The Latchathon: A Mini Hackathon | Latch AI

Vibe coding changed the game. Tools like Wispr Flow proved that the interface to our machines is collapsing into something faster and more human — you think, you speak, it happens. Prompting stopped feeling like programming and started feeling like telling a very fast colleague what you want. But here's the catch every builder has hit: telling is lossy. Your agent only knows what you remember to say. The richest context — how the problem actually looks, sounds, and behaves — never makes it into the prompt. So we're trying something different. At the Latchathon, you won't just tell your coding agent about your problem. You'll show it. Using Latch, you'll record the real context around a problem — the workflow, the mess, the domain knowledge that lives in your head — and turn it into a rich context layer your agent can actually build on. Then you hack. And when your agent is working from what the problem is instead of a paragraph describing what you hope it is, the difference is not subtle. This is a unique chance to get hands-on with Latch, meet the team that builds it, and walk out with something genuinely impressive — built on context you captured yourself. It's a mini hackathon on purpose. When you show your agent the context instead of typing it all out, you skip the slowest part of vibe coding — so one 60-minute sprint is all you need. Schedule 11:00 — Doors open, lunch served.1:00 — Lightning talks: three 10-minute talks from people working on context capture, agent reliability, and enterprise AI in production. Real war stories, no vendor pitches.1:45 — Live demo: watch a real workflow get captured and turned into structured context an agent can actually run on.2:00–5:00 — Open co-work. Plug in, ship something, talk to the person next to you. New to Latch? Latch is a conversational, context elicitation tool. Via the desktop app, users teach Latch by recording sessions. During the sessions, a series of agents observe and analyze your screen, listen attentively to your explanation, and ask astute clarifying questions to form a deep understanding and agent-ready context layer. It feels like you’re sitting next to a super sharp colleague, walking them through something. It's the "show" in show-don't-tell. No prior experience needed; we'll get you up and running at kickoff. Thought Starters Not sure what problem you want to tackle? Need some help figuring out how to leverage Latch? Our team will be here to offer ideas, support, and guidance throughout the Latchathon. To get you wheels turning, here are a problems we’re brainstormed: A debugger with instincts. Pick a nasty bug in your product and talk through how you'd hunt it down. Turn that context into an agent that takes the first pass at the next one.API docs that don't suck. Walk through an API you know cold. Have Latch turn it into a choose-your-own-adventure implementation guide that plays more like a game than a wall of reference text.A newsletter with taste. Scroll your Twitter feed out loud — what's interesting vs. noise, what you trust vs. what needs a second look. Then build a scheduled task that crawls Twitter, summarizes what you'd actually care about, digs deeper when something smells off, and drops it in your inbox every morning.A rival to the software you hate most. Pick the enterprise tool that ruins your week, show what makes it miserable, and build the version you wish existed.