SToneX

Yet another builder.

[SToneX @SToneoneX]

Even though it was just released, I’ve actually been using it (previous internal versions at ByteDance) for three years. I’d honestly say it might be the best tool I’ve ever used for debugging prompts and agent behavior. Definitely give it a try!

[Henry Li @henry19840301]

Today we open source LLM Space because debugging agents in a terminal is misery.

It's the tool the DeerFlow team ships on — every release is built and debugged in it.

Native desktop app. Your threads are just files.

https://t.co/ceBkx1ZwMD

[SToneX @SToneoneX]

After trying a lot of different agent workflows over the past few months, my current takeaway is to start with small, focused loops. They often end up being much more effective than I expect. Stack enough of them together, and they might even outperform one giant "do-everything" agent we've all been trying to build.

[Jason Zhou @jasonzhou1993]

Everyone talks about loops - but no one shows you a real one

@SToneoneX and I spent a month running loops to automate most of @SuperDesignDev - many many of them. Some worked, some didn't

Recorded a 14-min walkthrough on key learnings with the real examples we run 👇

-------

Open-sourced our internal tool to manage loops: https://t.co/zzgasUOip0

+ Skill for setup verifier: https://t.co/jnY4YXWQFz

[Jason Zhou @jasonzhou1993]

https://t.co/6iCX0j0aBP

[SToneX @SToneoneX]

@0xMovez @ShenSeanChen

[SToneX @SToneoneX]

@wenhaoqi @MicrosoftAI @MaximeHeckel 太强了！

[Jason Zhou @jasonzhou1993]

https://t.co/Ra4YZLY7dc

[Jason Zhou @jasonzhou1993]

This 10x our PR review speed & quality 👇

One problem my team had as AI writing most of code:
The architecture in our head VS the architecture AI ships slowly drift apart.

So @SToneoneX shipped archlet:
- It allows you review PR diff in a graph
- Easily review the impact to architecture

Basically with one glance you understand if AI fucked up your architecture

It helped us review PR much faster so want to open-source it: https://t.co/rLbCPWeuqx

Feel free to try out 😎

[SToneX @SToneoneX]

when i vibe-code for a while, the architecture in my head and the actual codebase stop matching.

archlet came out of that. Let your coding agent run it on a repo, and you get a clickable map of the codebase. you can also overlay a PR to see what it actually touches.

https://t.co/5NmDLIIQJy

here it is mapping itself 👇

[SToneX @SToneoneX]

Tools like Claude Code, Codex are great, but still very code-first.

As AI writes more code, reviewing diffs feels like the bottleneck.

Maybe the next IDE should show the system: structure, contracts, data flow, behavior, and impact.

Any similar tools or ideas I should look at?

[SToneX @SToneoneX]

https://t.co/V0Db6SwDaO

[SToneX @SToneoneX]

Tools like Claude Code, Codex are great, but still very code-first.

As AI writes more code, reviewing diffs feels like the bottleneck.

Maybe the next IDE should show the system: structure, contracts, data flow, behavior, and impact.

Any similar tools or ideas I should look at?

[SToneX @SToneoneX]

@27upon2 Mine’s the JC3636W518. But if you’re buying a new one instead of repurposing a screen you already have, I’d probably go with the @M5Stack.

[SToneX @SToneoneX]

Claude Opus 4.7 is kind of wild. I had it port claude-desktop-buddy onto an unused round-screen ESP32 device ridiculously fast. I can approve tasks right from the little screen.

It’s actually pretty useful. 😄

[SToneX @SToneoneX]

@PopVerseYT @jasonzhou1993 I think it’s actually pretty useful. I open-sourced my version here: https://t.co/B8CWuw0fwD . Should be pretty easy to give it a try if you have the same version

[Jason Zhou @jasonzhou1993]

Introducing world's first AI growth team

- Monitor your business 24/7
- Take growth actions autonomously
- Self-improve w/ great memory
- Use any tools you use

Comment ‘@crewlet_ + your website’, I will pump you up in waiting list

[DanielWalnut @taohe669360]

AgentForge taught me how to wire IM into agents.

Now We are bringing it to DeerFlow.
Lark / Telegram / Slack → async research threads → persistent memory → your deer grows with you.

OpenClaw lets you raise your 🦞.
DeerFlow will let you raise your 🦌.
Same idea. Different beast. @henry19840301

[SToneX @SToneoneX]

@PaulSolt @SuperDesignDev with claude code

[SToneX @SToneoneX]

I once spent an entire day just to make this robotic arm move.

@openclaw did it in one minute.

Now it’s controlling a real arm on my desk �� and even picked the music itself: “Golden Snake Dance”
(the dance still needs work ha)

Happy Chinese New Year everyone #OpenClaw

[SToneX @SToneoneX]

@0xTykoo 可以再试试看 https://t.co/JQ6m4RTkyt

[ByteDance Open Source @ByteDanceOSS]

We just open-sourced 🦌 DeerFlow, a cutting-edge Multi-Agent framework built on LangChain. @hwchase17 🌟

Check it out, give a ⭐️, and start building:
👉 https://t.co/7Dt0GgOGfj

See it live at DeerFlow's official site: https://t.co/8gdCRvRPFd

[SToneX @SToneoneX]

@__enix__ https://t.co/lxYOqGiBUu 用的这个