris Gmyr (26:00)
Yeah, totally. I think there's so many good use cases for Prism in Geocodeo, especially like on the dev tooling side and on the API side for the dev portal and even all the Clickhouse stuff that you guys have been doing to be able to pull out and extract data from Clickhouse or seeing what calls are taking longer or like how can you fix them or I don't know.
So many things that you could do with Prism and AI integrated into the GeoCodeo service in general. I think there's some cool projects in there for sure.
TJ (26:32)
Yeah, like one project we've kind of earmarked is for our data pipeline, having it do research for us, trying to find data sources. So that's a big thing too, is we've got maybe a stale data source, and the original source of that's gone. And so we're running on our own cache of that. And so getting AI to go out and do research for us to go find these new versions of these sources.
We've used, we've successfully used Claude in research mode to do it. So like kind of building out something a little bit more dedicated for it. I think it'll be great.
Chris Gmyr (27:07)
Yeah, that sounds awesome. Because I can imagine for, I don't know, there's so many sources that you guys probably use that are local and very small or minimally or not maintained after a while. So that can be a huge challenge. So having a little bot that just kind of goes through and be like, hey, this thing is six months old. Let me see if I can find a comparable or maybe just the source of it.
change to a different location or government endpoint or something like that. So yeah, I think that would be pretty awesome too.
TJ (27:42)
Yeah, sick man. I mean kind of coming off of that Topic you had was like some like miscellaneous like ⁓ anthropic and clot updates
Chris Gmyr (27:49)
Yeah, I'm sure you've gotten a. Yeah, totally. I'm sure you've gotten ⁓ the email notifications, but it seems like Claude and Anthropic are moving away from the Anthropic company name and basically branding everything as Claude. I think that makes a lot of sense, because I'm sure people are just using Claude AI and that's kind of how they reference it and.
TJ (27:50)
I think it's a good segue into that stuff.
Chris Gmyr (28:14)
when news articles or whatever is talking about Anthropic, maybe people don't make the direct connection between the two. So basically, they're changing any reference that they have from Anthropic to Cloud. So like Anthropic platform is going to Cloud Developer Platform, Anthropic AI, Cloud AI, Anthropic Console to Cloud Console.
So on. So I think this will make a lot of sense. Basically, like, no functionality is changing. It's just the name update and maybe some URLs that are going to point it to the new location. yeah, I think that's a win, making things a little bit simpler in the AI world and what you can do with it.
TJ (28:52)
Yeah, I think there
was definitely some like brand and like naming confusion that was happening as as new stuff has come out. They've kind of like pushed more towards this like Claude name anyways. And then you still have like Anthropic in there. And so like there was definitely starting to get some confusion going on. So I think it's I think it's great for the space to kind of like
Chris Gmyr (29:04)
Mm-hmm.
TJ (29:14)
Unify the name behind Claude and like I don't know I'm pretty sure the parent company of Anthropic like that's all staying the same but just all of their like Claude offerings are gonna be now branded as Claude instead of like Anthropic and Yeah, I think I think it makes a ton of sense because it's definitely starting to get a little confusing and a little little all over the place so
Chris Gmyr (29:37)
Yeah, so I think it's a good move ⁓ and they're basically Have rolled that out now and some of the old anthropic branded URLs will just be deprecated by the end of the year, I think but Yeah, no, no major plans there, but just like a nice little branding update and then the second one that I have on here is For I believe the the max plans You can sign up for these like research previews and I know
You've mentioned these before with certain features. And I think Cloud Code was a research project or offering at some point. So yeah, they just pushed out this cloud for Chrome. So you can get it in the Chrome Web Store. And you do have to have a Max account. So even though you can find it in the Chrome Web Store, it won't actually allow you to sign in unless you have Max credentials.
TJ (30:03)
Mm-hmm.
It was.
Chris Gmyr (30:28)
I tried to install Integra with this the other day, but it does not work in Arc. So I'm still on Arc. Apparently, you have to do it in Chrome, because I'm used to extensions just having a little pop-out window ⁓ on top of them, like very small eighth of a screen size of just little mini modal, basically, popping up. Apparently, Claude opens up a whole side panel.
TJ (30:35)
Mmm
Chris Gmyr (30:54)
which only works in Chrome, apparently. So I saw some Reddit threads or something like that about why doesn't this work in Arc. But I guess there's something walking in Arc versus what you can do in Chrome for these slide out panels. So it doesn't work in Arc. So I have to download Chrome and play around with it, which is a little annoying.
TJ (31:13)
Yeah. And I mean,
I mean, at the end of the day, like Arc and like they had, they've pushed for their own AI integrations and stuff. So I don't, I don't know how much I see them like trying to make this work. And with Arc being kind of like a deprioritized project, like I don't, I don't see them making adjustments for that, but I have it. The problem is I use Zen browser, so I'm not in.
Chris Gmyr (31:38)
⁓
TJ (31:40)
I'm not in Chrome.
Chris Gmyr (31:40)
That's a Firefox base.
TJ (31:41)
Yeah, and that's a Firefox base. like I haven't really had a chance to play with it yet. And honestly, I'm not even like 100 % sure of like great use cases for it. I've just I've never I've never really messed too much with AI in the browser. So I tried other extensions and stuff in the past, but I just
I struggle with knowing what to use it for. Like, do I use it as a research mode? I know it can like take actions on your behalf. I might try next time is going to like my barber shop's like square website and seeing if it can like schedule me an appointment. Like that might be interesting of like, yeah, schedule my haircut appointment for me, you know?
so yeah, right. Yeah. So I, I don't know. It should be, ⁓ it should be interesting, but like, yeah, like I said, I just, I don't, I don't know of like great use cases for it. So I haven't really used it yet and it's not my primary browser. So it's not like, it's something that I have to like explicitly go into Chrome and like try out, which I'd like to,
Chris Gmyr (32:24)
ends up signing you up for a perm or something.
TJ (32:47)
So we'll see how that plays out, but it's definitely interesting. I like that they did an extension for it. Like it seems to, like I was just playing with it. Yeah, like it's got this little like side pop out and I don't know. We'll, I'll have to play with it because it definitely seems interesting.
Chris Gmyr (33:10)
Yeah, and I had a use case for it yesterday, but didn't want to do the lift to actually spin up Chrome, like sign into everything, and do all the things needed for it. It's kind of running out of time at the end of the day. But in one of my projects, I'm doing some research. So basically, I'm bouncing back and forth between AWS data and Datadog data. And certain pages you can export.
data reports or like CSVs or something like that. But other pages you can't. And one use case that I wanted to try out is on a Datadog page showing queries. So the amount of queries that were run or even like average time spent like in that query. It's just on a page you can't export it or do like anything with it. So I was gonna ask Claude like if you can.
pick up this data in this table and create a CSV or even just ⁓ copy and pasteable CSV data from what was shown on the page. So I was going to try that today and see if that worked because I think that would be super cool if you can scrape your own. Because what I had to do is just take screenshots of it, which is not ideal.
TJ (34:19)
Yeah, that sounds like great use case.
Chris Gmyr (34:30)
if you wanted to get actual data from it later on. So yeah, I'll circle back to that today and see if that works out with the actual Chrome install.
TJ (34:40)
Yeah, I think that's like a very interesting use case of like, doing some like scraping data extraction kind of stuff. yeah.
Chris Gmyr (34:46)
Yeah, so I can talk about that a little bit more probably next time. But yeah, nice things coming in Cloud. Still seeing those updates pop in in Cloud Code, so getting those updates. I did see, oh, we did talk about this before of get ignored files in Cloud, not being able to come up on the at file picker that they have. And that's definitely a well-requested feature.
TJ (34:58)
Yeah.
Mm-hmm.
Chris Gmyr (35:14)
There's a couple of GitHub issues on it. So a lot of people are talking about it. A lot of people are running into it. But yeah, it seems like in general, Cloud Code looks at the gitignore files and just doesn't bring those up in the auto picker file thing that it does with the add symbol. So hopefully they adjust that or at least have some controls or add an a Boolean config to turn that on and off depending on the project.
TJ (35:41)
Yeah,
think even just having a config value of like, yeah, on or get ignore or whatever would be nice because there are ignored files that I would like to reference. ⁓ So I'm glad it's well-requested feature. So we'll see what happens.
Chris Gmyr (35:51)
Yeah, totally.
Yeah, for sure. But yeah, I know we're getting a little bit late, but want to do some Prism updates?
TJ (36:04)
Yeah, so like I said in my update, I definitely feel like I'm way behind on open source management. So there's stuff that I'm going to try to get to this weekend. It's my wife's birthday this weekend, so we'll see how much time I actually have to work on it, because we've got some plans. But I would like to spend the weekend, instead of writing code, I'd like to spend the weekend just trying to get up to date on issues and pull requests and stuff.
Chris Gmyr (36:28)
Mm-hmm.
TJ (36:30)
As far as the streaming output refactor, so far I've got, two, three, four, I've got five out of the nine. I have five out of the nine providers done. What's left is like Maestro, OpenRouter, XAI and DeepSeq. And I've had Claude try to do Maestro like three or four times now, and it has absolutely not done it correctly.
all three times. So that's one I might have to manually go and convert. And I haven't tried the rest of them yet. So like, I'm thinking that the rest of these might be a little bit trickier for Claude to migrate. I hand did Anthropic, Claude did OpenAI, Gemini, Olaba, and Grok with a Q really pretty easily. But Maestro is just choking on so.
Chris Gmyr (36:55)
the
TJ (37:16)
Um, I, I've made decent progress on it. I've got more to go so far. It's a, uh, 84 changed files on the PR, uh, 6,200 additional lines and like 1100 deletions. Um, so it's, it's, and there's, like I said, four more providers to go. So there's, there's a lot, um, in this PR, but you know, once it's, once it's ready, it'll be ready. Um,
I'm hoping I can get it done next week. But what I did do, I have done some stuff. I know somebody had asked me and like, has been something that I've been asked quite a bit. So this is probably going to turn into a blog post, but I've been asked like how to do rag with prism, uh, like retrieval augmented generation. So, um, I went ahead and did build a like chat, the Laravel docs.
Chris Gmyr (38:04)
Mm-hmm.
TJ (38:11)
rag piece of functionality. So it's like I built an artisan command that loops the chat so that you can have a persistent chat across a multi-turn chat. Instead of just asking a single question, getting a singular answer, you can actually have a conversation with it. So I built a command that you can do that. And then I gave it a Laravel documentation tool, very similar to the MCP tool that Boost uses.
So there's a layer of all docs tool that it can like ask a query. And then inside of there, we use prism to generate embeddings on the question and then query though against like PG vector and Postgres to get the embeddings of the documentation chunks. And then there's like a, order to generate those, I have an artisan command that I built that goes through all the layer of all docs.
chunks them up into smaller discrete bits and then embeds them with Prism and then stores them in PG Vector for retrieval by the command. And I knew all the fundamental pieces were there. I just have never built it before. And honestly, it was not that bad. The bulk of the work was really in writing the command that parses the docs and stores them.
in PG Vector. I want to clean up the document command a bit and that like the document splitting command up a bit. And then it's probably going to be a blog post because I do get asked about it quite a bit. And the implementation wasn't as bad as I expected. And I think there's definitely a project that I wanted to do with Prism later on is called Fragments. And that like that
or I guess fragment, that package is going to be designed around doing document splitting and the storage of everything. So you'd pass it a database connection for PG Vector, and then you'd pass it the directory or something. And I'll crawl the directory and then take all the documents, break them up into semantic-sized chunks.
Chris Gmyr (40:06)
Nice.
TJ (40:21)
and then store them in like PG Vector or something. So I think there's a lot of opportunity to make that really easy and like the document splitting and storage. So I definitely think I want to pursue that later on with Prism. But I was super stoked that like I built it, it works. Like RAG is absolutely feasible with Prism. You know, we've got all the foundational level.
components and pieces of functionality for it. So I want to do a blog post on that. that definitely through that project that also put Cloudflare AI back on my radar because they have like an auto rag feature. And I had talked to somebody at Laricon about doing Cloudflare AI as a provider. that a fast follow after the streaming output, I think is going to be
adding Cloudflare AI as a provider. So yeah, I think that'll be pretty cool. so yeah, so not no work on Prism, just taking a little time to build with Prism instead of building on Prism. And then I'll get back to it, I think this weekend and next week and hopefully make some like big progress forward on especially the streaming output refactor. Like I just I want I want to be done with that so bad.
Chris Gmyr (41:33)
Yeah, yeah, totally. Yeah, I think that's a great blog post idea with the rag and documentation MCP. So I think you should definitely put that on your list to do, you know, with all of your free time. I would definitely read it. And then if you're up for sharing, like the actual repo or code base or whatever, you know, it is that you're working on locally, if you put that up on the Prism PHP.
I think that might be helpful for other people to see too. I don't know if you had that planned or not, but that would be pretty cool.
TJ (42:01)
Yeah, yeah, I don't know about doing a repo, but ⁓ definitely a blog post with like the code snippets and everything ⁓ for sure.
Chris Gmyr (42:09)
Yeah, that work.
Very cool. Love it.
TJ (42:12)
Yeah, cool man. On that note, you want to wrap things up?
Chris Gmyr (42:16)
Yeah, let's wrap up. Thanks for listening to the Slightly Caffeinated podcast. Show notes and all the links and social channels are down below and also available at slightlycaffeinated.fm. Thank you for listening and we'll catch you all next week.
TJ (42:29)
Yep, we'll see ya.
Creators and Guests

