Jonathan Carter

I feel like the process of learning codebases should be more guided, interactive, and integrated into your editor. Not buried in documentation. To improve onboarding, and make it easy to create such learning assets, I created a "tutorial recorder" for @code. It's called Code Tour

@eliostruyf Let us know if you have any feedback after giving it a spin 👍

This new Visual Studio Code Agents app looks like my new #GitHub Copilot companion app. 😍 #VSCode

@davidfowl Scratch is a good example of what can’t just be vibe coded away, since its community has always been the most valuable part. Without the intrinsic motivation to share and discover and remix, I doubt my kids would have gotten into it so much

RIP scratch

@elmd_ @code If you use the VS Code Insiders build, then the experience should be even better 😁 The team is shipping updates to the Claude integration pretty rapidly

Unpopular take: You can just use Claude Code from @code directly. No fork required and it's better than the official extension IMO.

🙌 You can now use @claudeai and @OpenAI’s Codex in GitHub and @code with your GitHub Copilot Pro+ or Copilot Enterprise subscription. Define your intent, pick an agent, and they’ll get to work clearing backlogs and bottlenecks, all within your existing workflow. github.blog/news-insights/…

VS Code just announced MCP Apps support! 🔥 Here's one I just made that visualizes git commit history. You can also multi-select commits and have Copilot generate release notes for you! Check out how to get started with MCP Apps in VS Code here: code.visualstudio.com/blogs/2026/01/… #vscode

Today we're open sourcing a technical preview of the GitHub Copilot CLI SDK. Build agents with custom tools in Go, Python, TypeScript, and C#. Built on the same agent loop that powers the Copilot CLI and GitHub Coding Agent. Supports BYOK, and any model. Here is the Copilot CLI driving Excel:

@NickADobos I’ve been pushing on the CLI tabs++ UX, since I think there’s still a lot of gains to be had with that foundation: github.com/lostintangent/…. And it’s been really cool to see all the exploration in this space

What will the ideal AI agent orchestrator look like? - starcraft, age of empires, factorio - list of chats, aka cursor - CLI tabs - linear ???

@GeoffreyHuntley After building my own editor and terminal, I’ve been recently building my own terminal shell. The later feels totally superfluous, but it’s been a lot of fun 😁

alright; in the spirit of professional development in the form of play reply back with unhinged things you have got an llm to build for you me: a file system, a programming language, collection of network based protocols and countless transpiles from one lang to another lang not a competition though the spirit is play. you can build anything now what have you built for shits and giggles? why did you learn from it? from my pov i suspect people aren’t pushing the tools and models hard enough. if you just use them in your day job then you are missing out on some cold hard facts you can do anything now what have you been putting off on building? build it

@ishaansehgal @samuelcolvin @claudeai This looks really cool! One thing I noticed: is there a way to prevent the plan mode results from being automatically approved? I can appreciate why that’s useful for background jobs, but it’d be cool to try out ideas on the go, where I just want a plan first

@samuelcolvin @claudeai exactly why we made omnara, claude code from your phone

I want to be able to approve @claudeai code from my phone when I'm away from my desk. I run claude exclusively in ghostty, anyone know of a way to get these notifications and be able to hit "yes" on my phone - e.g. via slack?

We now support Agent Skills - the open standard created by @AnthropicAI for extending AI agents with specialized capabilities. Create skills once, use them everywhere. 🔗 aka.ms/vscode-agent-s…

A lot of critiques of vibe coding actually don't cut deep enough, imo Sure, "the AI wrote bad code" may be true...but that is changing fast! The deeper issue is that when you're designing a complex system, you need good mental models of how it works in order to invent great ideas for the next iteration. Until now, if you built it by hand, that would guarantee some amount of understanding. But now the two have been decoupled, so we need to avoid the temptation of skipping understanding Reminiscent of the debate around whether memory matters in the modern era. It turns out you can't "just Google facts" -- you actually need to have ideas represented in your head to work with them fluently. The good news is AI can actually help us develop understanding faster and better, if used correctly! So the takeaway isn't to stop using AI for coding. It's to find ways to use it to develop understanding and move faster

Over the weekend, I nerd sniped myself into building a custom terminal shell (as a compliment to Conduit). That's turned out to be a really fun project, and I'm officially a big Zig fan now. But the rabbit hole of complexity is pretty deep 😅

@iamsahaj_xyz Yeah I love the low-ceremony feeling of terminal-centric agent flows. I’ve been playing around with a “workspace manager” to see how it can feel to compose and automate terminals/editors/browsers: github.com/lostintangent/…

once again, I'm fully terminal-pilled nvim + claude-code + tmux + lazygit + ghostty couldn't be happier

New release coming tomorrow and its getting late, but in the meantime this is your friendly reminder the Copilot CLI can help you submit GitHub issues on your way to bed.

`copilot -i {prompt}` is coming to GitHub Copilot CLI in .366 so you can jump right into a session....but what can you do with it?

Over the last few weeks, I've been hacking on a tool to organize tasks with coding agents. It allows designing workspaces of terminals/editors/browsers, is automatable, and let's you zoom in/out/across work. As far as personal software goes, it's a fun example of what AI enables

@kcosr @notnotstorm Awesome! This is really cool. Conduit does a similar thing, by exposing a CLI into its hosted terminals, which allow discovering and automating the surrounding panes, which includes terminals and browsers (like a mini playwright). But I’ve got to tune it more for agent usability

@LostInTangent @notnotstorm It just provides a basic interface to my API. This shows the instructions I include. Also every session I orchestrate has some env vars set that the CLI uses directly to simplify things for the agents. The token is generated for the session and allows the agent to auth.

when running 24x claude code instances makes sense: 1. an initial agent scanned my repo looking for general improvements. it flagged 20 things. I liked 12 of them and told it to create a github issue for each 2. I opened up 12 tmux panes and ran `/fix ` in each one. this is a slash command that fixes a gh issue in a new worktree and submits a pr 3. I split each of those panes in half and ran `/review ` on each of the 12 prs 4. I ran `/respond` in each of the 12 original panes, to respond to the reviews and update the pr's with any necessary fixes 5. I ran `/summarize_prs` to figure out the best merge order and flag PR's that might have bad concerns or tradeoffs next step is decide which PR's I want to merge and merge them it's still a very manual process but every week I feel like I'm getting a lot faster

@kcosr @notnotstorm Yeah that makes sense. I currently rely on the serialize and search add-ons for xterm, and those don’t seem to be supported in ghostty-web yet. But I’ll be keeping an eye on this, since I’m a big fan of ghostty, and would love any free perf gains 😊

libghostty is supposed to be really performant. It can take 5-10 seconds to load some really long agent sessions when I attach (due to the PTY replay) so I thought I'd see if this was faster. Also xterm.js is occasionally janky in my app with control chars sometimes appearing and things getting corrupted so I have to clear history or refresh the page. It's probably something I'm doing wrong, but I haven't been able to fix it so I thought I'd try this. It looks like it's going to take some effort because I have mobile touch stuff build around xterm.js. So I just upgraded xterm.js for now and will see if it's any better. The version I pulled into my vendor dir was pretty old.

@kcosr @notnotstorm I’d also love to hear more about the CLI you have for agent communication. I’ve currently got a CLI for this as well, and it allows agents to enumerate other terminals, and read/write to them. And it also allows spawning new panes/tabs as needed.

@kcosr @notnotstorm This is incredibly useful! Thank you so much for sharing 🙏 out of curiosity: what are the main advantages of switching to ghostty-web? And in practice, how much of an xterm drop-in has it been for you?