The Lab
15 demos built15 ideas — each one researched, planned, audited, and prototyped with AI. Click any card to try the demo.
Porchlight
Porchlight has a genuinely warm insight wrapped around a real market, but between background checks, payor risk, and the ghost of Papa's near-collapse, the path from heartwarming prototype to durable business runs straight through operational infrastructure no one has built yet.
AI Grading Tool
The rubric-personalization angle is a genuine wedge in a real market, but between an unproven OCR layer, a brutal institutional sales cycle, and Google entering the room, this lives or dies on whether the founders can get three teachers to pay cash money before their runway runs out.
AI Voice Game
The voice loop is technically within reach for the first time in history and the storytelling appetite is real, but this prototype proves nothing yet and you're building in the shadow of OpenAI's roadmap — move fast or become a feature.
AI cooking assistant
A well-framed solution to genuinely painful problems, built on borrowed AI infrastructure in a market that has historically rewarded acquirers more than founders — proceed fast, stay capital-light, and engineer for a strategic exit, not a platform play.
AI-powered job marketplace
The pain is proven, the timing is decent, and the AI tooling finally makes the vision possible — but right now this is a Figma prototype with JSX, and the graveyard of cofounder-matching startups is a reminder that *connecting* founders is easy and *committing* them is the product.
AI sous chef
A well-framed solution to a genuinely felt problem, built on the right new technical primitives, but competing in a lane that big tech will pave over within 18 months unless it finds a distribution moat fast.
Fishi
Fishi has a genuinely clever conservation wedge that could attract NGO partnerships and government data deals that pure social apps can't access — but it's currently a leaderboard skin on top of an unbuilt AI verification engine that *is* the product.
Fitness Plan Generator
Colin has a genuinely charming wedge into a massive market, but right now it's a form attached to a promise — and in fitness tech, broken promises are the default, not the exception.
PlayDate
PlayDate has a genuinely clever insight — games create real conversation where profiles create performance — but insight without distribution is a fun demo, and two-sided cold start plus a content treadmill means this lives or dies on go-to-market execution, not product elegance.
Story Ideation Tool
A well-conceived creative graph tool with a genuinely smart human-in-the-loop twist, currently a typed wireframe facing a narrow window before Big Tech's ambient AI swallows the use case whole.
Treasure-hunting app
A genuinely underserved niche with a proven engagement model and a real user community, held back from being an easy bet by a powerful free incumbent, an unclear monetization ceiling, and a prototype that has dressed the shop window without yet building the store.
DiveSpot
DiveSpot is AllTrails for scuba diving — a proven model in an underserved niche — but Diveboard already died trying to build exactly this, and the only thing that matters is whether this team has a credible answer to why the outcome will be different.
AI-powered wine storage
The AI label-scan technology is finally good enough to make this work — but "finally good enough" won't matter if Vivino ships the same feature in their next update and the startup has no answer to *why they win* beyond the quality of their prompt engineering.
Life Wrapped
Life Wrapped is a beautifully framed idea chasing a proven emotional insight — the Moves acquisition proves there's strategic value here — but it needs to survive the Google Takeout parser, the retention cliff, and its own privacy liability before it's a business rather than a very good demo.
Subtitles for Humans
A genuinely important accessibility problem with a beautiful UI wrapper around a missing engine, entering a market where Apple and Google just showed up for free — the only path forward is a specific, defensible technical advantage in multi-speaker group settings that the OS giants have demonstrably failed to solve.
Sorted by market timing score · All prototypes built with claude-sonnet-4-6