Jesse

@lucasmeijer Try @capydotai

Who has a really nice setup like this? - Cloud based agents - Kanban board-esque overview - Full control over agent loop

@theo idk i read it as youre "incentivized to make it seem like the end of the world" for views, which doesnt seem like an accusation and more like an understood truth about the economics of content creation. no one will click on an even keeled, "it's ok, not great" take

Kind of wild to be told I have an "incentive" to shit on Anthropic by employees when I defended them against the world last week w/ the Mythos stuff 🙃

@aarondfrancis giving soloterm a try and its really nice! any reason you didnt go with libghostty for the terminal though? as a neovim user it wouldve been nice to have soloterm as the one stop shop for everything.

@mitchellh this means you have to use hyper as your terminal now

Excited to share that I've joined Vercel's Board of Directors. Vercel is made up of builders and tastemakers that continually ship things that deeply impact how developers work: Next.js, AI SDK, v0, etc. I can't think of a more exciting place to be. Let's fucking ship. ▲ My relationship with Vercel goes back to the earliest days. HashiCorp was an early adopter of NextJS and Vercel (~10 years ago!) and it remains my default tech stack and deployment platform to this day. Ghostty's website is all on Vercel, too! Beyond that, I've been continually impressed with the teams relentless focus on shipping meaningful software. And importantly, software that has incredible taste. Now we are in the age of agentic software development. Vercel is building agentic infrastructure that I think every app and agent will need (I certainly need it!) and I can't think of a more exciting place to be. Huge thanks to @rauchg , Jeanne, Marten, @cramforce, @tomocchino and the entire Vercel team for the warm welcome. Time to work.

@swyx @dillon_mulroy @badlogicgames if youre interested in how you can use /tree and /fork as a lightweight replacement for subagents, check out stacktoheap.com/blog/2026/02/2…

@dillon_mulroy explain the context mgmts thing more? also can u explain the webfetch websearch thing cc @badlogicgames

thoughts after day 1 of using pi full time - less is more - i don't miss subagents like i thought i would - /tree is an insanely good context management primitive (and partially why i havent reached for subagents yet) - based only on vibes, i think having a minimum system prompt is improving code quality - telling pi to copy opencodes webfetch and websearch tools was a good play

@swyx @dillon_mulroy @badlogicgames re: websearch, pi doesnt have it built in, but you could just point pi to the opencode webfetch tool on github, and say "build this as a pi extension" and then boom, you now have webfetch. it's the pi philosophy to use the harness to build extensions to enhance itself.

@banteg I’m mostly using extensions I’ve picked up from other setups, but I did make 1 extension for myself as a vim user: create a a file I can load into neovims quickfix list to view all edits from an agent session. Niche, but works for me github.com/jssee/pi-conf/…

am i dumb for never extending my pi agent? i mostly prefer vanilla tools if they do the job. i just don’t see anything missing from it, looks like a perfect reliable minimal harness. tell me how you extended it, maybe im thinking too small.

Come and join me on.design 👀 on.design/claim?ref=jesse

@badlogicgames any plans for pi to support reading skills from the `.agents/skills/` directory like amp and opencode (and others) do? i feel like i saw a post mention that it was planned, but i dont see it in the docs.

@badlogicgames oh now, please close more

Two more and then I'll close the PR tracker forever. It's surely complete software then. No more bugs or features.

@stolinski @intellectronica Yeah can here to mention agent-browser too. I prefer the “headless by default” approach vs mcp and it can even do recordings and screenshots to help agents verify their work.

@intellectronica Dope. Will give it a rip today.

These are the AI tools I'm currently using. - OpenCode - Dex - Sentry - @sveltejs/opencode - Opus 4.5 of Codex 5.2 - chrome-devtools mcp - A few custom agents I'm sure I'm missing out on things by not using skills more or various tools here and there, but honestly I feel very effective with what I'm doing.

@badlogicgames I 'forked' this and piped the changes to `lumen diff` via a floating zellij pane. fun. gist.github.com/jssee/465d0c96…

I love writting tiny pi extensions like this. github.com/badlogic/pi-mo… Lists all files the agent touched via read/write/edit, newest access first, opens the file in VS Code. Took 30 seconds for the clanker to write, hit /reload, continue work uninterrupted.

@stolinski i've moved to ticket: github.com/wedow/ticket its straight forward, portable, does what i need. even has a `migrate-beads` command to help with the transition. cc @gregwedow

If you aren't using or don't like Beads, what are you doing instead? I haven't personally found .md files with roadmaps and tasks to be as effective or reliable.

@benhylak Worktrees make sense to keep multiple branches checked out at once i.e if you have 1 branch in PR review, 1 WIP and 1 for reviewing someone else work locally. Worktrees makes this slightly easier but imo requires some scripting around the default impl to actuallly feel good

i've been using git for a long time and i still don't get why i should use work trees instead of just copying/pasting the folder

@dillon_mulroy Nice, reminds me of typescript-result.dev which was a favorite of mine (as opposed to neverthrow)

introducing better-result, my opinionated take on the rust result type for typescript after having written about 100 times npm i better-result *disclaimer*: if you can use effect-ts, you should use effect-ts

@kentdebruin obra/superpowers In particular the brainstorming skill, typically moving things to beads after the implementation plan phase.

What are your favorite Claude Skills so far?

@joelhooks Sick, yeah I typically pick one of two from his list based on his reviews. I don’t read novels that much so that’s enough to last me the year lol

@juhsee ya, his cuts go deep so i just get the full stack i do compare ratings on amazon and goodreads, but it aligns

hell ya

all my tall niggas be goin 2 sleep early 😭😭 tired as shit from standing

@mitchellh @ogirginc @flyasajet What are post facto ways? Im a daily ghostty user since beta and I don’t exactly care about scroll back search, but post facto streaming output sounds cool

@ogirginc @flyasajet My personal favorite is piping streaming output through fzf but depending on the use case openinf command output in an editor too (ghostty has provided post facto ways to do this since 1.0). I don’t use multiplexers

I'm not even joking I went to this ~300 person wedding last weekend, I'm in a sparkly tux, the music is on, dancing is starting, someone comes up to me and whispers in my ear "if you're here, it means Ghostty still doesn't have search" and walks off. Oh no! 🤣