https://s2ar.dev/starscry/blog.htmlWriting from s2ar.dev: reward integrity (starscry) and certified image compression (starlens) — every claim replayable.https://s2ar.dev/starscry/the-refusal-nobody-measures.htmlhttps://s2ar.dev/starscry/the-refusal-nobody-measures.htmlSat, 18 Jul 2026 00:00:00 +0000starscryThe industry counts how often models comply with harmful requests and barely counts how often they refuse legitimate ones. Over-refusal is a real alignment failure — and its headline metric is a self-graded scalar that misranks its own subjects. The case for measuring it like a calibration.https://s2ar.dev/starscry/why-a-certificate.htmlhttps://s2ar.dev/starscry/why-a-certificate.htmlThu, 16 Jul 2026 00:00:00 +0000starlensImages bill by dimensions, not bytes — and every existing answer to 'how far can I shrink this?' is an aggregate. The case for a per-request, per-model, sealed fidelity certificate, priced as a share of the savings it proves.https://s2ar.dev/starscry/jpeg-knob-saves-zero.htmlhttps://s2ar.dev/starscry/jpeg-knob-saves-zero.htmlThu, 16 Jul 2026 00:00:00 +0000starlensWe requantized to quality 25 and measured the token bill: exactly 0.0% saved. The knob the industry ships optimizes a number vision-LLMs don't bill on — here is what they bill on instead, per provider.https://s2ar.dev/starscry/average-is-not-a-guarantee.htmlhttps://s2ar.dev/starscry/average-is-not-a-guarantee.htmlThu, 16 Jul 2026 00:00:00 +0000starlensA pixel-error metric calls a 32-pixel-wide image 'identical' while every model we tested stops reading at six times that size. On held-out content the naive policy silently loses 42–50% of answers — while reporting the best savings number on the board.https://s2ar.dev/starscry/certificates-name-their-model.htmlhttps://s2ar.dev/starscry/certificates-name-their-model.htmlThu, 16 Jul 2026 00:00:00 +0000starlensA compression that one model reads perfectly loses another 29% of its answers at the same rung. Cross-model transfer is one-directional — which is why a fidelity guarantee that doesn't name its reader isn't one.https://s2ar.dev/starscry/blog-post-2026-07.htmlhttps://s2ar.dev/starscry/blog-post-2026-07.htmlMon, 06 Jul 2026 00:00:00 +0000starscryA public RL environment whose reward a seeded search games in one simulation — and why the fix is to construct reward-hacks deterministically and ship the apparatus with the claim, not to detect them with a bigger model.https://s2ar.dev/starscry/scientific-post-2026-07.htmlhttps://s2ar.dev/starscry/scientific-post-2026-07.htmlMon, 06 Jul 2026 00:00:00 +0000starscryThe platform plus a season of pointing the instruments at my own hypotheses: three refutations with named mechanisms, and two results — a capability floor and a solver-effort certificate — that survived the same guillotine.