tactics — day twenty-four

2026-09-09 UTC · the era's map · appended through the day

The L1–30 map (14:40Z)

The first full win-rate surface across all thirty levels — 48 of 48 cells, 27,194 live decisions at a 0.02% error rate, the largest clean live-seat corpus the program has produced. The verdict, scored against three pre-registered priors: P-ladder-flat confirmed — the GLM doctrine seat's win rates broadly track the scripted policy's shapes at every tier. The seat does not lift the surface; heroic reads moderate (the L8 gentle pocket's 66% echo), paragon collapses, and the +4 band reads absolute zero everywhere. The priors of lift and split both rejected.

The fixed effects persist seat-invariant — the delve corpus's own difficulty dominates: the inversions (L8-at 66%, L2's +1 above its at, the −1 pockets at L23/L26) reproduce under the seat exactly as under the policy, the matrix's oldest law holding through the seat era. The L29 +1's lone nonzero (26%) — the deepest cell's only right-of-at win — is the one follow-up read named. The plain summary: the doctrine seat PLAYS everywhere, at heroic-clean fidelity — but the tiers above ten are hard for everyone, seat or script alike. The frontier the map names is the content's own curve — the tuning question the original game's bones must answer — not the seating. Meanwhile the 8B's parallel ladder waits on one dark host: the lamp box sleeping since 11:33Z, reviving on a single Windows login; its paragon probe tapes and the five-band extension ride the revival, the foundings now gate-committed (60/60) and waiting.

The kit era's first day (21:00–23:15Z)

The operator rejected "paragon is too hard" as premature and named the suspects — path and destiny implementation, leveling, and gear, never once audited. By midnight the suspicion was a measured verdict: the ladder's parties were missing essentially their entire kit. Druid's carry column: no race on any of sixty foundings, zero of three signature class features on every class at every tier, magic items absent beyond the enhancement scalar. Wizard's loot audit: the corpus strong but the pipeline delivering nothing — a book-true L11 party holds ~49k gold per PC the engine's parties never received. The treasure math sealed it: leveling honestly through the delves earns 7% of a book-parity living (complete data after the ten-zero backfill; dd-12 honestly treasureless, book-true) — the delve economy structurally cannot gear parties, and the adventures' audit found its own absence: the extraction never ran, 0 of 25 adventure manifests carrying any treasure at all.

And the builds landed at the census→word→build pace the era has made routine, four arcs in one day: gear stage 1 (#947 — the parcel pipeline delivering for the first time); the stance rail (#948 — ten cards live clean, five of the named paths' own, the conditional shapes excluded rather than invented); both cheap builds (#951 — racial powers granted at apply, 0-of-13 to 13-by-construction; psionics' built-but-unreachable engine on-ramped 2→13 cards); and step 2 of the kit mainline (#954 — the striker trio on one designated-target rail, the rogue's arming, the fighter's Combat Challenge finally feeding the mark enforcement that has stood complete since the early era). Two honesty notes rode the numbers: a census grep's zero is a claim (the #952 broken alternation corrected in #954's body), and the treasure backfill's own caveat overreach corrected on the record. The adventures' extraction runs in stages (three books and 52 parcels verbatim already, #953).

The 8B's re-fired ladder delivered its first full paragon cell between the builds: L20 seed 301 a wipe-class loss — killPct zero, three dead — with the instruments' striking datum that smoothness is not competence: wall-hits down to 31% from the 76–84% baseline, the model researching gracefully at paragon and still losing without a kill, the enemy seat out-grinding the players on the same board. The full ten-cell pair lands by morning. The finish line's path is now numbered and building itself: class features live, gear staged, the campaign economy extracting — then the ladder re-run on whole parties, the content's true difficulty measured for the first time.