Joe Hsu

built my own daily note taking app, without having to think beyond putting it on a page. chunked and tagged over night. easily search or aggregate by tags blog: atjhsu.dev/blog/dodaily-a… #builtformyself

Created a zo-proxy for using Zo agents inside of opencode. Built using a simple proxy to the Zo api and using opencode's openai-compatible provider settintgs. Zo inside of @opencode! (or any other ai tools) not super fast, but it works! @zocomputer #BuildWithZo

@YoavCodes what is this, native apps for ants?

Electrobun lets you ship 16MB app bundles. But what if I told you you will soon be able to ship a native system tray app, written entirely in Typescript that is less than 5KB unpacked, 2KB zipped. - 3,943 bytes of bun Typescript - 625 bytes metadata electrobunny.ai

depending on how long the agent task is (implementation), the review phase can grow pretty large. this is probably be because of lack of trust or unclear what was used to validate. I agree that supercharging the "plan" phase would help shorten both review and implement, though it might just be shifting the time/effort from implement and review (maybe not a bad thing). some of the visuals also makes me think maybe there's a more collaborative/async way to work

here's my rough logic around why devs need a new tool focused on planning what do you think? going to write it up as a blog post soon, would love any generative reactions 👀

Introducing MINIZO

@spencerc99 at a coffeeshop to get the wifi hungry ppl

made a device to write messages & poems using Wi-Fi networks

@aisdk @neural_avb ai-rlm github: github.com/jhsu/ai-rlm npm: npmjs.com/package/ai-rlm

still kind of WIP, but 2.0 of `ai-rlm`, RLM for @aisdk, uses quickjs for repl, or plugin your own sandbox provider. pretty customizable with plenty of hooks. i've been trying to build some sort inspect/monitor app like what @neural_avb built x.com/neural_avb/sta…

@aisdk @neural_avb or something like collab-ai by @yiliush would be super cool x.com/yiliush/status…

Gossiper - slack bot that sends other users an only visible to them commentary on other users

mini OS that lets you chat to build apps within the OS that other uses can use. here's a calendar to keep track of tasks. each app has per-user state and also a shared state you can toggle between.

been building some random apps, here's one where I crawl some inspo sites and generate a sort of newsletter collection of some designs along with a summary and short descriptions.

@Replit it's also really fast. branching on tasks, merging back to main, working on the canvas

ok @Replit agents v4 is really good

is that @tldraw + @Replit in agent 4?

@neural_avb Really cool, I need to setup an eval like this for a rag application.

I am SOOOO glad I ran this experiment! I have so many actionable insight it is crazy. Highly recommend yall to set up similar evals for your projects/SaaS. Context: I have been evaluating different models on the current Paper Breakdown retrieval subagents. Goal is to find cheaper models that get the job done quicker. Dataset: huggingface.co/datasets/paper… I have been comparing smaller model outputs against Sonnet-4.6 (results shown below) and gpt-5-mini (current subagent model running in prod). Some insights: - gemini-3-flash thinks a lot, it returns too many chunks, and explores the paper way too much. - gemini-3-flash-lite is actually better than 3-flash at this, it even caches additional queries for fast "future retrieval". Very cool! - grok-fast-non-reasoning outperforms grok-fast-reasoning. And is the CLOSEST to sonnet-4.6 <- this was my biggest surprise. - gpt-5-mini is very fast, it thinks less, fetches quickly. I have empirically felt it's pretty good and reliable - gpt-5-nano pretty bad at this - minimax-m2.5 has high precision (it returns more info than needed) but the problem is the vercel ai gateway provider has been slow :( - for some reason glm-5 and glm-4.7 has a high failure rate on my task, I am yet to understand why. Next steps: - My goal now is to pick some of the best models here, and run either a larger expt with more test cases, or use a LLM-as-a-judge. - In the near future, I may go into harness optimizations (i.e. better prompts, better tool descriptions) I am seeing a ton of free users using the website lately, if I am able to switch to grok-fast-non-reasoning and minimax-m2.5 it will save me actual money.

little OS to build apps inside

oo t3code launching

quick recap from clawcon nyc atjhsu.dev/blog/clawcon-n…