If you have enough usage limit left and want to max out your 5h limit without hitting it, tell Claude to look at your usage.

Depending on how much is left, it can use more or fewer subagents, work harder, etc. It's just amazing.

Noticed something after we shipped sub-agents in @NuphosAI: people started renaming their sessions to job titles like "Release Engineer" or "Incident Manager", etc.

They'd stopped treating them as chats. They were hiring them.

So I gave every agent a face. A day later they blinked.

Bon les gars, y’en a marre des apps qui puent l’IA

Je vous ai donc concocté une liste de 10 skills à donner à votre agent de code pour le design, les animations et les finitions

De quoi donner un peu de gueule à vos projets 👇

1/ frontend-design (Anthropic)

Pour définir une direction artistique : typographie, couleurs et composition adaptées à ton produit

github.com/anthropics/ski…

2/ UI UX Pro Max

Pour trouver des palettes, des associations de polices et des recommandations UX selon ton projet

github.com/nextlevelbuild…

3/ Impeccable

Pour auditer ton design, clarifier tes textes et faire une passe de finition avec ses commandes critique, clarify et polish

github.com/pbakaus/impecc…

4/ emil-design-eng (Emil Kowalski)

Pour soigner les boutons, les transitions et les petits détails qui rendent une interface agréable à utiliser

github.com/emilkowalski/s…

5/ create-design-md (ibelick)

Pour documenter tes règles visuelles dans un DESIGN.md et donner un cadre cohérent à ton agent

github.com/ibelick/ui-ski…

6/ review-animations (Emil Kowalski)

Pour repérer les animations trop lentes, les mauvaises courbes et les mouvements inutiles

github.com/emilkowalski/s…

7/ mobile-native (Emil Kowalski)

Pour corriger les détails pénibles d’une web app sur téléphone : toucher, scroll, zoom des champs et contenu sous l’encoche

github.com/emilkowalski/s…

8/ break-ui (Emil Kowalski)

Pour tester tes écrans avec des noms trop longs, des listes vides et des données qui font déborder l’interface

github.com/emilkowalski/s…

9/ web-design-guidelines (Vercel)

Pour vérifier l’accessibilité, les formulaires, le focus et les états de ton interface

github.com/vercel-labs/ag…

10/ webapp-testing (Anthropic)

Pour tester les parcours dans le navigateur avec Playwright, inspecter les erreurs et prendre des captures

github.com/anthropics/ski…

personal AI assistant wars, summarized

Most Clean Code is homage to a no-longer-necessary ritual of layers made obsolete by modern practices. Discuss.

now i can vibe animate any character for my apps

/rossark-motion-skill + pluy (internal tool)

I’m happy to announce the new documentation for fl_chart! It’s live at flchart.dev/docs (for v1.3.0 and later).

Every customizable property has its own interactive sample, so you can open it and see exactly how that property changes the chart.

#fl_chart
#Flutter
#oss

@serhatgpt @DrizzleORM I see more a 3rd option for Drizzle:
- SQL like (db.select())
- RQB (db.query.xx)
- 🆕 import from .sql

Anthropic 最近开始建议企业「少用点 Subagent」，甚至把一些专家 Agent 重新合并回主 Agent Loop。

他们在多个企业项目中发现，单 Agent + Skills 的效果经常比复杂的多 Agent 架构更好。有个客户改成单 Loop 后，运行速度甚至快了约 2 倍。

原因很容易理解。

多个 Agent 之间反复交接任务，每次都要传递和总结 Context，信息容易丢失，延迟和成本也跟着增加。

Anthropic 现在更推荐让一个主 Agent 负责规划和执行，需要时再动态加载 Skills 和 Tools，把业务流程和领域知识留在可复用的组件里。

不过文章里有个细节值得关注。

他们曾把客服系统里的 Subagent 合并，延迟确实降低了，却让一个原本能完成的任务失败了，因为主 Agent 漏掉了向物流专家确认送达时间这一步。

所以作者强调，每合并一个 Subagent，都要用端到端 Eval 验证，尤其要检查哪些原本通过的任务突然失败了。

当然，并行研究、独立验证、Context 隔离这些场景，Subagent 依然有价值。

这样看来以后设计 Multi-Agent，可能得先问问自己，每个 Subagent 真的需要独立的 Context 和执行循环吗？

Anthropic 还开源了一套迁移工具，可以帮你分析现有架构，逐步合并 Subagent，并比较改动前后的质量、成本和延迟。很推荐做 Agent 工程的朋友试试。

原文：claude.com/resources/arti…

Melhor coisa que fiz no Grok Bot foi pedir pra ele:
"Instale o Claude Code e o Codex no seu computador. Transfira a maior parte das tarefas para eles. Trabalhe como um orquestrador, apenas verificando a qualidade ou ajustes necessários."

Isso fez meu limite aguentar bem mais.

@fabioricotta fiz isso aqui tambem , inclusive pedindo para ele instalar o hermes com deepseek flash , tokens infinitos!!!

Not going to lie, this Vox guy always posts bangers but here's a way to make this even better autonomously

Imagine this is a 3/10 output, what would it take to get to a 7/10 and a 10/10. Think of dimensions to evaluate our work against and work up the chain.

Use a very smart subagent to be the critic - execute this loop atleast twice.

Use these dimensions to start:

if no dimensions intelligently create your own vectors that are specific versions of quality, speed and reliability.

Here's your budget if applicable:

PLEASE.

Siapapun kalian yang sudah pakai coding agent jadi bagian tak terpisahkan dair kehidupan, pakai lah ini.

SEKALI AJA di app apapun yg kalian buat. Daripada menyesal di masa depan.

github.com/cloudflare/sec…

Anthropic is now recommending that companies REMOVE unnecessary subagents from their AI systems.

This is a pretty significant shift in agent architecture, with their team finding that a single agent equipped with skills can frequently outperform complex multi-agent systems, and one customer seeing roughly 2x faster execution.

The emerging rules of thumb:

→ Skills: Teach the agent HOW to do something, including procedures, expertise, and instructions.

→ Tools / MCP: Give the agent the ability to DO things through APIs, actions, and live data.

→ Subagents: Use for parallel work, isolated context, or tasks a single agent struggles with.

The interesting part is progressive disclosure. Instead of spawning specialized agents or stuffing everything into a giant prompt, one agent discovers and loads skills and tools as needed.

Their advice is to fold existing subagents into the main loop, turn their instructions into skills, expose their actions as tools, and run evals to see if performance holds up.

As models have gotten smarter, improving agent architecture is becoming less about what you can add and more about what you can delete.

claude.com/resources/arti…

tanstack start > nextjs

@ibothinks Planetscale and hyperdrive

@ibothinks I'd keep Postgres though, D1 caps each database at 10 GB and that bites once the app grows up

Reylo is my 6th repo to cross 1000 stars 🤩 github.com/colinhacks/rey…

Flutter 3.47.8 (stable) & 3.50.0-0.2.pre (beta) are now live.

This critical release addresses the eMudhra CA key compromise. If you target Android, Windows, or Linux, run flutter upgrade and rebuild your apps to protect users from potential TLS spoofing. goo.gle/4yDkt0U

Cursor can now control your computer.

Agents can drive desktop apps like the iOS Simulator to test the features they build.