There's a weird relationship between reliability, risk-taking, and product innovation.
To ensure the most reliable user experience, you have to take enormous institutional risks to build rainy-day things you are not sure you'll need. But once you have those things, you have the
Co-founder/CEO of @convex. Former Sr. Director/Principal Engineer of Storage & Databases at Dropbox. Shameless startup junkie.
- The @convex LLM leaderboard started as our own internal evals. We were measuring how well models could write Convex backends. Now, with Fable, Opus, and Sol, we often find ourselves deleting a lot of our Agents[dot]md context. The models need a lot less help.
- There are a lot of low value interactions you can have with VCs, but the good ones make all the difference. Michael and I spent many hours riffing on what we were doing and where it could go long before we were actually raising money. So once we were, most natural thing in theEvery shift in how software is built creates opportunity to reimagine infrastructure. AI is no different. @convex is reimagining the backend for an AI-first world — making it dramatically simpler and more reliable for both human developers and AI agents to build and scale
- I use AI a lot, but it doesn't particularly inspire me. What I love about tools is precision. Something that helps a beautiful, messy human do their work more precisely. AI is a faster way to do what humans already do (which is still valuable) but it's not the thing that makes
- We got lucky with AI three times at @convex 1/ LLM-assisted development made people reevaluate their backend tools. We'd been waiting for something like that but didn't know that would be it. 2/ the batch of founders building on us when we launched were all building AI



