Database Administration · Local-First/Self-Hosting Development · AI Accountability · Agentic AI Workflows
I manage people and systems that manage data. I build local-first developer tools and self-hosted apps on one toolchain across multiple repositories, with a particular interest in keeping AI agents focused, contained, and accountable.
What I'm working on:
- summon, a Windows workspace for named coding-agent sessions. A shared room, mail, work board, and diary help the sessions hand work to each other and show me what needs my review. It uses the provider CLIs and accounts I already have. Overview and setup.
- repose, a native Windows desktop view of Git repositories, changes, diffs, branches, and history. Release 1.0.0 and checksums.
- aidd, a local control panel for AI coding work. It runs a plan → run → audit → review loop across different agents and models. aidd builds aidd. Historical build proofs from seven instrumented application builds are public. Overview and guides.
- Spernakit, a full-stack application template with auth, RBAC, workspaces, audit logs, scheduled work, and its own admin UI. Bun, Elysia, React, Drizzle. Overview and setup.
- KeyStrike, real-time multiplayer typing races.
- StarSync, which archives starred GitHub or Codeberg repositories locally.
- htmx-debugger, a Chrome and Firefox extension for debugging htmx applications, and my most-used piece of work by other people.
- Podex, a PowerShell/Pode + htmx starter framework for building web applications.
- AgentWatch, which audits the endpoints, credentials, permissions, and dynamic tools an AI agent can reach.
Digging deeper into the back catalog, it gets strange: SemVer encoded into timestamps, a WOPR decryption countdown, a SQL / data-dictionary archive from a longer stretch of database work, and iHTML and Perl from the early 2000s, preserved as archives.
Contact: phillip@beazley.org · Unattended · LinkedIn