Who builds it
ProductBrain is built by Anthony Lee. He started as a self-taught hacker, shipping end-to-end bespoke systems on contract with no formal training — and spent the years since on work where the stakes were real:- Leica Geosystems & Caterpillar — real-time operational dashboards for fleets of expensive equipment in motion, where a stopped machine costs real money by the minute.
- Wotif (acquired by Expedia) — funnelling sprawling systems into something simple enough to be a consumer convenience.
- Tatts / Tabcorp — innovation on legacy, high-volume, low-latency financial systems, where real money moves in real time.
- ABC — high-volume media, brand, and SEO; running and measuring hypotheses to win. He led the team through the period ABC News became Australia’s #1 news site (Nielsen, January 2020), overtaking News Corp’s news.com.au.
- Best Practice — data integrity, security, and trust, in Australia’s largest GP practice-management system.
Why it exists
He built ProductBrain part-time, to solve his own problem: holding a vision coherent all the way from idea to shipped product — the thing every side project loses. He looked for a tool first. Nothing affordable existed — the one real candidate was aimed at big business, with opaque pricing you had to request a quote for. And he needed something that would keep up with him, not slow him down. So he built it, used it on itself, and for the first time the thread held. It stayed coherent enough to reach market without drifting. That was the signal. On 25 April 2026, he left his job to build it full-time.What problems it solves
- You lose the thread. The idea that was sharp in your head dissolves as you build — six months in, you can’t remember why you started. Every piece of work traces to its why, so the vision holds from idea to ship.
- You can’t tell what’s actually worth your time. Everything feels urgent; you can’t weigh one bet against another. The tree lays your bets side by side, so you can see what matters before you spend the week.
- Your AI agent has no memory of why. It ships features but doesn’t know which need they serve, or whether they’re even the priority. The agent reads and writes the plan, so it works from the why, not just the ticket.
- “Done” is a claim, not proof. A closed ticket only means someone thinks it’s finished. Jobs are verifiable acceptance criteria — done means verified in the product.
- Planning tools are toys or bloat. Markdown loses structure the moment it matters; the heavyweight suites are ceremony for any team that moves fast — whether you’re solo or a high-performing, tightly knit team. Opinionated structure, zero ceremony, fast enough to keep up.
- Strategy and execution drift apart. Two tools, and the roadmap rots away from the work. One canvas — strategy tree ↔ execution matrix, the same data.
- Lock-in. Your whole plan exports as JSON in one command — the API your agents read it with is the one you leave with.