Long-form notes on the things I’ve been wrestling with: running engineering organisations, productising data work, and the awkward gap between research-grade systems and commercial product. Collected in chronological order.
A code review on a small UX change collided with a notebook full of 45,451 free-text plant-science questions. The data made a stronger argument than any hunch could: don’t hardcode what your users call things.
First entry in a recurring short-form series. My second week on the job at JII, I was on a plane to Accra to help train plant breeders, crop physiologists, statisticians and data scientists at the first dedicated photosynthesis hackathon. A few notes on what made it work, on what I learned about training a room full of domain experts, and on what the platform underneath quietly enabled.
In the ever-changing world of software development, the need to adapt and innovate is constant. One way to maintain efficiency and relevance is to adopt a data-driven approach, and integrate it deliberately into the development lifecycle.
Pre-commit hooks promise to catch issues early. In practice they punish work-in-progress commits, get bypassed when convenient, and let the actual quality bar sag. Branch-protection and PR-level checks are a better place for the rules to live.