I'm a CS grad student in Tokyo building developer tools and researching how models can reason from very little data. Most of what I make is open source — terminal utilities, AI agent tooling, and experiments that fall out of my research.
No ads, no paywall, nothing gated. Your support covers the unglamorous parts: server time for training runs, the hours spent fixing edge cases nobody else hits, and keeping these tools free for everyone who uses them.
If something I built saved you time or taught you something, that's the connection. Thanks for being here.