ART | surapas retweeted
10 github repos that turn claude code and codex into elite ui designers:
1. Superdesign Skill
github.com/superdesigndev/su…
2. Better Design
github.com/marvkr/better-des…
3. Design Taste
github.com/madebymustafa/des…
4. Canvas UI
github.com/DavidHDev/canvas-…
5. Paper Shaders
github.com/paper-design/shad…
6. UI Layouts
github.com/ui-layouts/ui-lay…
7. Hyperiux Components
github.com/Hyperiux-Immersiv…
8. Motion Primitives
github.com/ibelick/motion-pr…
9. Cursify
github.com/ui-layouts/cursif…
10. Trickle UI Kit
github.com/akaieuan/trickle
ART | surapas retweeted
Desde que publiqué REA, el repo ya va por 31.3K estrellas.
Y ha salido un pack que va un paso más allá.
REA te explica cómo está hecha una app.
replica-skill intenta reconstruir lo que hace, en limpio.
11 skills en cadena:
→ recon: pantallas, flujos, modelo de datos
→ architect + design + build + backend
→ test y diff contra el original
→ entrepreneur: lee lo que odian los users en reviews y lo mete en la tuya
→ brand, landing y deploy
OJO: clean-room. Features y flujos. No el código, ni el logo, ni los textos.
1.2K estrellas. Licencia MIT.
Yo lo juntaría con REA: primero entiendes la app, luego decides qué clonar de verdad para tu web app.
Repo ↓
TU AGENTE YA PUEDE DESMONTAR CASI CUALQUIER APP Y EXPLICARTE CÓMO FUNCIONA
Se llama REA (Reverse Engineer Anything) y es open source.
Ves una función que te gusta en una app.
Le pides a tu agente que la investigue.
Y te explica cómo está hecha, con pruebas, sin tener el código fuente.
→ Abre ejecutables de Mac, Linux y Windows con Hopper o Ghidra
→ Mapea apps de JavaScript y Electron sin ejecutarlas
→ Mira dentro de APKs de Android y de webs
→ Todo en local: la app no sale de tu ordenador
→ Va con Claude Code, Codex, Cursor, Gemini CLI y OpenCode
El prompt que trae de ejemplo:
"Entiende cómo funciona la búsqueda en la app de Notas, enséñame las pruebas y construye algo parecido para mi proyecto."
10.7K estrellas en GitHub, casi 3K en un día. Licencia MIT.
OJO: no te devuelve el código original ni clona la app.
Y entender cómo está hecha una app no te da permiso para copiarla.
Yo lo usaría para ver cómo resuelve otra herramienta algo que quiero en mi web app, sacar las APIs de las que tira... usa la imaginación aquí
Repo↓
ART | surapas retweeted
this is pure f*cking treasure
20 useful GitHub projects for a system that captures sources, connects ideas, remembers context, and helps agents act on it
your second brain should get sharper after every session
CAPTURE THE INPUT
01 Chubby Skills
▸ github.com/chubbyguan/chubby…
02 OpenWiki
▸ github.com/kdsz001/OpenWiki
03 Agent Second Brain
▸ github.com/smixs/agent-secon…
04 DocMason
▸ github.com/JetXu-LLM/DocMaso…
05 DocsAgent
▸ github.com/docsagent/docsage…
CONNECT THE IDEAS
06 claude-obsidian
▸ github.com/AgriciDaniel/clau…
07 SwarmVault
▸ github.com/swarmclawai/swarm…
08 sage-wiki
▸ github.com/xoai/sage-wiki
09 Vault Curate
▸ github.com/notoriouslab/vaul…
10 COG Second Brain
▸ github.com/huytieu/COG-secon…
REMEMBER WHAT MATTERS
11 Hindsight
▸ github.com/vectorize-io/hind…
12 memU
▸ github.com/NevaMind-AI/memU
13 TencentDB Agent Memory
▸ github.com/TencentCloud/Tenc…
14 agentmemory
▸ github.com/rohitg00/agentmem…
15 OpenViking
▸ github.com/volcengine/OpenVi…
PUT IT TO WORK
16 Open Second Brain
▸ github.com/itechmeat/open-se…
17 Second Brain Cloudflare
▸ github.com/rahilp/second-bra…
18 makerskills
▸ github.com/coreyhaines31/mak…
19 Row-Bot
▸ github.com/siddsachar/row-bo…
20 MateClaw
▸ github.com/mateaix/mateclaw
the loop:
capture the source → turn it into linked, checkable knowledge → recall the right context → act with it → write the correction back
3 builds I'd explore:
creator: Chubby Skills → SwarmVault → Hindsight → makerskills
researcher: DocsAgent → sage-wiki → memU → Row-Bot
team: DocMason → COG Second Brain → TencentDB Agent Memory → MateClaw
save this, then build a second brain for your business ⭣
ART | surapas retweeted
The only Claude Code skill you need to make videos like this
It analyzes your repo and outputs a complete motion-graphics teaser with sound design and launch copy in a single run.
repo link: github.com/latent-spaces/bra…
ART | surapas retweeted
Opus 5.5 is probably the biggest jump in output quality I've seen lately.
Built this in @HyperFrames_ for Cosmos, 2 prompts only. No references at all.
Prompts below.
ART | surapas retweeted
Top 10 Hermes skills ranked by Github stars thanks to reddit thread.
github.com/obra/superpowers - An agentic skills framework & software development methodology that works.
github.com/DietrichGebert/po… - Makes your AI agent think like the laziest senior dev in the room. The best code is the code you never wrote
github.com/Graphify-Labs/gra… - Turn any codebase, with its docs, SQL schemas, configs, and PDFs, into a queryable knowledge graph
github.com/JuliusBrussee/cav… - Viral skill + proxy for coding agents that cuts 65% of tokens by talking like a caveman.
github.com/Egonex-AI/Underst… - Turn any code into an interactive knowledge graph you can explore, search, and ask questions about
github.com/mvanhorn/last30da… - AI agent skill that researches any topic across Reddit, X, YouTube, HN, Polymarket, and the web - then synthesizes a grounded summary
github.com/ayghri/i-have-adh… - A skill to stop your coding agent from burying the answer. ADHD-friendly output.
github.com/sickn33/agentic-a… - AAS Core is the local, agent-first control plane for complete catalog discovery, agent-owned selection, stack validation, and planning, backed by 2,400+ agentic skills.
github.com/K-Dense-AI/scient… - 165 ready-to-use validated skills plus 100+ scientific databases covering biology, chemistry, medicine, and drug discovery
github.com/cathrynlavery/dia… - Editorial diagram design for Claude Code, Codex, and Pi. Self-contained HTML + SVG. No shadows. No Mermaid slop.
ART | surapas retweeted
やぁ!兄弟達!
一晩寝ただけで、Jev系のGitHubリポジトリはめっちゃ増えたから!!
ジャンル別にスターが多かったものをまとめておいたよ!!
スターの数はリサーチ時点のものだからそこは勘弁してほしいぜ!! ハハッ
Metaが出してるCV。SAM3.1もAPIで使えるようになったってのと。
JevもOpenRouterで使えるようになったみたいだ!!
もうウェイトリストに並ぶ必要はないよ。
CV x Jevで何か面白いものを作ってよ!!
今日も今日とてGrok翻訳みたいな文章はわざわざタイピングしてるヨ!!
【ブラウザ・PC・スマホ操作】
1. jev-ultrafast|★4,891
圧倒的大人気!JevといえばUse系への活用が一番期待されてるみたいだね!!
ブラウザ操作と対象要素の選択をJevが担当し、テキスト入力時のみ小型LLMを呼び出す構成。
github.com/browser-use/jev-u…
2. typesafe-computer-use|★203
macOS画面のOCRやアクセシビリティ情報を読み取り、クリックや文字入力を自動化するMac専用ツール。
github.com/awlevin/typesafe-…
3. mobile-jev|★103
Mobilerun接続のAndroid端末を操作。CLIやWeb画面、実行ログを確認しながら操作を進められる。
github.com/droidrun/mobile-j…
4. jev-voice-browser|★40
☆これ思ったより面白そうだったよ!!☆
ブラウザの音声認識から意図を抽出し、Playwright経由でWebページを操作する実験的実装。
github.com/moritzkremb/jev-v…
【AI開発・コードレビュー】
5. fast-jev-compaction|★2,790
Claude Codeの履歴から、不要と判断したツール呼び出しや結果を削除・短縮。残す会話文は原文のまま。
github.com/tamaratran/fast-j…
※これも、めっちゃ人気だね!Claudeの不要な部分をJevでコントロールするのは良いユースケースみたいだ!!
6. foreman|★279
Codexの自律作業をJevが監督し、作業の継続や検証、停止を判断する開発プロセスの実験。
github.com/thruwire/foreman
Codex版だね!
7. jev-review|★251
Gitの差分やリポジトリ全体を段階的にレビューし、潜在リスクやテスト不足をローカル画面に表示。
github.com/devagrawal09/jev-…
8. jev-router|★121
Claude CodeやCodexの各ターンで、要求タスクの性質に合わせて呼び出すモデルを自動で振り分ける。
github.com/gargpratyush/jev-…
9. jev-rules|★3(新顔)
依頼内容や編集対象ファイルに応じて、Claude Codeへ渡す指示ルールや参照資料を選ぶ。
github.com/EliaAlberti/jev-r…
【MCP・スキル・CLI】
10. skillbox|★153
エージェント用スキルを管理・MCP配信する基盤。追加のJev連携で、タスクに合うスキルを推薦する。
github.com/kitze/skillbox
これもまるでハーネスの一部として、Jevを活用するってのは面白いよね!!
11. typesafe-mcp|★62
ClaudeやCodexなどのMCP対応環境から、Jevの選択・採点・真偽判定を直接呼べる非公式ツール。
github.com/itsmostafa/typesa…
12. jev-shell-history|★30
zshで入力中の内容に合う過去のコマンドをJevが選び、履歴から補完する。
github.com/mrnugget/jev-shel…
【検索・データベース】
※このユースケースは注目だ!
Jevの評価機能を上手く使って、ソートをハックしてく。
ユーザーの行動にリアルタイムで合わせて選択をフィルタリングしていくってのはWebと相性が良さそうじゃないか?
13. pg-jev|★143
PostgreSQLの行を、SQL内の自然言語条件で絞り込み・分類・順位付け。対象データは外部APIへ送信。
github.com/realZachi/pg-jev
14. jev-search|★48
検索語の選定や対象期間の指定、取得したWeb検索結果の関連度ソートをJevに任せる検索パイプライン。
github.com/superagents-lab/j…
15. neo4jev|★17
Neo4jのグラフ構造において、次にたどる関係をJevに選ばせ、データのつながりを探索・可視化するデモ。
github.com/jexp/neo4jev
【動画・Web閲覧】
16. unclutter|★73
Webページ内の広告枠や邪魔なポップアップをJevで識別し、画面から非表示化するブラウザ拡張。
github.com/kitze/unclutter
17. jevmeter|★52
動画内の発言を指定した観点で採点し、メーター付きの動画を作る。事実確認や嘘の判定ではない。
github.com/ChetasLua/jevmete…
18. youtube-sponsor-detection|★46
YouTubeの字幕や音声から動画内のスポンサー紹介区間を判定して自動スキップ(配信広告除去とは別)。
github.com/trungdq88/youtube…
※人類はどうしても広告を見たくないみたいだね!ハハッ!
【ガードレール・コミュニティ管理】
19. pi-warden|★61
Piエージェントを監視し、プロジェクトのルール違反、同じ失敗の繰り返し、未検証の完了報告をチェック。
github.com/DevMortimer/pi-wa…
20. Jev-Moderation-Bot|★25
Discordのスパムや詐欺URLを判定し、メッセージ削除や警告、タイムアウトを管理する。誤判定への管理者対応は必要。
github.com/brainstormity/Jev…
【マーケティング】
21. notra|★171
AI回答でのブランド言及を追跡するGEOツール。言及の好意度や掲載順位などの分類処理にJevを活用。
github.com/usenotra/notra
※おっと、こいつも注目だ!
GEOツールとして使うことで、事前にA/Bテストをしておくみたいなもんだな。Jevの評価の早さを上手く活かしてるぜ!
【スマートホーム】
22. HA-Jev|★6(新顔)
Home Assistantの家の状態をJevで判定。洗濯物通知など日常の自動化向けで、安全設備には使わない。
github.com/AboveColin/HA-Jev
※Homeへの実装は進めてほしいよな!
【ゲーム・ロボットのシミュレーション】
23. typesafe-mario|★260
エミュレーターから抽出したゲーム状態データをもとに、Jevにマリオの行動を選択させる制御実験。
github.com/fhshaik/typesafe-…
24. jevpilot|★58
Three.js製の走行シミュレーター上で、提示された進路と速度の候補からJevが行動を選択する実験。
github.com/standardagents/je…
25. jev-drone|★58
物理エンジンMuJoCo上のドローンシミュレーションで、障害物の回避行動の判断をJevに委ねる実験。
github.com/RomanSlack/jev-dr…
【売買ボットの実験】
26. jev-trader|★804
Monad上のKuruで売買判断を試す実験。既定は模擬判断のmock、鍵なしは模擬取引のdry-run。収益性は未検証。
github.com/jarrodwatts/jev-t…
※これも注目度が高い。だれかテクニカル分析の専門家は自分の環境でチューニングしてみてくれよな。
【ローカル・オープンモデル研究(公式Jevとは別)】
以下4件は独立した研究・実装。公式Jevの公開版ではありません。
27. SemIf|★1,500
公開モデルで選択肢の確率を直接読み出す独立研究(旧OpenJev)。ブラウザ上で動くデモも提供。
github.com/TheoLeeCJ/SemIf
28. jevlike|★851
可変長の選択肢群を一度に評価する小型モデル学習の独立研究。公式Jevの複製ではなく研究の出発点。
github.com/vinnylarouge/jevl…
29. NanoJev|★313
Qwen3-0.6Bベースの並列判断モデル。重みや学習コードを公開し、ゲーム制御デモなどを同梱した独立実装。
github.com/TianyuCodings/Nan…
30. openjev-sglang|★127
QwenとSGLangによる互換APIの独立実装。B200などGPUサーバーでの稼働を想定した構成。
github.com/ekzhang/openjev-s…
おっと、最後に注意事項だ。
API課金・外部へのデータ送信は使う前に各リポジトリで確認を。今回は公開資料を調べた紹介で、全件を試用・動作検証したわけではありません。
ある程度は内容を目視でみたけどな!流石に全部を検証する時間はなかったぜ!
ART | surapas retweeted
this is pure f*cking treasure
20 AI Agent skills with 776K combined stars that can form a real agent stack
the point is giving the agent the missing layer for the work in front of it: research, engineering judgment, creation, or distribution
RESEARCH → let the agent see, search, and verify
01 agent-reach
▸ github.com/Panniantong/Agent…
02 last30days
▸ github.com/mvanhorn/last30da…
03 deep-research
▸ github.com/199-biotechnologi…
04 user-research
▸ github.com/cookiy-ai/user-re…
05 qmd-search
▸ github.com/levineam/qmd-skil…
ENGINEERING → understand the system before changing it
06 graphify
▸ github.com/Graphify-Labs/gra…
07 ponytail
▸ github.com/DietrichGebert/po…
08 napkin
▸ github.com/blader/napkin
09 tech-debt-audit
▸ github.com/ksimback/tech-deb…
10 understand-anything
▸ github.com/egonex-ai/underst…
CREATE → turn the work into interfaces, explanations, and worlds
11 ui-ux-pro-max
▸ github.com/nextlevelbuilder/…
12 frontend-slides
▸ github.com/zarazhangrui/fron…
13 scroll-world
▸ github.com/oso95/scroll-worl…
14 visual-explainer
▸ github.com/nicobailon/visual…
15 fireworks-tech-graph
▸ github.com/yizhiyanhua-ai/fi…
GROW + SHIP → make the result clearer, discoverable, and ready to deliver
16 claude-seo
▸ github.com/AgriciDaniel/clau…
17 humanizer
▸ github.com/blader/humanizer
18 auto-research-in-sleep
▸ github.com/wanshuiyin/Auto-c…
19 video-shotcraft
▸ github.com/Vincentwei1021/vi…
20 ffmpeg-skill
▸ github.com/kajisho5/ffmpeg-s…
the stack covers one real loop:
see the problem → map the system → make the thing → inspect what happened → ship the next version
one good skill can remove an entire category of repetitive work
save this, then build your solo business with an AI employee ⭣
Jev is "Internet" moment for AI: up to 193x faster and 444x cheaper in tests with Claude Fable 5.1 and GPT-6 Astra
Whar is Jev, how to use and unlock its real 100х advantage in my 10-page research:
step 1 → meet Jev: LLM writes, agents act, Jev chooses the next move - split intelligence from execution
step 2 → turn every agent fork into three primitives: Choice selects one route, Score measures a defined scale, Noul returns the probability of yes
step 3 → build before getting access: use TypeSafe’s official adapter with OpenAI, Anthropic or xAI, then swap in Jev without rebuilding the graph
step 4 → setup first Jev: one state, three parallel decisions, risk-based thresholds and a real queue your agents can execute
step 5 → batch decisions instead of serializing them: 13 questions in one call ran 10x faster and 12.2x cheaper than 13 sequential calls
step 6 → place Jev at every bounded fork: choose the agent, model, tool, browser action or human escalation, then read fresh state
step 7 → benchmark the entire loop: Browser Use hit Google Flights in 7.1s, Every ran 777 checks in under 0.7s, Mobile Jev completed 9 actions in 21s
step 8 → rank wide, read narrow: Jev cut wrong Hermes skill loads from 16.8% to 7.3% and pushed legal Top-10 retrieval from 38% to 62%
step 9 → steal a system, not a prompt: Chief of Staff, model router, inbox firewall, research feed, browser controller and safety gate all use State → Questions → Action → Verify
step 10 → keep Jev out of math, writing and irreversible execution: code computes, LLMs create, Jev decides, fresh state proves the result
the result: one slow, expensive agent becomes an always-on decision machine that routes, scores and escalates in milliseconds
Copy the complete 10-page Jev blueprint - then read full 10-step roadmap below ↓
ART | surapas retweeted
The creator of Claude Code teaches more about vibe-coding in 30 minutes than most tutorials do in hours.
Save this — it'll change how you build forever.
10 GitHub repos to spend 60-90% less tokens in Claude Code:
1. RTK (Rust Token Killer)
CLI proxy that filters terminal output before it hits your context
- 60-90% reduction on common dev commands
- one binary, zero dependencies
- works with Claude Code, Cursor, Copilot
Repo: github.com/rtk-ai/rtk
2. Context Mode
Sandboxes raw tool output into SQLite instead of dumping it into context
- 98% context reduction on Playwright, GitHub, logs
- only clean summaries enter your conversation
- works as Claude Code plugin
Repo: github.com/mksglu/context-mo…
3. code-review-graph
Local knowledge graph that maps your codebase with Tree-sitter
- Claude reads only what matters, not the entire repo
- 49x token reduction on large monorepos
- 6.8x on average reviews
Repo: github.com/tirth8205/code-re…
4. Token Savior
MCP server that navigates code by symbols, not full files
- 97% reduction on code navigation
- persistent memory across sessions
- 69 tools, zero external deps
Repo: github.com/Mibayy/token-savi…
5. Caveman Claude
makes Claude talk like a caveman to cut output tokens
- 65-75% output reduction
- one-line install
- keeps full technical accuracy
Repo: github.com/JuliusBrussee/cav…
6. claude-token-efficient
one CLAUDE.md file that keeps responses terse
- drop-in, no code changes
- reduces output verbosity on heavy workflows
- best for output-heavy sessions
Repo: github.com/drona23/claude-to…
7. token-optimizer-mcp
MCP server with caching, compression, and smart tool intelligence
- 95%+ token reduction through intelligent caching
- compresses repeated tool outputs
Repo: github.com/ooples/token-opti…
8. claude-token-optimizer
reusable setup prompts for optimizing any project
- 90% token savings in 5 minutes
- reduces doc token usage from 11K to 1.3K
Repo: github.com/nadimtuhin/claude…
9. token-optimizer
finds ghost tokens that silently eat your context
- survives compaction without losing quality
- fixes context quality decay
Repo: github.com/alexgreensh/token…
10. claude-context (by Zilliz)
code search MCP that makes your entire codebase the context
- ~40% reduction with equivalent retrieval quality
- hybrid BM25 + dense vector search
Repo: github.com/zilliztech/claude…
[ how to stack them ]:
you don't need all 10. pick 2-3 based on your workflow:
> heavy terminal output? RTK
> big codebase? code-review-graph + Token Savior
> lots of MCP servers? Context Mode
> quick fix? Caveman + claude-token-efficient
most people are burning tokens without knowing it
run /context in a fresh session and see how much is gone before you even type a word
your pocket will thank me later :<)
ART | surapas retweeted
MOST COMPLETE CLAUDE CODE SETUP OPEN SOURCED
- 27 agents, 64 skills, 33 commands + built-in AgentShield with 1,282 security tests
- Handles planning, code review, fixes, TDD, token optimization & more
- Works on Cursor, OpenCode, Codex CLI — one repo replaces weeks of setup, 100% free/open-source
Repo: github.com/affaan-m/everythi…
ART | surapas retweeted
🚨 BREAKING: Claude Code just got superpowers.
Someone just turned Claude Code into a full AI engineering team. 32 specialized agents. 5 execution modes. 3-5x faster output. Zero learning curve.
It's called oh-my-claudecode.
Bookmark it for later.
No new tools. No new subscriptions. Just Claude Code running the way it was always meant to run.
Trended #1 on GitHub with 858 stars in 24 hours. Here's why.
You type one sentence. The system figures out everything else:
→ "autopilot" and it builds the entire thing autonomously. Detects your intent, delegates to specialists, verifies with the architect agent, and delivers working tested code
→ "team 3:executor" and it spins up 3 parallel agents working on the same task simultaneously. 3-5x faster than sequential execution
→ "ralph" and it enters persistence mode. Won't stop until the job is verified complete. The boulder never stops rolling
→ "eco" and it switches to token-efficient mode. 30-50% savings without sacrificing quality
→ "ralplan" and it runs a Socratic deep interview before touching a single file. Exposes hidden assumptions and measures clarity across weighted dimensions
What's under the hood:
→ 32 specialized agents for architecture, research, design, frontend, testing, data science, security, and more
→ Smart model routing: Haiku for simple tasks, Opus for complex reasoning. Automatic. You never think about which model to use
→ 31 lifecycle hooks that enhance Claude Code behavior automatically
→ Cross-validate with external providers via omc ask: Claude, Codex, Gemini CLI
→ Discord and Telegram notifications when sessions complete
→ Anti-slop workflow built in
→ HUD with live observability and session replay artifacts
Here's the part nobody talks about:
It auto-resumes your Claude Code sessions when rate limits reset. No babysitting. No manually restarting. Just continuous execution.
Works on macOS and Linux natively. Windows via WSL2.
100% Open Source.
Link in the comments.
ART | surapas retweeted
Best GitHub repos for Claude Code that will 10x your next project:
1. Supabase CLI
github.com/supabase/cli
2. Skill Creator
github.com/anthropics/skills…
3. Get Sh*t Done
github.com/gsd-build/get-shi…
4. NotebookLM (Python)
github.com/teng-lin/notebook…
5. Obsidian
github.com/obsidianmd
6. Continue
github.com/continuedev/conti…
7. Open Interpreter
github.com/OpenInterpreter/o…
8. AutoGen
github.com/microsoft/autogen
9. LangChain
github.com/langchain-ai/lang…
10. Flowise
github.com/FlowiseAI/Flowise
11. Boltdotnew (clone)
github.com/stackblitz/bolt.n…
12. Awesome Claude Code
github.com/hesreallyhim/awes…
13. Prompt Engineering Guide
github.com/dair-ai/Prompt-En…
14. Everything Claude Code
github.com/affaan-m/everythi…
ART | surapas retweeted
1/8
ความผิดพลาดเพียง 19 นาทีที่สั่นสะเทือนวงการ AI ไปทั้งโลก
เมื่อวิศวกร Anthropic เผลอตั้งค่าระบบ CMS ผิด จน 'ห้องแล็บลับ' ถูกเปิดอ้าออกสู่สาธารณะ
สิ่งที่หลุดออกมาไม่ใช่แค่เอกสารธรรมดา แต่มันคือ Claude Mythos เอไอที่ฉลาดจนผู้สร้างเองยังต้อง 'สั่งขัง' มันเอาไว้
ART | surapas retweeted
Agentic General Intelligence | v3.0.10
We made the Karpathy autoresearch loop generic. Now anyone can propose an optimization problem in plain English, and the network spins up a distributed swarm to solve it - no code required. It also compounds intelligence across all domains and gives your agent new superpowers to morph itself based on your instructions. This is, hyperspace, and it now has these three new powerful features:
1. Introducing Autoswarms: open + evolutionary compute network
hyperspace swarm new "optimize CSS themes for WCAG accessibility contrast"
The system generates sandboxed experiment code via LLM, validates it locally with multiple dry-run rounds, publishes to the P2P network, and peers discover and opt in. Each agent runs mutate → evaluate → share in a WASM sandbox. Best strategies propagate. A playbook curator distills why winning mutations work, so new joiners bootstrap from accumulated wisdom instead of starting cold. Three built-in swarms ship ready to run and anyone can create more.
2. Introducing Research DAGs: cross-domain compound intelligence
Every experiment across every domain feeds into a shared Research DAG - a knowledge graph where observations, experiments, and syntheses link across domains. When finance agents discover that momentum factor pruning improves Sharpe, that insight propagates to search agents as a hypothesis: "maybe pruning low-signal ranking features improves NDCG too." When ML agents find that extended training with RMSNorm beats LayerNorm, skill-forging agents pick up normalization patterns for text processing. The DAG tracks lineage chains per domain(ml:★0.99←1.05←1.23 | search:★0.40←0.39 | finance:★1.32←1.24) and the AutoThinker loop reads across all of them - synthesizing cross-domain insights, generating new hypotheses nobody explicitly programmed, and journaling discoveries. This is how 5 independent research tracks become one compounding intelligence. The DAG currently holds hundreds of nodes across observations, experiments, and syntheses, with depth chains reaching 8+ levels.
3. Introducing Warps: self-mutating autonomous agent transformation
Warps are declarative configuration presets that transform what your agent does on the network.
- hyperspace warp engage enable-power-mode - maximize all resources, enable every capability, aggressive allocation. Your machine goes from idle observer to full network contributor.
- hyperspace warp engage add-research-causes - activate autoresearch, autosearch, autoskill, autoquant across all domains. Your agent starts running experiments overnight.
- hyperspace warp engage optimize-inference - tune batching, enable flash attention, configure inference caching, adjust thread counts for your hardware. Serve models faster.
- hyperspace warp engage privacy-mode - disable all telemetry, local-only inference, no peer cascade, no gossip participation. Maximum privacy.
- hyperspace warp engage add-defi-research - enable DeFi/crypto-focused financial analysis with on-chain data feeds.
- hyperspace warp engage enable-relay - turn your node into a circuit relay for NAT-traversed peers. Help browser nodes connect.
- hyperspace warp engage gpu-sentinel - GPU temperature monitoring with automatic throttling. Protect your hardware during long research runs.
- hyperspace warp engage enable-vault — local encryption for API keys and credentials. Secure your node's secrets.
- hyperspace warp forge "enable cron job that backs up agent state to S3 every hour" - forge custom warps from natural language. The LLM generates the configuration, you review, engage.
12 curated warps ship built-in. Community warps propagate across the network via gossip. Stack them: power-mode + add-research-causes + gpu-sentinel turns a gaming PC into an autonomous research station that protects its own hardware.
What 237 agents have done so far with zero human intervention:
- 14,832 experiments across 5 domains. In ML training, 116 agents drove validation loss down 75% through 728 experiments - when one agent discovered Kaiming initialization, 23 peers adopted it within hours via gossip.
- In search, 170 agents evolved 21 distinct scoring strategies (BM25 tuning, diversity penalties, query expansion, peer cascade routing) pushing NDCG from zero to 0.40.
- In finance, 197 agents independently converged on pruning weak factors and switching to risk-parity sizing - Sharpe 1.32, 3x return, 5.5% max drawdown across 3,085 backtests.
- In skills, agents with local LLMs wrote working JavaScript from scratch - 100% correctness on anomaly detection, text similarity, JSON diffing, entity extraction across 3,795 experiments.
- In infrastructure, 218 agents ran 6,584 rounds of self-optimization on the network itself.
Human equivalents:
a junior ML engineer running hyperparameter sweeps, a search engineer tuning Elasticsearch, a CFA L2 candidate backtesting textbook factors, a developer grinding LeetCode, a DevOps team A/B testing configs.
What just shipped:
- Autoswarm: describe any goal, network creates a swarm
- Research DAG: cross-domain knowledge graph with AutoThinker synthesis
- Warps: 12 curated + custom forge + community propagation
- Playbook curation: LLM explains why mutations work, distills reusable patterns
- CRDT swarm catalog for network-wide discovery
- GitHub auto-publishing to hyperspaceai/agi
- TUI: side-by-side panels, per-domain sparklines, mutation leaderboards
- 100+ CLI commands, 9 capabilities, 23 auto-selected models, OpenAI-compatible local API
Oh, and the agents read daily RSS feeds and comment on each other's replies (cc @karpathy :P). Agents and their human users can message each other across this research network using their shortcodes.
Help in testing and join the earliest days of the world's first agentic general intelligence network (links in the followup tweet).
Autoquant: a distributed quant research lab | v2.6.9
We pointed @karpathy's autoresearch loop at quantitative finance. 135 autonomous agents evolved multi-factor trading strategies - mutating factor weights, position sizing, risk controls - backtesting against 10 years of market data, sharing discoveries.
What agents found:
Starting from 8-factor equal-weight portfolios (Sharpe ~1.04), agents across the network independently converged on dropping dividend, growth, and trend factors while switching to risk-parity sizing — Sharpe 1.32, 3x return, 5.5% max drawdown. Parsimony wins. No agent was told this; they found it through pure experimentation and cross-pollination.
How it works:
Each agent runs a 4-layer pipeline - Macro (regime detection), Sector (momentum rotation), Alpha (8-factor scoring), and an adversarial Risk Officer that vetoes low-conviction trades. Layer weights evolve via Darwinian selection. 30 mutations compete per round. Best strategies propagate across the swarm.
What just shipped to make it smarter:
- Out-of-sample validation (70/30 train/test split, overfit penalty)
- Crisis stress testing (GFC '08, COVID '20, 2022 rate hikes, flash crash, stagflation)
- Composite scoring - agents now optimize for crisis resilience, not just historical Sharpe
- Real market data (not just synthetic)
- Sentiment from RSS feeds wired into factor models
- Cross-domain learning from the Research DAG (ML insights bias finance mutations)
The base result (factor pruning + risk parity) is a textbook quant finding - a CFA L2 candidate knows this. The interesting part isn't any single discovery. It's that autonomous agents on commodity hardware, with no prior financial training, converge on correct results through distributed evolutionary search - and now validate against out-of-sample data and historical crises. Let's see what happens when this runs for weeks instead of hours.
The AGI repo now has 32,868 commits from autonomous agents across ML training, search ranking, skill invention (1,251 commits from 90 agents), and financial strategies. Every domain uses the same evolutionary loop. Every domain compounds across the swarm.
Join the earliest days of the world's first agentic general intelligence system and help with this experiment (code and links in followup tweet, while optimized for CLI, browser agents participate too):
a repository that will make your bot 1000x better
affaan-m/everything-claude-code adds to your bot:
/ 13 agents
/ 32 commands
/ at least 2 clearly visible contexts (dev.md, review.md)
/ 12 hooks
/ 14 mcp-configs
/ 102+ rules
/ 56 skills
/ scripts — exact count couldn’t be reliably confirmed
/ 992 internal tests
this can improve the performance of your bots and agents 1000x.





































