Your sponsorship pays for the hours I sink into open-source AI tooling. Here's what I work on and where the money goes.

• Caveman hit the Hacker News front page in April 2026 and crossed 6,600 GitHub stars in a few days, sitting now at top 400 repositories of all time with 57000+ stars. It's a Claude Code and Codex skill that strips filler from AI agent output, cutting ~65% of tokens without touching technical accuracy. One-line install, works with 40+ agents. I maintain it, respond to community issues, run evals, and ship features like intensity levels, a compression tool for memory files, and 文言文 (Classical Chinese) mode.
• Cavekit (formerly Blueprint) turns natural language specifications into parallel build plans and working software. You write what you want. It builds, validates, and iterates across models.
• Revu is a native macOS study app with FSRS spaced repetition, decks, exams, and study guides. The open-source SwiftUI codebase lives at JuliusBrussee/revu-swift. The commercial product runs at revu.cards.

I'm 19, studying Data Science & AI at Leiden University, and sit on the Education Committee shaping curriculum for the program. I write about AI infrastructure and indie development at Neurabridge on Substack.
Sponsorship keeps these projects MIT-licensed and free. It covers API costs for benchmarks and evals, pays for infrastructure I can't get on student plans, and buys me time I'd otherwise spend freelancing. If Caveman saved you tokens, or Revu helped you study, or any of this work made your day shorter, consider sponsoring.