build

A clean, native macOS markdown editor that Claude can read and edit while you have the document open. No copy-paste loop, no uploading your notes to someone else’s server — the model works against the file on your Mac, live, through local MCP.

A Mac and iOS AI-connected organizational and context tool. With a simple MCP connection you can connect your AI agent to your phone, easily sharing actions and documents between the two.

Agents discover the APIs they need, then call them directly. Joshua hands over the spec and a pointer to your credential — never the credential itself, and never sits in the request path.

Parsed Analytics started as gathering training data for an AI interview tool and evolved into a podcast intelligence tool and workflow.

Calendly-type booking where no calendar detail ever leaves the Mac. Select time in the week view and send a one-shot scoped link, or let people take a slot from a standing page — the server only learns that a block is busy, never what’s in it.

A client with no AI products and no roadmap. I built the roadmap with their leadership, then launched AI extensions of the offerings their clients already trusted. In year one those were upsold onto accounts representing over a third of total client revenue, with retention up 20%.

A lawyer certifies every filing, so citations cannot be automated away. I built a stoplight process that puts attention on the passages carrying the most risk instead of re-reading everything — a multi-LLM panel that triages cheaply and escalates only where warranted.

A twelve-person in-house B2B marketing team across LinkedIn, email, blog, SEO and sales outreach, rebuilt as a coordinated AI system behind a deterministic harness. Twelve people to 2.5 FTE, at roughly $33 a month in model spend.