Projects
Three running systems and one that gets cut out of plywood. Each began as a question I wanted answered properly rather than argued about, so I built something that answers it.
A daily read on ad tech, with separate editions for AI and for AI agents. Each edition works through the day's trade press, earnings and industry podcasts, then argues about what actually matters rather than listing everything that happened.
Where an edition makes a call about where the market is heading, that call is recorded and graded later on a public scoreboard — including the ones that turn out to be wrong.
Public companies put in front of a council of investor lenses — value, growth, technical, macro, AI disruption and others — each arguing its own case on the same company, then reconciled into a single scorecard with the disagreements left visible.
The AI disruption lens exists because most equity frameworks were built before displacement by AI was a live risk to a business model, and it changes the answer on more companies than you would expect.
A working terminal for prediction markets. Prices, movement, market depth and the case for and against a position, assembled in one place rather than reconstructed across a dozen browser tabs each time.
Watchlists and positions are per-person, so it needs a free account before it becomes useful.
A four-way active studio monitor system — a fleet of them — designed with an AI from acoustic first principles through CNC-ready fabrication files. Every dimension lives in code, and every joint was verified computationally in assembled coordinates before a sheet of Baltic birch was cut.
It is the AI-trust question in miniature, with a falsifiable ending: either the boxes fit or they don't. The blueprints, fabrication package and assembly guide are all published.