30
Projects worldwide
Spring '26
Cohort
AI SDK
Adopted in Caracal
The unglamorous problem
Docs, dashboards and demo environments cost real money, and they are the first thing to degrade when a project runs on a personal budget.
For infrastructure asking people to trust it, that reliability is part of the argument.
Engineering
What we adopted
- Problem
- The operator features needed a model layer we would not have to rewrite, and the public surface of the project had to be fast everywhere.
- Approach
- Integrated the Vercel AI SDK while building Caracal's AI-powered operator capabilities, and moved the documentation and static web experiences onto Vercel.
- Result
- Fast global delivery for the docs, and a deployment workflow short enough that shipping them stopped being a decision.
Vercel AI SDKVercelTypeScriptStatic hosting
The part I didn't expect
The cohort itself — maintainers building modern AI and developer infrastructure, willing to say plainly what had not worked for them.
Most of the useful conversations were about open source sustainability, developer experience and what production-ready AI tooling actually requires, rather than about any one framework.