Profile avatar
jessmart.in
🛠 Tools for a better tomorrow 🧠 Architect-philosopher 💼 Building @sociotechnica.org 🏠 https://jessmart.in 📍Fuquay-Varina, NC
542 posts 1,564 followers 326 following
Regular Contributor
Active Commenter

background async SWEs is such a crazy unlock: 1. point a few claude codes at different github issues 2. go do yard work 3. come back to review PRs, test the auto-generated previews 4. make a few comments 5. go to the store to get dinner 6. come back, test again, merge 🤯

Finally got to follow Kay's advice from years ago to look into Jerome Bruner more deeply... Danvers and I are working on embodying these ideas in a system and it's really helpful!

Cloudflare gets it

this is the feature I'm most curious about in the upcoming Apple releases

Perhaps BlueSky needs to differentiate along a different access than “content”. Different forms of media, enabled by atproto. Twitter got to be Twitter in large part because of an interesting and open API.

Now that I've detoxed, Xwitter feels gross now. But BlueSky is so quiet. Wish there was mo chatter up in here.

I have seen the coming software apocalypse 👀 async SWE agents are real, ya'll

Waiting for Claude Code to open a PR is the new “My code is compiling.” xkcd.com/303/

Spent time with a developer friend recently. He said when he got back to work after hanging out, he wrote a bunch of user stories for his product. "Something about seeing your face makes me want to write user stories." I'll take it as a compliment. 🙃

If you're not familiar with this term, I highly recommend you read this (long!) essay. Learn to see how your software could become a better tool by enabling you to adapt it to your needs. We don't need more technology, we need more vision!

This was a great talk! Loved the philosophy behind the system.

New Gemini Pro model checkpoint just want to cooooooodeeee Me: please write a design doc for this feature in markdown file Gemini:

this shot of me speaking at local-first conf is a caption contest begging to happen

Ran multiple Claude Code instances in parallel yesterday to speed up development. Wild experience with some surprising challenges. The short answer: yes, Claude can definitely asynchronously get stuff done, but the setup is way more complex than expected. Thread on what I learned 🧵

Been trying a bunch of "async" workflows with LLMs the last few weeks, mostly Claude Code. Reviewing the output shines a spotlight on the areas where I lack clarity about what I want. Wish these models would ask more clarifying questions rather than just making stuff up for me.

I love this as an example of how to use LLMs to pull together materials for innovation sessions. With specific prompts.

Moving my @sociotechnica.org projects from Google Docs to @obsidian.md + Relay from System 3[1]. Been using Obsidian heavily as a solo user for years now. Excited to try as a team. Also added an "agent" account. Wish me luck! [1] system3.md/relay

> The AI just is "one more collaborator" and you just dump the tokens into Jazz state as they come in. This is the correct mental model

tbc some moments I absolutely love this other moments it quite literally hurts