Application security and AI, in production. Not a demo, not a notebook.
The platform I work on scans hundreds of thousands of repositories and absorbs tens of thousands of events a day. Agents that reason about vulnerabilities instead of pattern-matching them, running against real codebases, with a pager attached. The model was never the hard part. Git internals were, and the 40 to 60x we clawed out of clone times to make any of it viable at that volume.
Before this: security research and reverse engineering, then a long stretch in automation.
Almost all of that lives in private repos, so this profile is the other half of the desk. Mostly Claude Code skills, each written because something annoyed me exactly one time too many.
- human-comments teaches agents to stop writing essays in your source
- interactive-spec turns a spec into a clickable mockup that sends your comments back
- marp-deck-architect writes the deck, not just the theme
- interviewer-kit turns a job description into an interview you can run twice
- morning-briefing reads the inbox so I don't have to
Away from all that, check-please opens your chess.com games in Lichess's engine. No signup, no paywall, no server, because there is no backend at all. Try it
Tip
Install steps live in each repo, and they differ more than you would expect. Start there rather than guessing the path.