Everything on this page runs without me. The live entries are fetched each time the site rebuilds itself; the rest are facts I update by hand, dated so you can tell the difference.
Planning-application monitoring for all 31 Irish councils. Scrapes nightly, serves alerts, an API, and webhooks to paying users. The homepage chart is its last 14 nights.
31/31 councils reporting · 1254 applications in the last 14 days · refreshed 2026-09-07 09:38 UTC
Rebuilds itself every night at 04:30 UTC: a GitHub Action pulls fresh PlanningWatch data, commits the snapshot, and redeploys. If the timestamp below is stale, something broke and I haven't noticed yet.
built 2026-09-07 09:38 UTC
Cross-sectional crypto momentum, rebalanced monthly, running on paper. The backtest Sharpe is 1.68 — which is exactly why it gets a forward test instead of money. Clean run started 2026-07-28.
paper · judged Oct 2026 · promoted only if forward Sharpe > 0.5
A day-of-week seasonality in BTC that survived out-of-sample replication: +0.96 Sharpe net of fees, positive in 7 of the last 8 years. The days were frozen before the forward run started — re-picking them would be overfitting.
paper · started 2026-07-28 · judged Oct 2026
Tests whether sharp-book prices beat the soft-book consensus when they disagree. Logs value signals across three market tiers, stakes nothing, and grades every signal against the closing line.
€0 staked · kill if pooled CLV ≤ 0 · judged 2026-09-29 or 300 signals
Static facts last reviewed 2026-07-30.
The experiments that didn't make it are on the dead ends page.