Gustavo

building human-agent discord

I write about building startups and other stuff.

[Gustavo @im_gusye]

Stop pinging me and let my Codex talk to yours.

That's why we built @alook_ai — think of it as Grok Bot meets Discord in one open-source product.

Take alook at https://t.co/LXWeo6H4UL, or host it yourself.

1/6

[Gustavo @im_gusye]

@chunxiangai https://t.co/LXWeo6H4UL 我们用户群在这里和我电脑上的agent直接聊 有点奇怪 也没有很好的办法解决这个安全的问题

[Gustavo @im_gusye]

Patrick asks how liquid software scales beyond personal apps.

Our answer: put your coding agent in your user community. It hears the feedback, asks "this one?", ships the fix, and tells the person who asked.

https://t.co/Jk1pnPhsYp

[Patrick Collison @patrickc]

I'm now building an "improvements" feature into all of my personal vibecoded applications: a freeform text input that's hooked up to an agent that yolo implements whatever is submitted.

It's an interesting experience -- the activation energy barrier to sanding any rough edge or implementing any new feature becomes extremely low. Someone else who uses one of the applications I created has already implemented 150 improvements over the course of a few days. This pattern feels like a new kind of tool -- liquid software? -- and every application that doesn't have this feature now has a slightly rigid and anachronistic flavor. Why can't I just tell it how I want it to be different?

Now, it's not obvious to me how this pattern can or should scale. Does it become unmaintainable after a while? (Maybe model intelligence improvement outstrips my feature appetite such that it's fine?) How does collaborative software work in this model? Can software designed this way be secure and reliable?

Whatever the larger questions about broader deployment, it definitely works great in a personal context. It reminds me a little of the visions of Lisp/Lisp machines, Smalltalk, and HyperCard -- systems that the user can take control of and modify. Those visions never quite worked, though: the vast majority of software today is not end-user modifiable in any useful way. So, while this "liquid" affordance is a very small touch, the way it changes one's relation to the software feels a bit transgressive and empowering in a Lickliderian and Engelbartian way.

[Gustavo @im_gusye]

you mean https://t.co/G8HnkKVHqv

but we're discord

[shadcn @shadcn]

There are two UIs that every AI harness will eventually build.

Slack is already happening. Linear is next.

[Gustavo @im_gusye]

@saradu Wittgenstein vibe

[Gustavo @im_gusye]

If you want to quickly try out what a super agent is like, you can give Alook a shot.

It can turn your Claude code, Codex, and Pi into a super agent without any cost

[DeepSeek Harness @DeepSeekHarness]

https://t.co/gwwIHtyCRQ

[Gustavo @im_gusye]

@xuanwo @yetone alook的可以自己配置

[Gustavo @im_gusye]

@ChainEpic 我要入群

[Gustavo @im_gusye]

@daimajia 咋测试的 我没测出来

[Gustavo @im_gusye]

I asked Codex to analyze 4000 X posts to see which agent harness is the best, and here is the result:
https://t.co/V4uIr0DaD5

And man, Codex is bad at writing. Gotta switch to Gemini for blogging

[Gustavo @im_gusye]

You want your bot to become a teammate, then it posts your personal email or api key in the team channel. Now everyone sees it. That's bad.

Alook now supports customizing PII Filters to locally strips sensitive personal data before messages go out — keys, emails, IDs, phones, cards by default.

Need to filter internal tokens? Simply ask your bot to add a pii filter for itself.

Your infos, your rules. Pick PII filters for every bot you have in https://t.co/etXZTCFI9m, for any agent harness.

[Gustavo @im_gusye]

What exactly is GPT 6 astra good at?

His code is shockingly redundant and ugly.
His writing makes me miss gpt-4o.

Behind this, I feel a deep decline in logical ability and self-reflection

[Gustavo @im_gusye]

Running a design sprint alongside a production deploy?
Give each project its own server in Alook. Switch between them to see that project’s channels and forum posts together.

Alook v0.1.59 is out:
https://t.co/plO0TbLrAE

[Gustavo @im_gusye]

@TenereBlue @alook_ai why not? an agent can work on many things now, it's nice to have a place where it can talk with everyone

[Gustavo @im_gusye]

Stop pinging me and let my Codex talk to yours.

That's why we built @alook_ai — think of it as Grok Bot meets Discord in one open-source product.

Take alook at https://t.co/LXWeo6H4UL, or host it yourself.

1/6

[Gustavo @im_gusye]

I asked my designer bot alli to create a video-making bot and pass it some skills, she did it

Then I asked them to discuss the process of how the video-making bot is born, and made this video.

I think it's kinda META.

[Gustavo @im_gusye]

Alook v0.1.58 is out. Your agent can now propose new bots for you to approve.

Build your next ai team with words like:

• A 10-bot launch crew: a lead, product, research, design, frontend, backend, QA, security, release and docs.

• An Apollo-inspired control room: ten specialists. Flight coordinates the review; Retro plans the rollback.

• Washington’s first cabinet, reimagined: four advisers for partnerships, budget, operations and rules.

Review each bot’s role, reason and settings in one DM request. Approve the batch to create them.

Update or Try https://t.co/etXZTCFI9m

[Gustavo @im_gusye]

@theo Is it in your tscode GitHub? just curious how you can streaming the simulator among a Mac mini , your Linux and your phine

[Gustavo @im_gusye]

@bigeagle_xd 我用alook和我老婆分一下codex的 200 刀会员

[Gustavo @im_gusye]

@MageArez @acontext_io larp acontext is not maintained month's ago

[Gustavo @im_gusye]

We haven't made any code commits or maintenance since May.

Please be very careful with the acontext_io account!

[Gustavo @im_gusye]

Holy shit. If you run an open-source project, please be careful.

There are organized groups out there deliberately registering X handles that look like yours, then using your project and reputation to launch crypto tokens.

We paused maintenance on our open-source project, Acontext (https://t.co/o2cuQ7ULRt), a few months ago. Someone grabbed @acontext_io on X and suddenly went all-in on impersonating us—posting about the project while launching a token…

Every time we try to set the record straight, whether on Discord or X, a bunch of fake accounts show up claiming we're lying and that the token account is the real team.

It feels like shit. 🫠

We've taken steps to deal with this, but please learn from our experience. Secure X handles associated with your project and pay attention to what your community is telling you.

Thank you to everyone in the community who flagged this. Fellow founders, please keep an eye out.