953: Why v0 creator left Vercel to fix GitHub (GOAT Jared Palmer)

I'm in. Hey Jared. How's it going? Thanks a lot for coming out. Thanks for the response. Yeah. Sick. Great to see you. No, I don't think so. We just finished recording our recap and I was like, let me check my DMs. Yeah, just there. It's fine, I think. Yeah, anywhere you want. What's up, everybody? We got Jared Palmer in the house today. We're at GitHub Universe. Oh, he needs to pull him in. Jared just popped in, and we're stoked to talk to him. You recently started...

a job at Microsoft slash GitHub and are fixing all the problems is what's going on. I don't know about all the problems. Yeah. Yeah, I'm doing my best. Yeah. I'm like... It's like day 13, 14. I was going to say, yeah, it feels like it's fresh. It's fresh. It's fresh. So I'm still acclimating. Okay. But there's a lot of shared DNA between – previously I was at Vercel. Yeah.

and Guillermo, CEO of Resell, and Nat Friedman are very close friends, and Nat used to be the CEO of GitHub, and so there's a lot of corporate DNA, I think. that is permeated through the evolution of the companies and so and there's been a lot of uh former githubbers that were that came to versell yeah uh and so that's it's it's there's hints of similarity, if that makes sense. Yeah. Still getting acclimated. Totally. That's awesome. So you're the creator of VZero, right? I am. I'm...

I'm curious to get your thought on just workflow in general with this type of stuff. So developers are listening to this podcast. We saw all these new tools be announced today. You obviously have a lot of experience in the V0, where it's like... happens in the browser and then at some point you you bring that out of there like what are your thoughts on like what the developer workflow will look like with all these agents um yeah i mean i think that

we need to do a better job as a community, building out like continuity, if you will, right? Yeah. Remember, it still feels like, remember before, it still feels like maybe... I'm going to date myself. In high school, when you worked on a word, we didn't really have like Dropbox, right? And like the pre-Dropbox days.

I know that it's like dating myself there, but it does kind of feel like that where these are so isolated and they're not fluid. There's no continuity. Your flow state gets broken. You start one task and you go off.

you kind of forget yeah it's called final.js drag and drop underscore underscore final final yeah yeah totally so so anyway so I think that in the future And ideally, one of the things I'm focused on is bringing continuity across all of the Microsoft, if you will, touch points, which is like. And that's what's been so important about this reorganization around the core AI unit, which is Visual Studio, VS Code, GitHub, and Azure AI Foundry. Okay.

and parts of Azure too. And the idea here is that we should be working really tightly with the VS Code team at GitHub and we should be making handoff super seamless. Right. And so when you start tasks in one, you should be able to pick them up on the other and vice versa. And you're seeing that today. And the other thing is that model choice is real. And I think developers want that choice because they always want to use the bleeding edge.

That's one of the announcements we had today that we're bringing third-party models to GitHub co-pilot and GitHub. Agent HQ, which is the new control plane from which you can kick off kind of background tasks and work through PRs and other things like that. So I'm really excited about that future. And then I don't know if you caught it in the in the keynote.

If you do start a task in AgentHQ or on GitHub, you can open that in VS Code. Yeah, that was a killer thing to me. I think that was the most encouraging new feature shown today. Just because, like we were saying, it's like so many times... when those responses come back. If it's unusable at that point, you're like, okay, now I'm forking this. Or I scrap it entirely. Let me type in the box again and let it go run for 15 minutes and then come back to it. You lose that flow state. Totally.

One of the things that I did testing that feature. So last week, Next.js announced version 16. Very exciting. You guys probably covered it if you didn't. Whatever. But I was excited. Scott's a huge Knicks fan. So the first thing I did was, and I thought this was pretty cool, because one of the cool things about AGNHQ is the prompt is actually set up to create PRs, which is a little different than...

like the VS Code, super interactive, fast Twitch, if you will, like cursor. Yeah. Not background, but like cursor sidebar situation. Yeah. The prompt that I gave was. Research the latest Next.js 16 release and upgrade guide. Then run all the code mods and upgrade my site. And it just did it. And it created five different commits. It created a draft PR.

And it did take, I think it took, I don't know, it's like 10 minutes, but that's okay. But it did it. And then it ran all the, it not only did that, it ran all five code mods, migrated from NextLint to the new ESLint CLI. It verified it migrated to TurboPack, changed configuration, validated the builds, even included like now built in nine seconds of TurboPack. Anyway, I guess the point was like, I don't think a year ago we were.

at this point. Oh, yeah. No, a year ago, we were sitting in this podcast booth talking about like better tab completions. Or tab completions at all. Yeah, yeah, yeah. It's kind of amazing, right? And so, anyway, to your point though, if that PR is like, 80% ready and we don't have an escape hatch. That's just like so much product pressure.

on us to like right so this new open and vs code fit feature i'm very bullish on because it just keeps you in the flow right yeah it's kind of like horseshoes and hand grenades we only need to get a little we kind of need to get close yeah yeah yeah So I'm very excited about that. Yeah, me personally, I think that's the thing I'll probably end up using the most beyond bringing in...

Claude or any of these things, because I was always assigning things to co-pilot out of issues right now. But being able to, like you said, model choice is important. And now it's less of, hey, I'm assigning it to co-pilot and I'm assigning it to what I'm used to. I'm assigning it to what my flow is.

which is great. I think that's awesome. And being able to use Cloud Code or Codex or whatever agent you want, can you talk about the background behind that as well, having some choice? Yeah, I think it's important that we...

From my short time at the company. Yeah, I was going to say, 13 days, yeah. 13 days. But obviously, like, I talked to... to Satya before joining and we had talked about some of the these these big decisions right for the platform yeah whether we should like allow third parties and stuff like that and we both agree I think it seems like this past

at least like three to six months have shown that like giving model choice is what developers want. It is ultimately leads to the best product. And, you know, if you're talking about GitHub's mission, which is to empower developers to build awesome stuff, well then like.

yeah they should have model choice yeah it's kind of a no-brainer um and we'll do well if we stay true to the community yeah and focus on developers and everything else will take care of itself yeah oh that's great that's great so

Once you got hired, you went on a tear on Twitter. I've been throwing heaters. What can we fix? How can we do it? Are you just coming in and fixing all the rendering issues? You have a pretty... Not aggressive, but, like, you're good at getting shit done, right? Obviously, Vercel was like that as well. Yeah. Thank you. I... A couple of things. I think that feedback is a gift. And we've got such passionate developers in this community and they have such awesome opinions that like...

Sometimes you just got to ask, like, how can we make GitHub? I don't think I don't know of any time anybody from GitHub has asked that on X. Like, how do we make GitHub pull requests better? Yeah. Right. So like, how else will you know? Right. And it turned out that the top ask was stack diffs. Yeah. Right. Which I kind of knew was coming, but I was I was and I did I did ask. I can get into that in a second, but I'll answer it after. But no, you know.

I've got a lot to learn. It's my 13th, 14th day here. And GitHub is a very... It's like a legendary product, let me say that. Yeah, it's an institution. It's an institution, right? So change is coming in the sense that we want to make sure that we are evolving with the times and building the best possible developer experience we can. And I think that's going to also change because...

Just like we discussed just a couple of minutes ago, we went from autocomplete to almost like a co-worker. Yeah, robots going off and doing work. An intern. Let's call it an intern. Fire it off from your phone. That's just different. And also it's going to require...

a different workflow. And so some of the stuff needs to evolve with that new workflow. And we're committed to seeing that vision through. And so I think about what does it look like? What is a pull request in 2027? And I don't know the answer yet, but I'm pretty sure GitHub is going to have to answer that question.

yeah yeah yeah it's gonna look it might look a little bit different a little bit different and that's exciting and I think that's what we're driving and focusing if that makes sense cool and then also to the performance side like this is something that's near and dear to my heart And so we're also working on that, too. That's been in the works. But I'm really excited about the team and the stuff that I've seen so far. Yeah, but now you get to show up and take...

credit, right? Because that's your thing. Right. Show up and take credit. Yeah, exactly. I contributed to not as much as you could expect for this, but I'm very happy and proud of the team. They put in a lot of effort. A lot of the stuff you're seeing today was started months and months ago, obviously. And so happy to cheerlead. but also like deferential to the team. They kick butt. All right, last question. You need to build something.

What's your interface? Where are you starting? Are you starting in V0 in a browser? Are you NPM scaffolding something? Are you kicking off an agent? What are we building? You're building a podcast player website. Wes is asking for personal reasons here. He wants to know where to start. What are you guys? What would you use right now?

For me, I would NPM scaffold out something so that I have my like basics, you know, like I have them in place, maybe get feet and everything running. Would you do that by hand? I would. I usually do that by hand because way too often it picks the wrong tech. And would you like install like ShatCN or like. Yeah. Yeah. I install all the depths that I decide I want. Yeah. Which I know was part of the announcements today as well, which is planning mode.

once I have that, I'll, I head into like the agent tab and start kind of going back and forth with, with agents and it scaffolds it out and I kind of talk back and forth to it. So that's my UI, but I'm always curious about like, I do that because like, I'm a.

I was a developer before this stuff, and that's what I'm comfortable with. I'm also curious, like, people who are not necessarily, or not that you're not a developer, obviously you are, but people who are new to this stuff, like, what does their UI look like? And I'm always curious to see what people... What about you? Tackle, yeah. What would you start with? Yeah, I use Svelte for everything, so I start with, like, SVE.

There's a number of options, boilerplates and stuff that get going. Is SV the Svelte CLI? Yeah, yeah. And so you couldn't even get database and all that stuff going right from there. Shout out to Rich. Yeah. Oh, yeah. And by the time you're done with that, the CLI you have. Oh, pretty much everything like ORM, everything configured, database, you have your containers all set up. And just from there, that's when I'll hit and go into a planning mode or I'm a little bit more on the.

over-engineer on the spec-driven mode myself, so I'll start getting into specs and then go from there. Yeah. Yeah. So to answer the podcast question, I think I would do the same thing. I think I would start out with like my I would scaffold. Yeah. An XJS app biased.

slightly biased i would but you know ssr it's important yeah i would get like um like biome set up and like like everything like to that level yeah that makes sense i would install shad cn i probably would pre-install the sub subset of the components that I thought would be useful in the UI.

I actually did this, by the way. This is a specific question. For the old undefined site, I actually use that sometimes. It does like a... sort of demo thing and like i actually migrated it was it hadn't touched the site in so long so it was in gatsby and i did the migration to next and what i did was i think the tailwind guys did a a podcast template in one of their um oh yeah which i bought but then

I copied that thing into a folder. Tailwind UI, yeah. Yeah, I copied that into a folder. I downloaded a folder and then I was like, yo, look at this folder. But one thing I'll give as a tip for those who are prompting, it doesn't matter what. which one of these agents you use a trick that um actually picked this up from the claude code team that works really really well is to prompt

Even before you either write a spec or actually go to execute is to ask the model to do some research. Yes. Yeah. It's so powerful. It's been powerful for me. And even like getting into like database design or those types of things to pre. research like go research this yeah like

Get a sense of my code. Even that, those two prompts, go research this and get a sense of the code base or paste in a bunch of links and get context. Packing that context can be so powerful. And then either write a plan.

There are also some special keywords, by the way, for a lot of these models. I don't know if you guys know this. No. Okay, so have you heard of Ultrathink? Yes. Okay, so you can say Ultrathink, and that forces, usually force, at least in the anthropic models, will force high reasoning. Okay. Another sort of secret tip I'll share, which I found useful, something like emergent behavior, is if you ask these models to add logs and assert statements.

which is like low key, like not many people use these days, but add assert statements and logs to maximize debuggability. That can be very helpful when you go to like hit it out. And the third one that I found interesting, and this has helped me with planning. When I'm getting it to write a spec, I'll put that in a markdown file like to do or plan and sort of ephemeral. And I think we need to develop as a community like a better way of doing this. It's absolutely so kind of out there. Yeah.

But, like, maybe that's beautiful, but it's kind of... It just seems, like, not right. It seems like... I've, like, redone my own custom... you know, spec flow like eight times. Yeah, so there's like some spec, figuring it out, but let's just put it in a markdown file. But then what I do is I ask it to create a rubric for grading the spec or this plan. then grade it, then revise it. And that extra step of like it thinking through the rubric and then self-evaluating I've found to be quite useful.

it can go overboard and so it ends up like i've probably seen this too it's like week one to two oh week three to four like i haven't figured out how to get it to stop planning like 12 weeks oh yeah what is that about we did like a vibe vibe what is that what is that yeah yeah

How do I get it out? Yeah, and mine was like, yeah, week one. Yeah, we were seeing who could get it done the fastest. Like, oh, can we get this done in 10 minutes? Oh, my gosh. And then it was like, okay, week one. I'm like, no. Why does it do that? I don't know. That and like the yapping at the end. Yeah, the yapping. Where it tells you what it did. You got to tell it, stop yapping. Stop yapping. Yeah. Just give me the code.

It's so funny. But, yeah, so if I could figure out how to not have it make a 12-week plan, that would be great. And I'm too lazy to, like, correct it. Yeah, no, you're just like, yeah, yeah. Maybe it's good. I don't know. So that's some of my tips. The rubric thing is helpful. Ultra think is helpful. And then from there, what I'm really excited about now is with some of these, I don't think Cloud Code can do this yet, but I know that VS Code can do this, I think.

I saw something, the cursor's gonna have this very soon too, which is the ability to have like N attempts or variants at the same prompt. Oh yeah. Just, you know, throw, throw money. Yeah. Yeah. Yeah. Just run them all away in like, like branches or something like that. I think they, yeah. Yeah. I think that's probably, I mean, yeah. Like let's just, you know.

Yeah, right, yeah, yeah. Whatever, we're going to rip. Fire three of them, hopefully one hits. Yeah, let the interns do it. Hedging your code. Put a team on it, yeah. Put a team on that. Anyway, so I think that's what I would do for a podcast site. That would be awesome.

Awesome. Well, appreciate your time. This was really good. I'm glad you could make it. Likewise. Big fan. It's finally great to meet IRL in the flesh. Yeah, absolutely. See you guys around and happy to come on again. Yeah, anytime. All right, thanks. Peace.