63 field notes about workflow. Quick, practical, no fluff.
OpenMontage is an open-source system that takes a plain-language prompt and handles the research, script, footage, editing, and music — delivering a finished video.
Emdash lets a single developer run dozens of AI coding agents at the same time — each on a separate task, never stepping on each other's work.
Ponytail gives AI coding assistants a simple instinct: before writing anything, check if something already does the job. Fewer lines, faster results.
Alibaba ran this AI code review tool internally for two years across tens of thousands of developers — and just made it free for everyone.
Headroom quietly compresses what your AI reads before it reads it — slashing API costs by up to 95% without changing how anything works.
Odysseus is a free, open-source workspace that runs AI assistants, email triage, and research agents entirely on your own computer — no subscriptions, no data sharing.
A widely-shared principles guide reveals that the most reliable AI products are mostly structured code, with AI sprinkled in at precisely the right moments.
Microsoft Research built a browser agent that writes reusable scripts instead of clicking blindly — so every task you solve once, stays solved forever.
A free tool that pre-reads your entire codebase so AI coding agents stop wasting time searching — and cost you far less per session.
Google's new WebMCP standard lets AI agents interact with websites cleanly — no guessing, no fumbling — and Booking.com, Shopify, and Expedia are already on board.
A free tool that reads your entire project and draws a living map of how everything connects — built for teams who inherited someone else's mess.
A free tool that turns Claude into a full research team — and won't let it invent fake sources, which is the part that actually matters.
OpenHuman quietly reads all your tools every 20 minutes and builds a private memory of your work — so it never starts from scratch.
A new open-source tool flips the AI coding workflow: instead of prompting and hoping, you write a structured brief first — and the AI executes against it.
A small Mac app sits quietly in your menu bar, watches you repeat the same task three times, and then writes a detailed playbook your AI agent can follow — in your voice, with your logic baked in.
Sandcastle lets you spin up multiple AI coding agents at once — each working in its own isolated space — and then quietly merges everything back together.
OpenAI open-sourced a tool that turns a project management board into a team of coding agents — each ticket gets its own worker that runs until the job is done.
Warp just open-sourced a developer tool where AI agents do the actual coding — and humans just check the work.
A developer published his personal AI instruction set as a free, installable collection — and 31,000 people starred it in days.
PentAGI is an open-source AI agent that runs full security tests on your systems autonomously — no specialist required to set it in motion.
Wanman runs a coordinated crew of AI agents on your own machine — each with a role, each talking to the others — while you sit back and watch it happen.
An open-source AI agent that reads academic papers, finds datasets, and trains machine learning models — without anyone holding its hand.
A new open-source tool reads PDFs the way humans do — charts, tables, diagrams and all — instead of pretending they're just paragraphs.
A Google engineer published 20 ready-made instruction sets that stop AI coding tools from cutting corners — and they're free to use today.
A free, open library that teaches AI assistants to think like a cybersecurity expert — covering 26 areas of digital security across the frameworks big companies actually use.
A tiny file you drop into any AI-built project that stops the agent from producing the same generic, purple-gradient website it made for everyone else.
A new open-source tool turns plain web code into rendered video — which means AI agents can finally produce video content without touching a timeline.
Archon gives AI coding tools a fixed process to follow — so every task runs the same way, every time, no matter the model's mood.
A tiny AI agent that gets smarter and cheaper every time it completes a task — and reportedly built its own codebase without a human ever touching the keyboard.
A tiny config file inspired by one of AI's sharpest minds turns Claude Code from a confident intern into something closer to a careful engineer.
An open-source project that works like a tiny film studio on your computer — you describe a video, it writes the script, plans the shots, and produces the whole thing while you sleep.
A small YC-backed agency figured out how to fix broken developer docs in weeks, not months — by letting AI do the tedious parts and humans do the thinking.
A tiny add-on makes Claude respond like a caveman — and somehow that cuts costs by up to 87% while keeping every useful answer intact.
Free, open-source Claude Code skill that turns any codebase into a knowledge graph. Ask questions in plain English, runs 71x cheaper, works with a local mode.
Someone built an AI-powered job search system for themselves, used it to evaluate 740 offers and write 354 tailored CVs — then got hired as Head of Applied AI.
A plain text file called DESIGN.md can now tell an AI agent exactly how a website should look — colors, fonts, spacing — so it builds interfaces that actually match your style.
A small open-source tool called OMX quietly turned a single AI coding assistant into a coordinated team of agents — and someone used it to do a week's worth of work in one night.
A tiny open-source tool just went viral for solving something deceptively boring: measuring how tall a block of text is, without making the browser stutter.
TREK is a free, self-hosted trip planner your whole team can use in real time — maps, budgets, weather, packing lists — and it lives on your own server.
OpenWork turns solo AI coding sessions into something a whole team can audit, repeat, and trust, with logs and shared workflows.
A new plugin for Claude Code gives you an instant team of 32 specialized AI agents — architect, designer, librarian, and more — without writing a single line of setup.
A single command sends an AI agent across Reddit, YouTube, X, and five other sources — and brings back a grounded, cited summary of what people are actually saying in the last 30 days.
A small Mac app quietly bridges the gap between Claude and your actual life — calendar, messages, contacts, reminders, and location included.
DeerFlow is an open-source AI agent that does not just answer questions, it plans the work and finishes whole projects on its own.
Researchers built an open-source system where a whole cast of AI specialists — analysts, a trader, a risk manager — debate each other and arrive at a single investment decision.
OpenRAG lets you build a private search brain for all your business documents — PDFs, audio recordings, spreadsheets — without sending a single file to anyone else's cloud.
A tiny open-source tool that shows you exactly how much of Claude's attention budget you've used — before you run out mid-task.
Researchers at one of China's top universities just released a tool that turns any document or topic into a live, interactive AI classroom — complete with teachers, curious classmates, and quizzes.
ByteDance just open-sourced a smarter way for AI agents to store and retrieve what they know — and it could quietly cut costs while making agents a lot more reliable.
A new tool lets AI agents operate desktop software like GIMP or Blender the same way a human would — but faster, quietly, and without anyone touching a mouse.
Hermes Agent remembers past conversations, learns from mistakes, and connects to Slack, Discord and Telegram. Free, self-hosted, built by Nous Research.
100+ ready-made AI agent personas in one repo - frontend devs, brand managers, growth leads. Copy-paste into Claude, ChatGPT or any AI tool.
Andrej Karpathy just released a tiny tool that lets an AI agent run experiments on its own, all night, and hand you the results by morning.
IronClaw is an open-source runtime that lets AI agents do real work without ever seeing — or accidentally exposing — your API keys and secrets.
Paperclip lets you build a fully staffed AI organisation — with a real org chart, budgets, and accountability — that actually runs itself.
A free, set-it-and-forget-it system that watches Reddit for your brand name and pings you when someone mentions it, so you don't have to.
WorldMonitor pulls together global news, markets, and geopolitical signals into one screen — with AI analysis that runs entirely on your own machine.
GitNexus turns any code repository into a visual map your AI assistant can actually read — privately, for free, in your browser.
Google quietly embedded a powerful AI forecasting model inside Google Sheets — and you don't need to know anything technical to use it.
A free tool that gives your AI coding assistant a real methodology — so it stops winging it and starts working like a disciplined engineer.
A new open-source tool lets AI navigate long documents the way an expert would — by understanding structure, not just scanning for similar words.
An open-source platform that takes your team's little scripts and automations and gives them a proper home — with scheduling, logs, and a real interface.
n8n lets you automate anything between your apps — Slack, email, CRM, spreadsheets. It's free, it's visual, and you own the whole thing.
The AI tools and moves I actually use to win more business. A couple a week, nothing I haven't run myself.