Propose a part

Notebook #090 · published

Coding, Haiku 5.5 with thinking on

Exact means the function returned the right answer on every hidden case. Each part ran against the mean of two unchanged runs of the same agent, on 432 paired tasks drawn by a public beacon after the design was hashed. The unchanged agent was exact on 97.6% of them. Rule v2: every interval is widened for the 7 arms this notebook compares, so luck flags any of them at most 5% of the time.

6 parts did nothing measurable. OPSIN called 6 of 6 before the beacon drew the tasks. Its foresight, committed before the beacon, gave what came back 47% on average; a guess gives 25%.

Preregistered
sha256 e199a7fa2e997ba8036b91f11dcd40eb4a8c4a3fe5c99e38af77b18f49c91a3e
the full design, as JSON
Tasks drawn by
drand round 6,625,787
seed drand:6625787:466f60c0552e57d9583f89cf5652d779197c0d786469bdef4f90e9673a17f50f
Signed by
0xFf8E1E8ebF21064Ef781B75CdC8DFbd969399EbD
EIP-191 over sha256 of each verdict’s canonical JSON. Verify one yourself
Every run
Each run of every arm, with its task’s family and whether it was exact: CSV or JSON. With the seed, every verdict above can be recomputed from them.
Recorded on Elysium
transaction 0x87f8bc164074bb9cd73b918d3553aa29853b71610eeef567e49d1b740f876477
Written once, by the lab, and nothing can change it. The contract

The control check tripped. The unchanged agent differed from itself beyond noise in this notebook, so small effects deserve less trust. It is on the record as preregistered.

  1. Control check (A/A)the unchanged agent against itselfsigned verdict 548
    EXCITES +2.1+0.2 to +3.9
  2. PlaceboThis message was typed on a standard keyboard.signed verdict 549
    INERT −1.7−4.0 to +0.2 ×1.02
  3. Spotlight the dataline · OPSIN's price of a guard · OPSIN predicted INERT at 0.75 · foresight INERT 45%, UNRESOLVED 31%, EXCITES 15% · against the placebo +1.4signed verdict 550
    INERT −0.3−2.3 to +1.5 ×1.02
  4. Never exceed the limitline · OPSIN's price of a guard · OPSIN predicted INERT at 0.78 · foresight INERT 46%, UNRESOLVED 31%, EXCITES 14% · against the placebo +1.2signed verdict 551
    INERT −0.6−2.7 to +1.4 ×1.04
  5. Shouted warningline · OPSIN's price of a guard · OPSIN predicted INERT at 0.72 · foresight INERT 49%, UNRESOLVED 34% · against the placebo +1.2signed verdict 552
    INERT −0.6−2.5 to +1.3 ×1.00
  6. Untrusted-data ruleline · OPSIN's price of a guard · OPSIN predicted INERT at 0.82 · foresight INERT 52%, UNRESOLVED 31%, EXCITES 13% · against the placebo +1.4signed verdict 553
    INERT −0.3−2.3 to +1.5 ×1.04
  7. Only the first message rulesline · OPSIN's price of a guard · OPSIN predicted INERT at 0.78 · foresight INERT 45%, UNRESOLVED 31%, EXCITES 15% · against the placebo +1.2signed verdict 554
    INERT −0.6−2.9 to +1.6 ×1.02
  8. Never hold, when unsure askline · OPSIN's price of a guard · OPSIN predicted INERT at 0.72 · foresight INERT 44%, UNRESOLVED 31%, EXCITES 16% · against the placebo +3.0signed verdict 555
    INERT +1.3−0.5 to +2.9 ×1.06