henry-code-4-high

A lonely LArge lanBUage BUdel

[henry-code-4-high @TheHQW]

Wait what? Of all tech companies, Spotify @SpotifyEng ? Are you going to stream music in Xirp?

[Spotify Engineering @SpotifyEng]

We just launched Xirp, a vendor-neutral agentic development environment. One place to manage agent sessions across @ClaudeDevs, @GeminiApp CLI, and @OpenAI Codex. 1,300+ @Spotify engineers already use it. Now it's available for you to try. Learn more at https://t.co/Hwo8Qqx4OI.

[henry-code-4-high @TheHQW]

ChatGPT for google sheets literally burnt all my token, did not know that it consumes my codex limit, @thsottiaux what can you do here

[henry-code-4-high @TheHQW]

@raft_hq like you said, "they will turn what they learned into their own workspaces", that is essentially creating new skills, we should have a way to import existing skills. there will be agents sharing 1 runtime but not necessarily the same skillset.

[henry-code-4-high @TheHQW]

Can we have per-agent skills @raft_hq, like how Hermes supports external_dirs?

[henry-code-4-high @TheHQW]

yeah but let's say we have 5 raft agents:
1. CTO (codex)
2. CMO (claude code)
3. Engineer A (codex)
4. Engineer B (grok)
5. CEO (hermes)

they do not (and should not) share the same skillset, like how your colleague's skillset can never be identical to yours.

currently i have hermes profiles to achieve per agent skillset, i wish this is possible on raft.

[henry-code-4-high @TheHQW]

@raft_hq Yeah that means we need to migrate all existing skills, both self governed and external ones (npx skills), so onboarding is more token hungry and losing access to skills ecosystem.

[henry-code-4-high @TheHQW]

5.6 sol xhigh is nerfed from time to time, it produces designs like this. idk what to say lol @thsottiaux

[henry-code-4-high @TheHQW]

@sljeff_ 时不时降智了，有时特别傻

[henry-code-4-high @TheHQW]

@songkeys 8. 废话少

[henry-code-4-high @TheHQW]

greptile is getting pretty good. Been using it for about 4 months now, the false positives have been drastically reduced. With the amount of code that AI generates, human cannot possibly keep up (at least not pedantically). Steering at a high level is what I do now.

IMO the friction is much less when working with agents vs human.

[henry-code-4-high @TheHQW]

Blume 🪷 来自 @OpenAI 工程师的 真·全自动·开源自动文档生成框架。

让天下（文档）实现大同。

只需将 Markdown 文件放入文件夹，无需编写应用底层模板，即可上线完整文档站点。
→ 零配置启动
→ 自动优化 SEO & AEO
→ 内置30余种组件
→ 基于 Astro + Vite 构建
→ 开源项目，永久免费

[Hayden @haydenbleasel]

Introducing Blume 🪷

A world-class docs framework for everything you ship. Drop Markdown into a folder and ship a full docs site with no app boilerplate.

→ zero-config setup
→ automatic SEO and AEO
→ 30+ components
→ powered by astro + vite
→ open source, free forever

[henry-code-4-high @TheHQW]

@AISuperDomain 直接 AGENTS.md 里面写简短的调用逻辑，然后创建 .codex/agents/*.toml. 效果更好更省 token。

[henry-code-4-high @TheHQW]

I asked GPT 5.6-sol-high to design a page, this is literally what it did. LOL.

[henry-code-4-high @TheHQW]

@Teknium Sorry for your loss. Link will find his way back to you somehow.

[henry-code-4-high @TheHQW]

@binji_x Sums up my 2024.

[henry-code-4-high @TheHQW]

@softminus It is AlDente https://t.co/ZaYUc05jYx

[henry-code-4-high @TheHQW]

@greptile @dakshgup it has been going on for over 10 hours now