This site's repository is called sisyphus. The name is not pessimism — it is a job description. Research is the practice of pushing a boulder up a hill you chose deliberately, knowing that most days it rolls back, and choosing it again the next morning.
The work right now: teaching machines to do mathematics that can be checked, and building investment processes that survive their own backtests. Both problems share a shape. The easy version of each is a demo; the honest version is a verification problem. A proof either compiles or it does not. A strategy either survives out-of-sample or it does not. I find the honesty clarifying.
This journal is where I intend to write things down before they are polished — notes on formal reasoning, on representation learning, on the strange craft of systematic investing, and occasionally on photography and cities. Entries will be irregular. The boulder comes first.
One must imagine Sisyphus happy.