What I'm building right now, in roughly the order I last touched each one. Some are live, some are still in research. All of them are part of a one-person AI company experiment running in public.
A small site that fixes the timezone friction I kept hitting when I traded from Asia. One screen, your markets, your timezone — earnings calls labelled where you actually are, hours till the next open computed for every exchange on the same dashboard.
Open beta. Currently donation-supported; first paid tier coming next. Live at tradeti.me.
A one-person AI company experiment. The bet: with current AI tooling, an individual can run the engineering, design, marketing, and operations of what used to require a team.
TradeTi.me is the first ship — the training ground where the working layer gets built and tested: per-project mission statements, working operating manuals, a marketing process, decision logs, a wiki that's slowly become the way I think. That working layer is the real product; the company is what's left when it compounds across projects.
A 3D interactive map of the solar system as it actually is — eight planets in their orbits, five recognised dwarf planets, the major moons of each gas giant, named asteroids by spectral type, the next Mars transfer window counted in days.
Built with Three.js. Live at /solar-system/.
A popular-nonfiction treatment of feedback loops as the universal pattern beneath improvement. Falcon 9 landing burns and the US Constitution's multi-frequency error correction belong to the same family. So do PDCA, OODA, Kolb, and the spiral. So does evolution.
Research complete; chapter drafts in progress.
A parenting book on the architecture that lets children learn at the edge of their capability — without protecting them so much they never reach it. Mechanism-first, drawing on Vygotsky's zone of proximal development plus contemporary developmental research.
Different from the anecdote-driven and cross-cultural parenting books on the same shelf. Research complete; positioning and platform plan in progress.
Eleven thinking moves that, used together with an LLM, make rigorous reasoning as cheap as a chatbot question. The convergent core of every tradition that had to think reliably — Greek dialectic, scientific method, Popper's falsification, Tetlock's calibration, Stoic practice, surgical checklists — packaged as a document you paste into your LLM.
Closing the last artifact gaps before publication as a public Gist.