Commit graph

199 commits

Author SHA1 Message Date
8d4dc3b5f7 Rewrite the research prompt against the live orchestrator norm and fix what the refuter found: the zone belongs to the orchestrator and one hard-contour fact contradicted its own source 2026-08-30 23:55:54 +03:00
ebff194631 Write the harness-topology research prompt to orchestrator norm: a blocking repository sync precedes any search, because this phase proposed already-finished work five times in one day 2026-08-30 23:36:42 +03:00
cdc5a5f156 Land stage 0 item 1: the bank holds its terms, and the audit found on disk a production-profile run, a July measurement of the same question, and a ratified editor-price input this phase never executed 2026-08-30 20:10:54 +03:00
62f8f6bef9 Land the consilium: what the editor does to the text, eight retracted conclusions with the class of each error, and a staged plan that starts free 2026-08-30 19:16:55 +03:00
860c337387 Teach the rig that DeepSeek peak pricing is weekdays only, and record the vendor wording with the near-miss that a single page read caused 2026-08-30 12:55:42 +03:00
8326e35d07 Raise the cap by five so the second circle fits, and record that my own arithmetic left it one session short 2026-08-30 12:02:06 +03:00
c83aea8a0b Raise the agent-session cap on the owner word so the luna panel and its second circle both fit 2026-08-30 02:43:05 +03:00
6d85a91b20 Retract the time-shift reading of the thirty-eight-fold gap: it is three sigma against the only spread estimate we have, and price the repeats that would settle it 2026-08-30 02:33:15 +03:00
8837a0aa02 Report the interleaved run: the default path itself moved thirty-eight-fold in nine days, which measures the confound and leaves the knob magnitude unestablished 2026-08-30 02:13:39 +03:00
2acaedc910 Record that the effort knob now reaches the anchor, retract my claim that no unconfounded design fits the money, and freeze the interleaved pre-registration 2026-08-30 01:51:47 +03:00
7fabdd0634 Correct the battle-config comment that justified the vendor default by a premise the 30.08 probe retired, leaving the configuration itself untouched 2026-08-30 01:51:47 +03:00
cabc158f63 Record the probe: the effort knob on the anchor does act but only twofold, which lands in the pre-registered middle band and stops the purchase 2026-08-30 01:04:44 +03:00
75e8b44571 Record the third silent breakage of the machine move: every API key name in the polygon env lost its underscore, which blocks the sanctioned purchase 2026-08-30 00:58:38 +03:00
6bc10f9669 Write the post-compaction restoration file: the sanctioned purchase is one command away with its decision rule frozen, and the seven mistakes of this session are listed so the next one does not repeat them 2026-08-30 00:56:26 +03:00
87dd12485e Freeze the pre-registration of the anchor at low effort, with the one-cell probe rule written before the probe and the ranking endpoint declared out of reach on sessions 2026-08-30 00:53:56 +03:00
7a74cfbf90 Land the adversarial pass over my own work: one blocker on a decoy control I read after the fact, six important overstatements, and the disposition of each 2026-08-29 23:04:15 +03:00
f52466a811 Report the delivery check red with a diagnosis per line, fix the checker that counted table headings as requirements, and record the gate I almost turned green by quoting it 2026-08-29 22:44:37 +03:00
9e9e757a07 Give the polygon README the two steps a clean machine needs, the map of the phase-D zone it described as one file, and the mtime gates the move silently broke 2026-08-29 22:08:47 +03:00
d7a202edc6 Freeze the step-4 pre-registration and prove from bought cells that Gemini drops the FIRST system message, so the engine loses the stage prompt rather than the glossary 2026-08-29 21:32:01 +03:00
28c2d0fd50 Carry the polygon's uncommitted phase-D work onto its own branch: judging rigs, session logs and plans that had been living in the shared tree since the machine move 2026-08-29 13:30:17 +03:00
2d1e4250d2 Land the first cross-family reading comparison the phase has ever had: no arm of twenty-one is ordered differently, and say plainly why this still does not satisfy the two-family norm 2026-08-23 02:49:11 +03:00
c5b45e7c97 Correct a claim I published without checking: my readers ran on Opus while every earlier panel was read by Fable, so the second reading is cross-family and its headline metric is renamed 2026-08-23 02:44:13 +03:00
62710ab2d3 Register the quarantined no-prompt cells as a neutral pseudo-arm gated by token count, and add the second-reading tool that splits the threshold into reader noise and text 2026-08-23 02:31:59 +03:00
97ccf999d4 Move the reader launch wrapper into a frozen zone file with a hard rule that it may never name an evaluation axis, and cite it by hash 2026-08-23 02:09:18 +03:00
03f585755a Freeze the pre-registration of both second readings: name the baked-in answer path trap, the twelve-chapter subsample and the four metrics before any reader runs 2026-08-23 02:02:05 +03:00
bedc39a751 Freeze the panel-0 pre-registration and its counter result: register the foreign-phase arm E6, build the blind source etalon, and name three defects found in the instrument itself 2026-08-22 22:27:42 +03:00
Claude (backend session)
27fefa265e Set the panel standard at six Gu chapters by owner decision and record why it is a consequence, not a saving 2026-08-22 04:06:39 +03:00
Claude (backend session)
3cd9334f32 Fold the last consilium checks into the plan: critic-filter debt, character sheets, per-axis taste split and the open unit-conversion decision 2026-08-22 03:56:16 +03:00
Claude (backend session)
a0938dd143 Land the consilium outcome and the execution plan that doubles as a post-compaction restoration prompt 2026-08-22 03:49:06 +03:00
Claude (backend session)
8aa125f123 Add the Gemini low rung so its reasoning level is comparable with the other vendors 2026-08-22 01:07:54 +03:00
Claude (backend session)
29338dfbd3 Correct the Gemini thinking-knob and billing notes against vendor docs and ping the orchestrator about the engine losing the glossary on Gemini hops 2026-08-22 01:03:20 +03:00
Claude (backend session)
70966c219f Switch the Gemini thinking knob to the standard compatibility parameter after the vendor rejected the nested form 2026-08-22 00:39:26 +03:00
Claude (backend session)
457bc0854a Add the Gemini thinking variant with its own budget knob 2026-08-22 00:38:51 +03:00
Claude (backend session)
b222ac8a08 Record the Gemini quirk that silently drops extra system messages and merge them for that vendor only 2026-08-22 00:36:26 +03:00
Claude (backend session)
2b665fd5c4 Add the decisive probe case with two system messages, the shape the bought Gemini cells actually used 2026-08-22 00:35:02 +03:00
Claude (backend session)
20783511e0 Freeze a live probe that separates a wrong token counter from a system prompt that never arrives 2026-08-22 00:33:23 +03:00
Claude (backend session)
af0447aa6b Move the unused stopped-phase budget into the English role-prompt phase without touching the pack ceiling 2026-08-22 00:23:16 +03:00
Claude (backend session)
88a79a06f0 Add gates that prove each review fix behaves, surface the multi-instance pair mine, and raise the English role-prompt ceiling for the sanctioned purchases 2026-08-22 00:21:29 +03:00
Claude (backend session)
80aafdeb99 Fix the pair-path leak that sent Chinese editor rules to English one-pass cells, declare each arm reasoning level explicitly, and record the review errata 2026-08-22 00:12:32 +03:00
Claude (backend session)
7c63ec062c Record the vendor-portability verdict and make the tally anchor and controls configurable per panel 2026-08-21 01:19:02 +03:00
Claude (backend session)
22820288b2 Add the panel gate for truncated cells and record the judging-run audit with its two incidents 2026-08-21 00:17:20 +03:00
Claude (backend session)
26bf4f297a Freeze the erratum arm that runs glm-5 with its vendor-default thinking, the confound the bought cells revealed 2026-08-20 23:47:45 +03:00
Claude (backend session)
3433a85d1f Freeze the vendor-portability pre-registration and the panel gate that byte-checks every packet text against its cell 2026-08-20 23:38:59 +03:00
Claude (backend session)
7647579feb Skip chapters with an incomplete panel instead of aborting the whole emission, and print the skips 2026-08-20 22:45:10 +03:00
Claude (backend session)
db02b7ab64 Freeze the reading packet format: three judging axes and one chapter per packet 2026-08-20 20:38:56 +03:00
Claude (backend session)
276c85b4be Make the blind-reading panel and packet name parameters now that reading is the standing instrument 2026-08-20 19:56:14 +03:00
Claude (backend session)
3075566104 Freeze two ablation variants with the verification tail cut from both and opt-in trace capture 2026-08-20 18:52:19 +03:00
Claude (backend session)
eb49b651d4 Add the blind-reading emitter for this panel and record the role-prompt reading verdict 2026-08-20 18:06:48 +03:00
Claude (backend session)
273d2df41f Merge the two overlapping rule lists into one axis list, add the restored-paragraph arm and harden the coverage gate 2026-08-20 16:16:45 +03:00
Claude (backend session)
824bd209eb Restore three rules the first prompt draft dropped and gate every battle-prompt rule by declared disposition 2026-08-20 15:48:06 +03:00