Everything's an experiment.

Idk, I spend a lot of time these days thinking about most of the ethics trainings I’ve been in hammer on “it doesn’t matter if it’s all above-board, it’s how it looks”

Brandon Friedman (@brandonfriedman.bsky.social)

Schumer's daughter works for OpenAI. Gillibrand's son founded a new crypto exchange backed by billionaires. In case you were wondering why things are the way they are.

“Oh, inspired by the Starship Troopers uniforms? Well those seemed like some nice kids, this sounds great”

derek guy (@dieworkwear.bsky.social)

this is the kind of design costume designers make to tell you who are the bad guys in a space movie

Got some really sad news today. Life is too short. Gonna go tear up at the pothos cutting I’ve been nursing back to health after somehow killing its original planting

post: Claude'ld

Pushed 2 commits to mattpolicastro/nuxt-mattpolicastro.com • feat: add local draft previews and chat transcripts

Claude'ld

Actually, no, it is you.

Log the planner merge and the constant-covariate fix

Pushed 9 commits to mattpolicastro/causal-impact-wasm • Drop zero-variance covariates instead of failing on them • Merge PR #1: Add a "Plan a test" mode (prospective power simulation) • Follow the calendar, not the array index, when resampling noise • Read the event calendar from an exclude column in the CSV • Let the analyst exclude known events when planning • Make the planner's answer reproducible, and match the analysis model • Add a "Plan a test" mode that estimates required runtime • Add prospective power simulation for choosing test duration

Landing: one stylesheet, one theme script, shared bundles, favicons wired

Pushed 4 commits to mattpolicastro/alphabeta • Lab: sample-size and detectable-lift are one app; schema bumped to v2 • graph-canvas: implement the settled brand spec; fix dark mode • Favicons: αβ from the real Commissioner 900 outlines, light/dark adaptive

graph-canvas: /bet/attach — a lab result becomes evidence on an existing bet

Pushed 26 commits to mattpolicastro/alphabeta • Lab: /lab/results, /lab/bayes, /lab/sequential; srm attaches rather than locks • packages/analysis: results, bayes, sequential, and the special functions • CLAUDE.md: no duplicated CSS or JS — extract it • Brand v0: settle the canvas weight/color rules; baseline-align the topbar • CLAUDE.md: lab CTA rule (lock vs attach) and closed-form engine policy • Scrub alphabeta.run as the app's domain; hold it for a future agentic name • WORKLOG: close out 2026-09-04 • graph-canvas: docket is a due-list; the facilitator knows what you're looking at • graph-canvas: prodbar back to places + two menus; one overlay shell • Lab: /lab/pre-post — causal-impact-wasm served under the lab path • CLAUDE.md: the static build ships the dock via BYO key now • CLAUDE.md: status, definition of done, registry rule, packages/analysis, funnel • graph-canvas: typecheck is clean and a script • graph-canvas: fix isFunnelLanding call in the tray initializer • graph-canvas: phase 4 — funnel landing, typed intake, BYO-key facilitator, resolve vs rung • Lab: /lab/detectable-lift and /lab/srm • WORKLOG: 2026-09-04 — roadmap, phases 1–2, first lab tool • graph-canvas: phase 2 — cockpit, instrument ladder, compiled criteria, ledger filters, admission • graph-canvas: guard clear/reset with a confirm and one undo • graph-canvas: phase 1 — registry chips, the loop tray, export/import, vertical default • Landing: capability registry, /capabilities page, /lab index • CLAUDE.md: pin the /lab/* route policy • Docs: September roadmap — quarry apps/web into the live canvas • graph-canvas: move layout controls out of the global nav onto the canvas • Docs: openexperiment — build brief and one-page flows

build: BASE_PATH env for deploying under a sub-path

Pushed 1 commit to mattpolicastro/causal-impact-wasm

WORKLOG: record the history rewrite

Pushed 9 commits to mattpolicastro/alphabeta • Untrack compiled rubric and eval grades; broaden ignore rules • Docs: alphabeta.tools is canonical; record live deployment • graph-canvas: VITE_STATIC build mode for the public demo • Landing site: publish αlphaβeta landing page at alphabeta.tools • WORKLOG: 2026-08-16 → 08-25 — foundations pivot, prototype, grammar, brand v0 • apps/web: allow LAN and tailnet dev origins • Eval harness: local-model replay and rubric-gaming experiments • Shape docs: event grammar v0–v0.3, stress tests, literature review, rubric, surfaces