a16z speedrun is one of the biggest accelerators in the world. We submitted our application a bit back on the recommendation of one of our friends. Our EdubaWare software has been in closed testing for a bit (get ready, next week thousands of you will get access to it and see our changes) We received and Email yesterday saying they want to hear more from us and to send them a video in under three minutes that explains: - Your backgrounds and greatest accomplishments, professional or personal - What makes you exceptional, and why you're the team to win this market - What you're building and what problem it solves - Why this is a big, underserved opportunity - What progress you've made so far  Now fitting all that into one video is challenging, but we did our best. I hope this gives you a view into the other side of my life which is my company Eduba!

Auditing Plumbline, the coding plugin I built, I found that it ships 43 files, about 50,000 tokens, that the AI never loads. ##What it is I built Walkdown, a pre-flight check for AI skills. You give it a GitHub link or a folder, and it reads the instruction layer: the SKILL.md files, agent definitions, hooks, manifests and CLAUDE.md, the English a model loads and runs with your permissions. For each finding it answers two questions, cited to a file and line. What does this text ask the model to do? What does installing it grant? Run it before you install someone else's skill, or before you ship your own. I built it for security. Most of what it taught me was about writing better skills. ## What you can learn from a run The summary page asks four questions: - **Size and shape.** What it costs in tokens every session and at its largest, against the context you expect. If you run local models, add `--context 8k` and watch the percentages move. It also lists the files that ship and never get used. - **Will it work somewhere else?** Hard-coded paths, programs it needs and never declares, and which AI tools besides Claude Code it's ready for. Got a CLAUDE.md and no AGENTS.md? You'll see it. - **Does it ask only for the tools it needs?** Tools granted and never used, loops with no stated limit, hooks that fire on every prompt. - **Anything to be careful about.** Hidden characters the model reads and you can't see. Dead links and files nothing points to. Instructions to go fetch more instructions from somewhere else. Simon Willison's lethal trifecta (can it read your data, take outside input, and send data out?). Leftover TODOs and pinned model IDs land here too. You can get deep into the details of everything it caught. Each finding also says how often that pattern turns up in other repos, so you can tell a common habit from an odd one. `review.html` walks you through each place one card at a time so you can mark it.

I compared @Jake Van Clief ’s animation workflow with a Short I made called “The Box Nobody Wanted to Open.” Jake’s saved system uses Canvas/JavaScript, Python, Playwright and FFmpeg. My Box was built with Python/Pillow, NumPy and FFmpeg. Both use a timed plan and export checks, but I haven’t found evidence that the Box ran through Jake’s engine. The useful discovery: my “preflight” checked a short clip only after the entire video had rendered. Next time I want to render and inspect a short section first, then commit to the full render. Simple change, potentially less wasted time. I’m learning that the point isn’t necessarily to replace one workflow with another. Sometimes it’s to understand what each does and borrow the part that makes the work better. For those making videos with code, what do you check before committing to a full render? @Danney Tee

Over the past few days, I’ve had some fascinating conversations here in this community, including a private chat with @Bryson Bossley . Those conversations got me thinking. And last night, I couldn’t stop thinking. One thought led to another. Ideas connected. Questions became bigger. And somehow, I ended up writing a speech and turning it into a video. So please, WATCH THE VIDEO FIRST. And if you recognize yourself in what I’m saying, if something in that message resonates with you… Come back and read this. ——— What Can We Build Together? 50,000 minds. Thousands of skills. Countless ideas. And yet… most of us are still trying to succeed alone. Think about that for a moment. What if the greatest opportunity in this community isn’t the next AI tool, another course, or another business? What if it’s the person reading this right now? Someone here might have exactly the knowledge you’re missing. Someone else might need exactly what you already know. And the strange thing is… Neither of you even knows the other exists. Isn’t that a tremendous waste of potential? I made this video because I believe we’re overlooking something extraordinary. It’s not another tutorial. It’s not a sales pitch. It’s a question we should all be asking ourselves. What could we achieve if we stopped trying to do everything alone and started combining what we’re already good at? Maybe we don’t need another thousand ideas. Maybe we need to connect the people who can bring them to life. And maybe it starts right here. Not with another complicated system. Not with another promise of success. But with people discovering each other. So let me ask you three simple questions. 1. What are you genuinely good at? 2. What do you need help with? 3. What would you build with others? Don’t overthink your answers. You don’t need to impress anyone. You don’t need to sell yourself. Just be yourself. Because somewhere among the people reading this, someone might have exactly the missing piece you’ve been searching for.

For a month my job was carrying text between chats: copy the plan out of one, paste it into the next, then check whether the second one actually did it. The template I ended up with splits the work this way. - One session at the root plans and dispatches. It's the only one I talk to about priorities. - Each worker (content, design, an assistant) lives in its own folder with its own CLAUDE.md, CONTEXT.md and STATUS.md. - Workers never read each other's files. Every handoff is a brief, written so a cold session can pick it up with no chat history, and every brief lands in one briefs folder. - STATUS.md is the first thing a worker reads at the start of a session and the last thing it writes at the end. That's how I know a dispatch landed without asking. It ships with four hooks: one asks before rm -rf or a force-push, one snapshots high-risk files at the start of the day, one bumps a "Last updated" date when a tracker changes, and one asks before editing a file that's been handed off. It also has a weekly and monthly clean-up routine, so the folders don't silt up. It carries the structure only, none of my brand. You run a builder routine that fills in your voice and visual rules, then dispatch your first brief. The layers follow Jake Van Clief's ICM: what loads every session, what routes, what gets read on demand. Try it yourself: count how many times today you pasted one chat's output into another. Each one is a handoff that could be a file. Question for the comments: what's the handoff you still carry by hand most often? Free, MIT, nothing to sign up for: https://github.com/NFTYoginis/agency-of-one