Geoffrey Litt's banner
Geoffrey Litt's profile picture

Geoffrey Litt

@geoffreylitt22,620 subscribers

malleable software @NotionHQ / prev @inkandswitch, @MIT_CSAIL / 🇯🇵🇺🇸

Shorts

casually making an evolution simulator in spare moments of holiday break... because opus 4.5 makes it so easy that it's fun!

casually making an evolution simulator in spare moments of holiday break... because opus 4.5 makes it so easy that it's fun!

160,115 Aufrufe

I wanted to convert a JSON file of a chat transcript into nice markdown text for sharing w/ people... so I had GPT generate an ephemeral React UI where I can drag in the JSON file and it outputs the markdown🤓 reflections on the process:

I wanted to convert a JSON file of a chat transcript into nice markdown text for sharing w/ people... so I had GPT generate an ephemeral React UI where I can drag in the JSON file and it outputs the markdown🤓 reflections on the process:

233,077 Aufrufe

underrated technique: vibe code a full-custom debug view for code you're writing by hand! in this case, visualizing a parsing algorithm 😎 a clear interactive visualization helps you really grok what's going on step by step, and of course helps with fixing bugs too

underrated technique: vibe code a full-custom debug view for code you're writing by hand! in this case, visualizing a parsing algorithm 😎 a clear interactive visualization helps you really grok what's going on step by step, and of course helps with fixing bugs too

76,484 Aufrufe

tired: AI generating Markdown docs to explain something to you wired: AI generating slide decks with interactive simulations for you to explore

tired: AI generating Markdown docs to explain something to you wired: AI generating slide decks with interactive simulations for you to explore

35,411 Aufrufe

was working on a prolog interpreter this week step 1: had Claude make me a custom debugger UI that could visualize execution traces of my interpreter! AI-generated custom dev tools ftw 😎

was working on a prolog interpreter this week step 1: had Claude make me a custom debugger UI that could visualize execution traces of my interpreter! AI-generated custom dev tools ftw 😎

93,980 Aufrufe

what if a chat produced a spreadsheet as the answer, so you could instantly tweak numbers and see the result?

what if a chat produced a spreadsheet as the answer, so you could instantly tweak numbers and see the result?

166,104 Aufrufe

Recently I've been hacking on a tiny collaborative Markdown editor at Ink & Switch. The neat thing: collaboration was trivial to add, because I built on automerge-repo, a new local-first library announced today! 🥳 Quick report on the experience and why it felt nice: 1/

Recently I've been hacking on a tiny collaborative Markdown editor at Ink & Switch. The neat thing: collaboration was trivial to add, because I built on automerge-repo, a new local-first library announced today! 🥳 Quick report on the experience and why it felt nice: 1/

42,382 Aufrufe

Videos

geoffreylitt's profile picture

✨New demo: what if vibe coding felt more visual? Brian Lovin Mary Rose Cook and I did a game jam using Notion as our "IDE": launching Cursor agents from a task board, and making a custom image for each task 😎 The demo shows 3 ideas for the future of agents: 1) Agents should collaborate across apps. Each app has its focus--Notion AI is good at drafting specs and organizing tasks; Cursor is good at coding. So let them specialize! Today we're launching a new integration where Notion AI can kick off Cursor Cloud Agents to do coding tasks. The Cursor API accepts natural language prompts, so I think of this as "cross-app sub-agents" -- it's kinda cute how it resembles humans hiring outside contractors 😊 BTW: the parallelism of cloud agents is incredibly freeing for creativity, but it also creates a new problem: sooo much work to keep track of! Which brings us to the next idea... 2) Agent orchestration is a data visualization problem. A powerful frame for designing agent UIs is to think of the chat transcripts as the "raw data" and ask: what visual projections might help people make sense of this data at scale? We need to engage our human GPUs -- our visual processing -- to understand what the computer GPUs are doing for us! One thing we can do is use AI to populate traditional UIs like progress bars and status updates. But there are also new possibilities now... For example: when you have a lot going on, it can be hard to identify tasks just by text titles. So we tried generating an AI image for each task -- turns out this helps a lot by giving it a unique visual identity! And of course, it also just makes it super fun to build with friends 😃 Speaking of friends... 3) The future of coding is collaborative. Sometimes it feels like IC engineers are being reduced to middle managers: shuffling information between the team's context and the coding agents that they individually manage. The solution: bring all the people and agents into one shared space, with shared context and visibility! In the video you can get a glimpse of how this feels. Mary, Brian and I record ourselves chatting about ideas, and then we use AI to turn that conversation into a list of tasks on a shared board. As the ideas get built in parallel, we can all monitor progress and review the work together, nothing is siloed. My main takeaway from this game jam was: damn, creativity with friends, at the speed of conversation, is incredibly fun. --- Our goal here is to let anyone use Notion as a fun and creative "software factory" to build software together with your team. Give the Cursor integration a shot and let us know what you think! (AI Image gen in Notion isn't GA yet, but coming soon and already out to some users) And let me know if you'd want a template or more detailed instructions on the setup we showed in this demo...

Geoffrey Litt

88,677 Aufrufe • vor 3 Monaten

geoffreylitt's profile picture

tired: Claude Code ports my personal website to a new framework wired: Claude Code generates a command center UI where I can execute the port! with live feedback, including a side-by-side preview of old / new site 😎 this is an example of what I call an "AI HUD" -- a "heads-up display" that gives me visibility into what's going on. i can see: - what terminal commands are running - what files are being created on disk - a live running preview of the old / new website (!) this HUD was actually my third attempt at doing this task, and I liked it the best. here's why I landed there: I wanted to move from middleman to astro for my personal site. These are both static site generators that output html, so I suspected an AI agent could do a good job with a port -- since it could compare old / new build output and try to match. Attempt 1: Agent I just had Claude Code do the port. It went pretty well. But it was hard for me to review what had happened! The git diff was dozens of new files and it was hard to tell how they mapped to the old site. I think the LLM also tried directly writing some files rather than doing deterministic copying, which felt dangerous. Attempt 2: Script I realized a better approach was to have Claude Code write a script that would do the port. The script would execute shell commands like "copy all the markdown files from this directory into this other directory in the new site". This worked much better because I could review the *process* rather than the *results*. Still, even with a nicely commented script, it felt like I was doing a lot of work to review. I wondered: how could I make it ridiculously easy to review this? Attempt 3: HUD Enter the command center! This is an entire web app that I had Claude build just for this one task. I can click buttons to run each step of the port, and watch the new website gradually materialize. It has a node server that runs on my local machine that executes commands on my filesystem and coordinates dev servers for the old/new site. And then a web UI that talks to that server to visualize terminal commands, file trees, and live running previews. It's a HUD because I feel like I can ambiently see everything that the port process is doing 🧐 Was this whole thing worth making for this one-time task? I'm not sure, maybe it was overkill 😅 But it did help me feel more confident that I understood what was going on, and total time was still less than it would have taken me to do the whole port by hand! One way to look at this experience is: "AI can put absurd amounts of effort into a pull request description." What do I mean by that? Well, a PR description is a way to communicate a code change to a human. Typically we do that by writing a bit of text, because that's all we have time for. But given more time and effort, we can do better! An interactive command center / HUD can help a person understand what a code change is actually doing. Historically you'd never imagine investing the effort to do that for a single PR. But now it's within the realm of possibility. So, next time you have an agent do something for you, and you wish you understood better what it was doing, consider: instead of just asking for textual descriptions, maybe ask the agent to make you a HUD!

Geoffrey Litt

62,839 Aufrufe • vor 9 Monaten

Keine weiteren Inhalte verfügbar