Benoît Devilliers

@elvissun Glad Codex was already setup as a backup :)

so what are you switching to?

@elvissun really cool, the subagent you trigger here are not CLI agents right? You only use CLI for coding

ok we just graduated from "too many users I can't manually check everything" to the "zoe also can't check everything because it no longer fits in her context window" territory exciting but also scary... the new workflow is zoe spawns subagent swarm to review result for every single user. then zoe reviews output and compiles a report. then I review that report with zoe. then she sends our notes to a codex agent to implement. wtf.

@thekitze Was on PC switched to PS5 to have clear separation between gaming and working

one of the best decisions i made like 5 yrs ago is making a gaming pc youtube.com/watch?v=XA49aD… playstation can fuck off with their hardware + game prices i was a ps fanboy just because of a few exclusives that are now on pc now with moonlight+sunshine i can stream my games to any device in my home (tvs, phones, tablets, etc) fu sony <3

@euboid @Railway I love Railway too, using it daily, are you hosting your DB/bucket there too? Having this overview of the project is really nice and makes me want to see everything about my project here. Maybe they could add a way to show external DB too

.@Railway feels a lot like Arc (browser) Arc permanently changed how I browse the web. Since they put it into maintenance a year ago, I've tried a dozen other browsers. Nothing has scratched that itch. For better or worse, Railway has also per changed the way I deploy and scale. I’ve tried a dozen other cloud providers, and nothing feels as well designed or thoughtfully built as Railway. It's not even close.

@robinebers I have been working on something like this for myself that I am improving day by day. It's using @openclaw and follows @elvissun idea of swarm agents. Where I am not aligned with you: I don't believe you need file editor, you can always open a PR in cursor or conductor to go into details. I think this as something that stands above your project as an orchestration layer. This is where you refine your tickets so that agents have better chances to one shot implementation. Ideally visibility should be built in so that you can view what agents are doing and unlock them if they raise their hand but they should run mainly in autonomy. I think each project needs its own context: Linear/gitghub + any other docs that can help. For now I have github and linear integrated on my projects. Behind that my openclaw agent has its own github account and is creating PRs.

@T_Zahil Yeah won't make a difference for me, I've banned social media apps from my phone

@benoit_dev_ Yeah that's the main issue 😬 I'm working on a mobile app, it should be available next week!

❓If you're not on Uneed Community yet, why? We're building something unique here 💸 Free and open to all 🤖 No AI, no bots, no spam 📣 Old "Build in Public" spirit 📈 More active users every day 🚫 No algo, payouts or anything like it 🔥 New features dropping almost every day 🧑🏻‍🤝🏻‍🧑🏻 Makers helping each other, asking for advice, and giving feedback Join on uneed[.]best/community

@vanschneider Some stairs are the worst, you can't get a good walking rythm. (elevators stairs for example)

@MistralAI Wow that's crazy!

🔊Introducing Voxtral TTS: our new frontier open-weight model for natural, expressive, and ultra-fast text-to-speech 🎭Realistic, emotionally expressive speech. 🌍Supports 9 languages and accurately captures diverse dialects. ⚡Very low latency for time-to-first-audio. 🔄Easily adaptable to new voices

@robinebers I think there are differents layers to ADE, in conductor it's always focused on the current session but to me there should be a layer above that gives you an overview. I'm picturing it as a kanban view from which you can check on what agents are doing/trigger agents

hear me out Conductor, but with actually good Codex support it's been clear since day one that the team is ahead in terms of thinking about ADEs, but they fucking hate Codex and clearly never use it someone should truly build the alternative, where every integration is treated the same (Codex, Claude, Cursor, OpenCode, etc) then kill half the features they added recently (especially "gstack" 😂) and you're sitting on a winner

@elvissun Let's go! I'd be down to watch :)

mac studio for zoe just arrived m4 max, 128gb ram - enough for a fully parallelized swarm of 10 agents running locally thinking about live streaming the setup. would anyone watch? maybe we can also figure out how to get zoe in the chat

@jamesdevonport IMO there is a chance that the next stage is to write tickets for agents to pick up and come back to review

I was a product manager for years before becoming a maker. So fun to be able to build my own stuff now, could never go back to writing Jira tickets all day!

@ToolsOttho J'ai aussi des troubles de l'attention et ça a clairement trigger un hyperfocus pour moi, impossible de faire marche arrière par contre après lol c'est ma nouvelle drogue haha

Claude Code c'est clairement un outil pour TDAH

@simonbs Love the idea! this is what it triggers for me open.spotify.com/intl-fr/track/…

I’ve been working on something… and it’s now on the App Store! 👀 Introducing brrr 🚀 Push notifications to your own devices. Instantly. No signup. No dashboard. No nonsense. Link in 🧵

@vanschneider One thing I like to do is to create a styleguide page where I can preview all of my components. It's often easier to get the component right from there than to tweak in while in context

@DrewAustin @0xfoobar @openclaw @claudeai @OpenAI I agree with you but we could make the same argument for Cursor yet people seem to go with Claude Code or Codex. Token price is a huge incentive right now. Local models getting better could flip that

@0xfoobar @openclaw @claudeai @OpenAI Yes, but openclaw allows for flexibility of models and tools and not lock in

It’s fascinating to me that people don’t really see the point of @openclaw yet as opposed to @claudeai or @OpenAI. But they will when the company paying for their AI asks them to leave their AI at the door when they leave. Thus making them basically unhireable. Owning your AI agent will be the only way forward.

I'm live on youtube building my Openclaw agent orchestration platform, come say hi! gaming.youtube.com/watch?v=sN-vpW…

@bestbubbledev Yeah I'm not against buying machine to host more of them later, currently working on having my own UI to manage my projects using Openclaw as the brain

@benoit_dev_ If we are betting on the future we are betting on, no one will have only 1 OpenClaw. So I’m assuming some will run online some macmini. However we like. PS if there’s any limitations of running Any Claws 🦞 online I’m sure some hosting startup will solve in next YC batch. Lol

Where do you run your OpenClaw and why? Mac Mini vs Cloud

When AI makes everything faster, friction becomes a feature that helps us pause and think

@petergyang Added this to my openclaw and memory is much better github.com/martian-engine…

My personal experience on the drawbacks of using Claude Code vs. OpenClaw as a personal assistant from my phone: Claude Code - Doesn't have dangerously skip permissions via remote control. Also remote control doesn't feel super reliable - Doesn't have voice replies so doesn't feel as personal OpenClaw - Tends to forget things randomly still