TanStack Run, September 24, 2026
Recording: https://x.com/i/spaces/1rxmqpnNjNNxy
Duration: approximately 70 minutes, 52 seconds

Automatically transcribed from the recording and lightly corrected for obvious product names. This transcript may contain errors. Speaker identities are not reliably separated. Timestamps are approximate and refer to the recording. Unreliable opening and closing text has been omitted rather than reconstructed.

[00:00] [Opening audio not reliably transcribed]

[02:48] But I'm about five minutes into my bionightly run. Something I've learned is that I get bored when I run. So I just got to find ways to take my mind off of it. But last time I ran, it was just like chatting with ChatGPT for like an hour and a half, trying to get it to build stuff. But it just kind of felt wrong to run and work at the same time.

[03:37] But I definitely didn't have to talk this much. Anyways. Good conditioning. I've got to run a half marathon at the end of November. I'm definitely not ready. Okay. It's okay. I've been told that all you got to do is be able to run 10 miles. That's exactly what I did last time.

[04:21] I just got up to 10 miles. And then I let adrenaline carry me the rest of the way. I might have to do that plan again. I suffered a foot injury about a year ago. In New York City, actually. I hurt my foot. And it took like seven months to get that thing back to normal. Sucks getting old.

[05:05] I mean, I'm not that old. I'm 36. But it definitely sucks. Okay. Okay. Something I've been marveling at lately. It's just how easy it is to build stuff now. I'm sure like everyone else. I'm just like... Almost like amazed at how...

[05:47] Like normal it feels now. Not to write code. That probably makes... That probably makes... Some people really sad. I know it does. There's a lot of people out there who just... You know, they... They stake their careers... And their education... And just like all of their usefulness... On just knowing how to code. I mean, I did too, to some extent. I thought... I mean, I know that... I was really fast.

[06:29] Man, I could code fast. Sometimes I used to think that like... When they call people good... Engineers... I swear a lot of the time... It was just turnaround time. Like how fast could you bring something... And so much of that just had to do with like... Formulating your ideas... And... Getting them out. And it's just... It's funny that a huge part of that... Used to be just... Typing. I'm trying to say this with the most respect I can.

[07:20] But... I... I thought... I thought I was going to be scared... When... I stopped... Typing. Or programming. I thought, oh no. If this goes on long enough... I'm going to lose... My personality. I'm going to lose... You know, my sense of worth. I think what I really discovered... Is the same thing I've been learning... Since I was young. Since I was... A little boy. Is that I like to create. I like to build stuff.

[08:04] I do like to understand how things work. But not as much as I like to build... I used to be really into music... Like a lot of other... Programmers. Man, I played everything. I took piano... Guitar... Bass... Drums... No strings. It's a big regret. Never really had the desire... To play any... Brass or woodwind.

[08:46] Thankfully. As you can tell. I don't have the lungs for it. But... I love making music. For a long time... I thought I was going to be a... I thought I was going to be a musician. And then... For a long time after that... I thought that I was going to be... A... Studio... Director... A producer. Because I loved to... Program music. You know?

[09:28] And I thought I wanted to do film. And... Went to film school for a while. I'm going to find out... I just liked... Building things... With technology. Which means I also loved Photoshop... And Illustrator. And... Man... It just goes on. And I think... I think building with AI... Is... Is... Is going to be... Exactly the same for me. You know? And I am excited about it. I've had so much fun...

[10:10] Over the last year. Because I've been able to iterate on so many ideas... So fast. And not in like the exhausting way... But just like... Trying crazy random things... Pushing... Pushing the tool from the limit. It's just a lot of fun. And it's also been a lot of fun for the serious stuff too. Like for TanStack. For all the TanStack libraries. It's just really crazy to see... Well, AI is...

[10:55] Responsibly... Changing... A lot of things for the way that we... Uproach our open source. You know? Not everything. Thankfully. There's definitely some bad... Some bad parts to it. I don't think they were just catching up there. Especially around like... All the crappy PRs and issues and stuff that we get. It's just... Dealing with all the AI garbage. Some of it is just like... Directly in your face.

[11:36] Just garbage. It's just garbage. It's okay. Because you know what? Even before AI... There were a lot of developers that... Would come in and... Drive by... Issues. Drive by PRs. Get into... These weird... GitHub discussions that are like... Really entitled. It almost made it worse sometimes that those were real people. Who could like actually feel that way. Feel like... They were owed something... From free software. And I've always believed that technology is just a...

[12:22] It's just an extension of our desires. I frequently talk... To some of the youth... In my church. And also... Friends in my church. And my kids. And family members. How... Technology is a double-edged sword. Or an axe. Depending on how nerdy you are. Want to get. And they're like... Oh, so what does that make AI? And I say... 50 foot long lightsaber. You know? It's still just an extension... Of your desires.

[13:04] And your motives. And I like that. I'm gonna stick with that. It's an incredible generator. I think we can do some really amazing things. Like template generator. Can you hear me? Can you hear me, Tanner? I'm gonna walk... For like 30 seconds to catch my breath. It's definitely harder to run... While you're talking. It's good conditioning. Check my phone. Oh, nice. There's like... 10 people in here.

[13:44] That's cool. I honestly thought I was talking to nobody. I figured... At least it's being recorded. But... Hey, if you guys can hear me... Throw up an emoji or something. So I know this is... At least working. Though I can't imagine you guys are like... Hanging out in here. Just listening to like silence. So it's gotta be working somewhere. Any of you guys wanna talk about anything? I'm... Happy to... Chat. Take questions... Open... Openly about whatever you want. I'm pretty sure I opened up the... Permissions. So you can just kinda unmute. And respectfully start talking.

[14:27] So if you wanna do that, go ahead. Otherwise... I'm gonna start running again. Feel free to interrupt me. Anytime you want. I'd much prefer talking to people. Than at people. Maybe it'd be cool to talk about TanStack charts. It's currently in alpha.

[15:10] I don't know if you guys are talking about TanStack charts. I've seen TanStack charts. But... I think there's a pretty cool... Pretty cool story behind... TanStack charts. It goes way back. It's like a... Like a history lesson. I like... Not too far though. I'm not that old. I like 2014, I think. I... Started getting into data visualization. Because I had to. I was working on a... A project called Nozzle.

[15:52] Nozzle is a... It's an SEO... Marketing data platform. It's a lot like... Moz. Ahrefs. It's a pretty sweet product. And... And... Oh, I just got a phone call. I gotta make sure that didn't... Screw it up. If you can hear me. Emoji. Yes. All right, we're going. So...

[16:33] TanStack start... Or sorry, TanStack charts. Yeah, so... DataViz was complicated back then. You know, D3 was... D3 was actually kind of like this new... Kind of up and coming thing. It's getting really popular. Tried out some charting stuff with a... Library called... C3. Which was built on top of D3. But it was for like dashboards. And... C3... Used this... Crazy... Crazy library under the hood called Crossfilter. And Crossfilter was like...

[17:14] This wild... Idea of using... Like... Bit swapping... To track... If... A specific record... Had matched a filter or not. So you could get this like... Crazy... Real-time... Filtering... At like... 60 frames per second... With Crossfilter. It was a library that was... Being built by Square. Or some developers of Square. Might have been like... Jason... Davies? I don't... I think it's Jason Davies. But... Used that for a little while. I even got into Crossfilter a little bit.

[17:58] But... Eventually... Landed on Chart.js. Mmm... Chart.js is great. It was okay... When I found it. Nick... Nick was the guy who built Chart.js. The first version. He was awesome. Nick... Nick was ready to throw in the towel. He was like... Yeah, I'm done. Open source. I'm over this. So... He gave me write access... To kind of... Keep it going. And then he also gave another guy... Everett. Everett Timberg. My Canadian pal.

[18:41] He gave us write access... So we could... Go and build... The next version of Chart.js. Man, we built... Like... We went to town. We rewrote everything. We did like... Not everything, but... A lot. We built a new animation system... That didn't exist before. Tons of stuff around animation. Twining. Which is all really wild and crazy stuff... When it comes to... And I learned a valuable lesson. That I never wanted to write...

[19:21] Directly with the Canvas API... Ever again. So we launched it. It was awesome. Chart.js 2.0. Stuck with that for a while. But then React happened. React was crazy. Changed the way that we wanted to think about UI. Changed the way that I wanted to think about charts for a while. I was like... Declarative charts. Yes. This has got to be better. So... Got into all these other libraries. Tried them all out. Ended up building my own. Called React Charts. How that empty name was not taken, I have no idea.

[20:07] But... Grabbed it. Built React Charts. Totally custom. Used React for all the rendering. D3 for all the calculation and layout. And... Just made it as simple... And cool as Chart.js was. Not as many chart types. As... Pie charts and... Donut charts are... I don't like them. Man, they look pretty. But... They're hard to read. For most people. Or they can lie to you. So... Kept it standard. X and Y. And we used that nozzle for a while.

[20:51] It was great. It was really automatic. Everything about it. You just kind of... Here's your data. Here's your chart type. Just figured everything out. Figured out like... Automatic margins. For... Labels. Which... You have no idea how hard that is. Because... If you start rotating labels... On anchors. On the ends. You start having to do some crazy trig. Get back into like... All your geometry stuff. I did not like that. Just doing it all by hand too. It's crazy. And... It was cool though.

[21:33] Very automatic. But then... I realized that like... It wasn't very expressive. You can only build a couple of chart types. It's hard to mix data. It's very performant though. Very cool. Could be faster. But I didn't have AI at the time. So... It's just a lot of... Profiling. And just like throwing spaghetti at the wall. But... After that... I'm gonna take another break here. Because... Pretty sure my pace was... Way too fast right there. Because I started talking about charts. Gets my blood pumping.

[22:18] Yeah. It just wasn't expressive enough. We wanted to start doing more stuff. You know... Node graphs. Directed force graphs. Or just like... Circular charts. Or... Geo charts. All that kind of stuff. And... Had to reach for another library. I didn't like that. Um... But it... Stood its ground for a long time. In nozzle. And then... What happened after that? Is I was like... I need... I need... More expression. I just need more power. Out of the library. And you know what? I don't care so much about this declarative thing anymore.

[22:59] I started studying charts like... You know... On a... On an academic level. It feels like just reading everything I could. In the D3 source. I read Lee Wilkinson's grammar of graphics. And... It just changed the way I thought about data. After that. And about data viz. And then I found out that... Matt... Or not Matt. Mike... Bostock. Mike was... He was working on a new library. And I was like... Whoa! The D3 creator is working on a new library. Like... This has gotta be... It's gotta be something really cool. It's called Observable Plot.

[23:44] And he was building it for Observable. It's like a Jupiter style notebook platform. So... It was like... I wouldn't say it was super opinionated. Towards Observable. But it did have some... Some trade-offs. That it made. For that application. But even just as an open source library. The way that it was. It was amazing. You could just take... To map it to shapes. Or layouts. Or whatever. It just worked. It's magical. Used that for a really long time. Not a really long time.

[24:24] I'd say like three years. But... Eventually... Caught up to me again. Because I realized I had all these wrappers. Around Observable Plot. To do tricky things. Had some features that I wanted. Around the ergonomics. That probably just... Weren't really the priority. Of the project. Maybe a little beyond scope. So... I just decided to throw in the towel. With Dataviz. Stuck with Observable Plot for a long time. Just continued to... Kind of build up. Build up my ideas.

[25:06] On what I liked and what I didn't like. Tried a couple others. Some paid versions. Nothing worth mentioning. The D-charts. And then AI happened. And I was like... Oh my gosh. This could change everything. And I waited... For a year. Before I tried it. I'm glad I did. Because... I took my time. Like gathering all the information that I... That I knew was really important. Around... API design. Not just for humans. But like API design around... How agents...

[25:47] Are gonna wanna write charts. How they understand data. The... Patterns they reach for. Out of the box. And I had all these... Libraries that I had tried. And I just knew all the things about them. That I... Loved. And hated. So I had this huge plan. Exactly how I wished... A charting library would look or behave. And when I started writing... TanStack charts. It only took me... A couple hours. With AI to have the first working version. And I had no idea how it worked under the hood.

[26:29] I mean... I gave it enough direction. You know... To kind of get... The overall architecture. Scene graph. Reactivity. Utility driven. But there were a lot of other things that I had to... Massage. To get the way that I wanted. Especially around type safety. You have to structure the APIs... In such a way that types flow really well. Something you have to think about... From the very beginning... Of designing a library. Type safety and type flow. And AI just loves that stuff. Eats it up. And there were just a lot of other nice polished items...

[27:16] That I had... Gathered up over the years with... Chart.js and React charts. And it came together so nicely. I launched the Alpha... In like less than a week. Or no... The just... Yeah... Alpha experimental. And I showed it to a couple of my friends. Showed it to the nozzle. Showed it to anybody I could that I knew was doing... Data viz. And... They just told their agent to switch. Just to move over. And I had written... Well, I instructed to write all the docs. All the API... Design sources.

[27:57] Everything. It was just all there. I even had migration guides. And... All these people were just like... Hey... I told my AI to switch to it. And it did. And guess what? Our charts are faster. We have more expression. And one of the things I really focused on was... Bundle size. And they're like... Oh man... My bundle is tiny now. I'm like... Yeah... Because it's only using the stuff that you're using. It's only bringing in what you need. And I... Spent like... A couple billion tokens on it. So... Make sure that... It was as small as possible without...

[28:38] Breaking expectations. And tests. And I think that... I think that was really... It's really one of the turning points... I've been developing right now. Is... I wouldn't necessarily... Go back and... Rewrite every library... That we have with AI. I think that would be a big mistake. And there's certainly projects that... I would not trust to AI right now still. Just because of the sheer... Complexity... That they bring in. Namely things like... TanStack router. TanStack start.

[29:22] Table. TanStack table is a big one. Form. Anything with like... A sufficient crazy amount of type safety involved. You know... AI can definitely help you get there. Which is why we're all using it. To like... You know... Extend... Our desires. Extension of our... Of our... Chalice. But... Charts was the very first one that... You know... What I needed... And what I wanted to build... Actually fit inside of the capabilities... Of the model. To do. So even though I know the architecture... It's because...

[30:03] I asked my agent to teach me... The underlying architecture that it landed on. Which is so crazy to go back through it... And see stuff. Or I'm like... Oh man... Yeah... scene graph... Of course... Like I did that in Chart.js... To see like... All this crazy math. You know... I'm like... Oh, I was doing that in React charts. And it's cool to see it... You know... Land on a lot of the same ideas... That I would have had. Or definitely... I nudged it in that direction... Because... I fed it. React charts. You know... There's some... Definitely some training data. But...

[30:45] Regardless... That's not the impressive part though. The impressive part is how... Performant it is... The smallest... And I'm surprised... How type safe it is. And I was able... To get exactly what I wanted. It's just weird. Yeah, it definitely took money. You know... It took money to do that. And that's the part that I kind of resent. I don't want... I don't know if I like a world where... You know... Tokens... Tokens become... The only way that you can write...

[31:26] Really, really great software. But I mean... Beforehand... We had a different kind of currency. You know... It was... It wasn't tokens... But it was definitely time... And money. If you were a company... It was money. If you were to dissolve... It was time. And I gladly would have spent a year... Building this library. I did. I tried it. I put a year... Or more... Into React charts. You know... But I think I'm ready to accept that... I like the trade-off now. Where... I know what I want...

[32:07] Out of a library. I know how to design it. I know what's going to work well. And at this point... I'm more than happy... To skip over... I don't want to say skip... I'm more than happy to... Trade. Endless, endless amounts of time... That I would be pouring into this thing... For... The cost of some tokens. To get to the parts that really, really matter. Skip over the trig. You know... Not worry about... Re-implementing yet another scene graph... Again and again. Get to the parts that I think are really going to make... Charts better.

[32:49] And... Make agents better at building charts. And I would make that trade-off again in a heartbeat. And a lot of that goes back to... Just like... The things that I care about. I don't... Like... I care about you guys. I care about developers and... Building great stuff but... Not enough to like... Open up my laptop... Every day and every night when I get in bed. And be thinking about charts. And dreaming about... About charting library stuff. I don't care about it that much. Not anymore. I used to, but I can't...

[33:29] I don't think anybody should care that much. I'd rather spend it with my family. And doing other things that I enjoy, you know. I think that's where I'm realizing that this is a good trade. And also, it's not like we're trading up for something that's different. We're trading up for... I think something that's better. It's just... It's just a different way of getting there. And it still takes a lot of work. My goodness. It still takes a lot of polish. So... Yeah, this is a... These are new thoughts. You know, thoughts that I've been grappling with for a while.

[34:10] Definitely not as radical as like... Some of the DHH thoughts I know that have come out this week. Or several other like, quote unquote, thought leaders in our industry. But... Like I'm... I'm more than happy to trade that time for something else. That I think is more valuable. And I want you to... I want people to put it to the test. Like I seriously think, go put it to the test. You don't even need to use AI. Just go and try... TanStack charts. And... Where I'm comfortable... Saying this... Is...

[34:52] You will enjoy it as a human. The API is fantastic. I designed it. I designed the API. The experience. I manufactured... Parts. A huge part of what you're going to feel. When you go to use that API. You say, wow. This just works. But not all of it. You know, it goes back to like... Observable plot. Mike Bostock. Lee Wilkinson. D3. Everett Timberg. Nick. You know, it's like... Even before this, we were always...

[35:32] We were always adopting and evolving ideas that came before us to be something better. And... I'm just still really proud to be able to call... TanStack charts something that... I definitely... Feel like I created... To... Take those ideas and... And like... Give them new life. You know, instead of people getting into their... Into plot or codex and saying... Build me a pie chart. And it just starts whipping together random... SVG and... Random math problems. Equations and things that are going to break.

[36:14] It's like, no. We've already solved this problem... Countless times. And it would be a real shame... To... Have the new generation... Of creators... Who... May not even care to think about that level of primitive... and have to relearn all that stuff over again. That would really suck. Which is why I think we do have a big responsibility as open source maintainers to make sure that we are embracing the new tools, using them for what they're good for,

[36:57] and making sure that the things we've learned don't get lost to the non-determinism It's a lot of fun, though. Man, that was pretty deep. All right, I'm going to come back and take a little break. Let's walk for a little bit. Hey, Tanner, can you hear me? But, yeah, it would be a real shame for that to happen. But it is truly just so easy to create now. Now, I just made a tweet about this, kind of a joke. You know, but I said, I'm working on a new project.

[37:40] And naturally, I said, hey, Codex, I want to build this new project. You know, outlined it for a couple hours. And then I said, all right, let's build this thing. You know, start with a proof of concept. And it knows me. It picked TanStack Start and Router and Query and a couple others out of the box. But I noticed that I didn't pick all of them, you know. I didn't pick all the TanStack libraries. And I didn't even bother to say, oh, make sure you're using TanStack for everything. I didn't say that. I just started building. Telling it what I liked, what I didn't like. And then I noticed something that I started seeing these little things.

[38:27] It's like performance problems. You know, I started seeing little glitches. Little user experience things. Where I'm like, you know what? 15 years of building open source and 20 years of doing software. That should be, like, we should never have to worry about that stuff again. And I started seeing these little things here and there. And I went to the thing. I said, hey, I'm seeing this weird little flicker here. What the heck's going on there? What is that? And it's like, oh, oh, yeah. We're not caching that thing correctly. So we've got to refetch it. I'm like, well, what are you doing?

[39:08] It's like, well, you know, TanStack Query is great. Like, real time. And I'm like, well, TanStack DB, go. It's like, okay, I'm on it. You know, TanStack DB comes in and it reads all the TanStack DB skills. And it's like, all right, problem solved. Oh, and by the way, it solved a bunch of other stuff too. I'm like, sweet. And then I get in there and I just noticed right out of the gate. I'm like, oh, we're going to need to virtualize this thing. It's a huge div. Saw it before it was even a problem. Said, hey, use TanStack Virtual. Said, got it. TanStack Virtual, I'm on it. Brought it in.

[39:50] Read all the skills. Knew how to do it. And now I've already avoided so many other problems. You know, it's like you get deep enough into building really serious stuff. Still got to know what you're doing. Even if it's as simple as picking one library over the other. Or just telling it to use a library. It's still not very different than it used to be. You know, it was still a question. Do I use React Query? Do I use SWR? Do I use Apollo? Apollo? Right?

[40:35] It's funny. We used to... We never used to fight over whether you should use a library. Some of us did. But a lot of us didn't. It's like, well, yeah. Of course you should be using something for that. You know, we're not an animal. We're not animals. But you got to use this one library. You know? So we all kind of came to the conclusion that you got to use something. But now that we're back, everybody's using AI. Now we're reverting that. And it's like, well, we got to care still. I don't want to make this...

[41:15] I don't want to turn this into an extension of Brooks' Do We Need Frameworks article. I agree with a lot of what he says in there. Not all of it. But with a lot of it. And... I'm just saying that the practicality of being able to spot a problem and have the confidence that if you know the magic word and you wave your wand around and you say, use TanStack Virtual. Use TanStack Charts. You know, use...

[41:58] Use... I don't care what it is. Use effect. Not React effect, but effect. You know, use anything. If you have the confidence that that's going to solve the problem, and it does. I mean... That's still super valuable. And those tools... Until... Until those magic words are baked into... Every system... Those magic words are going to be useful. And the technology that they represent... Is still going to be so important. Tanner, can you hear me? That's kind of where my tweet for the directives came from.

[42:41] Man, I hate directives. In JavaScript. I can hear you, but I don't think Tanner can. Okay, okay. But... I can hear. The feeling... That they give you... Is undeniably great. When I was writing code... I did not like directives. I'll still date for quite safe reasons. And a lot of other reasons. But... The idea that you can just tell... Your system to use something. And have that problem be solved. I mean, I'm pretty sure that... That's where... We've wanted... DevTools to go for...

[43:22] Since the beginning. You know, now that we're there... It would really be a big shame. To just... Get rid of it all. And roll the dice to chance. Which is why everywhere I go now... Just as a hobby... I look for people who are using AI... To do anything. And I have them show it to me. And... Usually I'm impressed. I'm just like... Wow. I can't believe you built this. This is amazing. And then I ask them... You know, what's wrong? Is there anything... That could be better?

[44:04] You know... After a little minute... They... Usually they'll land on something... That has to do with just... Orcs. Thugginess. Unexpected... Things happening where... No matter how many times they ask the agent... You know... It just kind of... Isn't quite right. Like... Sometimes you can get a brute force through it. Right? But then... You end up... Recreating all the work that we've done... Over the years. And I go... Well... It's funny you're doing this... Because we have a library. Just tell your agent to use TanStack... To solve that problem. I don't even tell him the library.

[44:47] I just say... Just tell it to use TanStack. We're like... What the heck was TanStack? Like... It's a magic word. Do it. We'll try it. And come back to me... And like... Dude... It's fixed. It's amazing. Yeah... And I... My agent told me about TanStack. That's crazy. It's a lot of fun. Not gonna lie... That's where I get a lot of my... Where I get a lot of my excitement lately. Time to take another break. Look at my phone. For a second.

[45:30] Okay, cool. We got some people here. Oh, Kevin's here. Nice. What's up, Kevin? You can come on and chat if you want. Please. Yeah, can you hear us? That'd be a lot of fun. But... Just to... Maybe wrap up that idea that... I'm just having a lot of fun. And... I think it's a... I think it's a bad idea to be... Completely... Polar or binary with any decision that we make. Or any stance that we take. Other than a few things in life. But most things have a lot of nuance. And...

[46:14] I'm definitely not in any camp of... Well... AI should be used to build everything. Or... AI's evil should be used to build nothing. But just like any other tool... It's... It's got its place. And... More and more confident that I am... That I can wield it in the right scenarios. You know... The more fun I'm having doing it. I'm not ashamed to say that I built... The entire TanStack Charts library... Without writing a single line of code... Myself. And... It was the greatest experience in my life. Building a library. I thought it was the most fun part. Coding and programming.

[46:54] But it wasn't. The most fun part about it was... Getting my ideas out. And seeing my ideas come to life... In a way that could help other people. Because programming was a part of that... I associated... I associated success... And happiness and fulfillment... With that action. Because it was required. But... I'm learning that... That's not... Anymore. It's great. And I'm going to keep going. We got... We've got all the...

[47:35] All the energy to program now... In the world. We just got to know what to build. Or really... What not to build. Right? And I think Kevin knows... Better than me... And maybe anybody else here... That we got a lot more in store... At TanStack. We're trying to... We're trying to... Pull back a little bit... And really... Get some things stable. So... TanStack starts... Going to be 1.0... Here in a little bit. We have the date. I'm not going to give you the date yet. But... We do have the date. And it's coming fast. And... A couple of other libraries... We're talking about... Either bumping... Either bumping from alpha to beta...

[48:16] Or beta to release candidate... Or 1.0 soon. So... Get ready. It's going to be... It's going to be a good holiday. Let's put it that way. Okay. Yeah. Are you silent tonight? Well... Can you hear us? At all, Tanner? I messaged you on Discord. Yeah. I don't think Tanner can hear us yet. Kat's just listening. That's blind here. Yeah. Well... That was a long... That was a long... That should have been like a podcast. You know? But hey... At least it's being recorded. That's pretty sweet. Um...

[48:57] Yeah. I don't... I don't really want to... I don't really want to ramble anymore. Might start getting like... Emotional or something like that. And we don't want to have any of that. Okay. So far. I say we do some Q&A. I'm... I'm wrapping up probably on the last... 20 minutes of my run. 30 minutes of my run. So... I'd love to get some question and answer... Rapid fire... Stuff going on here. If anybody's willing to talk. We all want to talk, Tanner. You can just unmute. And... All right. Let her... Let her fly. Can you hear us yet, Tanner?

[49:44] Put your AirPods in. Oh, man. Yes, Kevin. Okay. I don't want to get emotional. Oh, okay. Can you... All right. He still can't hear us. Come on. Somebody unmute or I'm going to go listen to Spotify. All right. Wait. Getting tired. We need a TanStack package to try and get Tanner to hear us. Yeah. Oh, no. Um... No. Kevin says, you can't hear us.

[50:28] No, I can't hear you. Why can't I hear you? It says nobody's talking. Yeah. X. Spaces. It says, listener, listener, listener. Everybody's a listener. People have been asking questions for a while. Oh, my gosh. I'm so sorry. Guys, I swear. I haven't been able to hear anything. Everybody can know that I'm fooling around with you. Yeah. Here. I'm going to... I'm going to... I'm going to post it on the thing so everybody knows I'm not... I'm not lying. All right. If I...

[51:09] I don't know how to, like, get this so it'll work. Anybody have any tips? I think if I close X and come back in, it'll just kill everything. Yeah, probably. I don't know. Let's see. Unless... No. All right. Hang on. Let's see. The spaces still have co-host. Oh, gosh. All right. I'm doing co-host with Kevin. All right. I'm writing co-host. Kevin's co-host.

[51:54] Okay. I'm just going to quit my app and come back in and hopefully that holds it in place. If not, I love you all. You've been fantastically patient with me. So sorry. Nobody can hear me. All right. Here we go. All right. Let's see if he rejoins. Hello.

[52:38] Hello. I think it's still just me. Just Kevin. Hey, Kevin. What are you working on at the moment? Um... Past week, I've mostly been focusing on a TanStack pacer and hotkeys. I'm just getting... Hey, Tanner. All right. I can hear you. Oh. Yeah, there we go. I'm sorry, guys. I'm sorry. I was trying to communicate through emojis, but I don't think that was working. Yeah. Well, I was like, why is everybody making these weird emojis? It's like sad to your eyes and sad faces.

[53:22] I didn't think I was like digging in that deep, you know? Oh, man. It's good to hear some voices. God. It's been... It's been... Five miles just in silence. Jeez. Yeah, I'm impressed. I really enjoyed, yeah, hearing that history and your work with, yeah, visualization. I just had a question around 3.js. Yeah, what are your thoughts on, yeah, TanStack integration with it? Have you been playing with it? I know all the agents have, but yeah, just curious on your thoughts on that. I mean, I feel like 3.js is just like foundational at this point.

[54:08] And it works great with everything. I don't think... There's not really a lot of libraries that I'm like, oh, that doesn't work well with 3.js. Because it's just... It's primitive enough, you know, that it's just... It just works. I think it was primitive enough that AI now loves it. Because it's not trying to undo a bunch of weird abstractions. I think sometimes you get into like... React 3. Or what is it? The 3. The React version of it, you know, the React wrapper. But... Yeah. Oh my gosh, the mosquito. The mosquito truck just drove by and I'm breathing in deep.

[54:51] Oh my gosh. I'm gonna get cancer. Jeez. I live out by the Salt Lake. So it's like... They have these mosquito trucks that go around and just like decimate entire colonies. Oh man. No, 3.js is off. I actually built the game. If you go to tanstack.com slash explore, you can see some... Some early... Some early... Odd... 3.js shenanigans. I built a little game with a tool up and...

[55:36] It's pretty sweet. But some of the stuff I've seen lately on Twitter... Like just today I was looking through some stuff. Just couldn't believe what I've seen. I was like, holy cow, did you guys... This hadn't come out of like Unreal Engine or Blender or something. Nope. Just a bunch of like... AI generated craziness with 3.js. Yeah, that's great. I'm just checking it out now. That's really cool. The explore page. Yeah, I got bored for a day or two. Besides it. Push the limits. That was back when... AI was...

[56:18] Amazing. But... Also still kind of sucky. So... It took... It took a better part of like... Like 3 or 4 days of playing around with it to get to what that was. But now you could generate something like that and like way better. In like probably a couple hours. Tell it to use some like crazy... Crazy shaders for the water or something like that. I don't know. It's not a bad idea. Like a first person view in the boat. Start recreating like... Black flag or something like that.

[57:01] Yeah. Wow. I just had a curiosity. Since I couldn't hear any of you. Was it... Was it hard to follow up? With me running? Could you... Could you visibly hear the strain in my voice? Yeah. You can hear it bounce in your step. But... I can still understand you. Oh, good. Could you hear the cars? Were the cars just like blaring? I'm not... No, I couldn't hear any cars. Oh, good. Good job, AirPods. Because I was on a busy road for a little bit. And I could barely...

[57:41] Hear myself talk. So... It's pretty impressive. Oh, man. You know what? I might do this for the half marathon. It's not a bad idea. Like... If I take the half marathon and say, okay... My goal is to beat... It's slow. Two and a half hours. Or I think it was like two hours, 15 minutes. Something like that. I can't remember what my last one was. And then I'll just split it up and say, hey... For one hour... I'm gonna go... And just talk... To...

[58:24] Codex. And I'm gonna build something. Just during... In the middle of the marathon. And then halfway through... I'll launch it. And then I'll switch over to spaces. And we can talk about it. But I know that I'm gonna have crappy service where I'm going. In fact, I might not even have service at all. So... That's not gonna work. So I'll be back to my... I'll be back to my 165 beats per... Beats per minute playlist. I'm a sissy runner. I think I...

[59:05] I think the fastest I can probably sustain a half marathon is probably like... An 11-minute mile. And that's optimistic. I... Kev... Did you... Did you already talk about anything you're working on? I got like two sentences in. But... Yeah. Hotkeys and Pacer for the past two weeks. I got back from traveling. Yeah.

[59:48] Actually, like... I've been using hotkeys where I work. Again. So like... I love it. Every time I work on hotkeys, it's pretty much just because where I work needs improvement. So... Yeah. I love it. I drop it into everything I have now. And I legitimately just tell the agent, like, hey, put it into my context somewhere. I don't care where. But just like tell it. That anytime there's something that should be a hockey, just do it. Just make it work. So... And use TanStack hotkeys. And it's like, aye, aye, Captain. Yeah. Although... So it's interesting you bring up your agent. Because so... I think so far, AI misuses TanStack hotkeys more than any other TanStack library right now.

[60:31] Really? What does it do? Anytime I've tried to add TanStack hotkeys to a library, it just creates tons of facade code around it. Like trying to force it to do things that it already does in another way. Part of it is we don't have TanStack intent skills set up for it yet. I also just... Oh, yeah. I just did a big internal refactor and a bit of an API change, too. That should be coming pretty soon. Yeah. Well, you know... I've been using it without looking. And it's working great.

[61:13] Yeah. I think Alec has a question. Sure. Yeah. Hand raised. Oh, sorry. That was just from before when I was trying to get your attention and you couldn't hear. Oh, okay. Oh, you just wanted to say hello. All right. About... 10 more minutes. Five minutes, 10 minutes. I don't know. Depends on how fast I can run this lab. But does anybody have any, like, existential questions? Really tough stuff? You know, question reality type things that they want to ask?

[62:00] Okay. Yeah. What is your code review process looking like within the team? Depends on the project. Right. For a lot of our open source libraries, we still enforce human code review. It's AI-assisted, for sure. But... It's definitely still very human hands-on. Just because we're dealing with... We're dealing with libraries used by a lot of people. Yeah.

[62:43] Security. You know, we can't just blindly just let anything happen, you know? Yeah. We'll throw in. Like, we have a lot of code owner stuff. Like, certain folders in every project is, like, highly scrutinized by all core maintainers for security reason and stuff like that. So, yeah, there's parts of it where, like, you can't touch the files unless you've got, like, two people turning their keys at the same time type of stuff. But there's other areas where, you know, like, I'll get a pull request from Allem for TanStack AI, and he's like, oh, hey, like, fix this annoying bug.

[63:27] Got to get this out. So, I go check, you know, I check out the PR really quick on GitHub and just kind of scan through it really quick. And I'm like, yep, it's good. And we merge it. And I feel good about that because it's not a backend library. Some of it is, but, like, the stuff that's not in our code owner is... It's pretty safe. So, it just depends on the library, right? But for pretty much all of them, when it comes to anything that has to do with, like, release, CI, publishing pipeline, it's super locked down.

[64:09] When it comes to the website, tanstack.com, well, that's just up to the individual you ask. Everybody on the team will say in a heartbeat that I'm super fast and loose with the website. And I am. And I love it. And that's not going to change. Because, you know, as fast as I can break it, that's about as fast as we can fix it, too. We always have instant rollbacks and everything, too. So, take some stuff super seriously with the website. If we're doing, like, SQL migrations or anything like that, take that pretty seriously.

[64:51] I mean, don't tell anybody, but I have admin privileges over the website. Man, I'm still merging right to main from my machine sometimes. And that's not manual. I'm just like, yo, GPT, fix this and merge it to main and use my admin override. Wow. And it's like, got it. Like, I had to do that the other day. Had to make a change to one of our partners that was overdue displaying partner logic. And they're like, hey, we need to get that updated. They're like, right now. I'm like, you're right.

[65:34] Codex. Bing, bong, bam. Get it done. And within, like, two minutes, it was fixed. So, pretty cool. Yeah, I feel like it's flipping around instead of, like, my company, you know, we're building kind of, you know, integrating with CMS project, you know, set up for non-technical members to, you know, change things on the website. But it's just, it's flipping around now to be, I couldn't, I could just get a code change in faster than it would take them to log into the CMS and make the change. Right. I think that defensive measures have gotten even better to, you know, I'm looking into, like,

[66:23] self-healing. I want it to be like, like, before it was like CI, you could set it up so that if you launch you caught a bunch of stuff, you could wire it up so that it would just, like, roll it back immediately and alert you. Instead, I just want to have, like, Jev running through the website all the time. Catching stuff. And if it finds something that's from one of our, from one of my PRs, it's like, roll it back and tell me. Or better yet, just spin up an agent and go fix it. So, I know that's pretty wild, I think, but a lot of people are doing that. Yeah, that's what really interests me.

[67:05] I saw a software factory ideas. Yeah. It seems where all of the fun's going to be now, now that AI's taken the coding away. I haven't come around 100% to software factory yet. Yeah, it's very new. I'm still very much in the loop. But I'm getting there. I didn't want to jump there out of the box because I wanted to understand the processes really well. Definitely. But I'm, I'm definitely getting there. But I saw something today that's, like, taking it too far. Somebody was like, oh, yeah, Vercel Labs. They're like, oh, yeah, we're a software factory. And now they've got their software factory telling them what they should build next.

[67:50] And I'm like, yeah, that's too far. I'm like, well, so, so you don't do anything, you know? You just, you just approve scope creep, basically. You know? I don't know. So, anyways. That's not fun and not safe. No. But, like, TanStack charts, like, it's weird because I haven't looked at any of the implementation code in a long time now. So, if there's a bug, I'm just fixing it with AI, you know? I've got all my testing harness set up, everything, like, tons of tests.

[68:31] Integration testing. Playwright testing. Like, all this stuff. I've got budget thresholds that it has to match. I've got just all sorts of stuff now. And I can just go in and just say, hey, you know what? Somebody requested this feature. And I've got a skill. I've got a skill specifically for implementing features that respects everything about the project size. One of my skills is implement this feature without increasing bundle size. Without making any breaking changes. And, like, a couple of other things.

[69:12] So, I'll just go in and say, hey, I approve. Like, I approve this feature. Here's some things I want to change about the way that it was conceived. This is the API I want. I'll make it happen. And it'll either come back to me and be like, success. We did it. I'm like, wow. Good job. Or it comes back to me and it's like, hey, we were able to implement it, but we had to add a bunch of bundle. We had to add a bunch of code to do it. And I'm like, oh, well, tell me why. And it's almost like this little therapy session between the model and me. I'm like, oh, really?

[69:53] Tell me why. Tell me why you had to do that. It's fun. All right, guys. I'm coming up on my house. It's time to shut this down. Sorry for the silence. Hope I didn't drive too many people away. Thanks for hanging out with me, Kev. Thanks for rescuing the stream. Thanks, Tanner. Thanks, Kevin. It was great. Well, see you around. See you, Tanner. Thanks, guys.

[70:29] [Closing audio not reliably transcribed]