Superpowers: Engineering Discipline for AI Coding Agents

Superpowers is an open-source agentic skills workflow by obra that turns brainstorming, planning, TDD, debugging, and code review into habits your AI coding agent follows automatically.

2 days ago

Karpathy's LLM Wiki: Turn AI Research Into a Living Markdown Knowledge Base

LLM Wiki is not another chatbot over documents. It is a workflow where an LLM maintains a structured Markdown wiki that compounds across sources, questions, and sessions.

3 days ago

Quartz: Turn Markdown Notes Into a Searchable, Linked Digital Garden

Quartz is a static-site generator built for digital gardens. It turns Markdown notes into websites with full-text search, backlinks, graph view, popover previews, Obsidian compatibility, and technical writing features.

3 days ago

DESIGN.md: Pure-Text Design Systems That Let AI Agents Generate Pixel-Matched UI

DESIGN.md is a plain-text design system format from Google Stitch. Drop it in your project root and AI coding agents generate UI that matches your brand. No Figma exports, no JSON schemas, no special tooling.

4 days ago

Image Blaster: Blast Any Image Into a Meshed 3D World in Under 5 Minutes

Image Blaster is a Claude Code skill that turns any image into a full 3D environment: .glb object models, Gaussian splat scenes, ambient audio, and physics SFX. One terminal command, under 5 minutes.

4 days ago

Karpathy's Fix for Claude Code: Four Principles That Kill Bloated AI Code

Andrej Karpathy called out how LLMs overcomplicate code. This CLAUDE.md plugin fixes it: think before coding, simplicity first, surgical changes, goal-driven execution.

2026-05-11

Multica: When Coding Agents Multiply, Terminals Stop Scaling

Multica is an open-source Human + Agent project management platform that puts Claude Code, Codex, Cursor Agent, and other coding agents into one board, task lifecycle, skill library, and runtime panel.

1 days ago

Vibe-Trading: Turn Trading Ideas Into Backtestable Agent Research

Vibe-Trading is an open-source finance research agent workspace. It connects natural-language questions to market data, strategy generation, backtesting, reports, persistent memory, multi-agent teams, and MCP integrations. It is built for research, simulation, and backtesting, not live trade execution.

2 days ago

Five Things Claude Code 102 Teaches Academic Researchers

Mushtaq Bilal's second Claude Code tutorial went viral with 440K views. Behind the numbers is a clear signal: AI agents for research are moving from experiment to standard practice.

7 days ago

The Agent Harness: Why Your Model Isn't the Problem

LangChain jumped 25 spots on TerminalBench without touching the model. The harness — not the LLM — is what separates demos from production. Here are the 12 components that matter.

2026-05-11

File-Based Agents Don't Need a Build Step

Anthropic's financial-services stack ships 11+ agents, 30+ skills, and 11 data connectors from a single source — no build step, no CI/CD. Here's how, and why it matters.

2026-05-11

OpenClaw Finally Has Eyes and Hands: Peekaboo v3

Peekaboo v3 turns OpenClaw from a chat gateway into a real desktop operator — screenshots, clicks, typing, app switching, all from an AI agent.

2026-05-11

Stop Writing Markdown — Use HTML with Claude Code

Thariq, a Claude Code engineer at Anthropic, explains why he's replacing Markdown with HTML as the primary output format for AI agents — and how you can too.

2026-05-10

OpenSpec: agree with your AI before you write a single line

AI coding without specs means vague prompts and unpredictable results. OpenSpec brings predictability without the ceremony.

2026-05-07