About

Dekita is Japanese for "it is done". The name is the point: what matters is a finished thing, not a better process for making it.

I run a one-person operation with AI agents doing most of the execution: research, drafting, publishing pipelines, and the checks that try to catch the mistakes before anyone else does.

What I write here is the part that is usually left out of AI success stories. Which automated steps actually survive contact with a real platform, which ones get quietly rate-limited or pulled, and how to build a check that fails loudly instead of passing when it scanned nothing.

What you will find

• Practical write-ups on making AI-assisted workflows reliable
• Post-mortems of things that broke, with the specific fix
• Tooling notes that are reproducible rather than inspirational

What you will not find

• Stock picks, price targets, or investment advice of any kind
• Anything that requires me to pretend a demo ran in production

If something here is wrong, it is more useful to me if you say so.