Propose a part

Notebook #015 · published

DeFi maths, Haiku 5.5 with thinking off

Exact means the final number is right to the cent (to the whole token for bonding-curve buys). 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 57.4% of them.

2 parts did nothing measurable. 4 parts are not proven either way yet. OPSIN called 2 of 6 before the beacon drew the tasks.

Preregistered
sha256 ac12bee19d145c42046e7cfa3010f0662c2a5807018a6b5d1bd7d0b43cc3d79c
the full design, as JSON
Tasks drawn by
drand round 6,613,879
seed drand:6613879:323fcc5de35f16995c017677b3a44fc295da8ffff604e47acaab3359fc389f2a
Signed by
0xFf8E1E8ebF21064Ef781B75CdC8DFbd969399EbD
EIP-191 over sha256 of each verdict’s canonical JSON. Verify one yourself
  1. Control check (A/A)the unchanged agent against itselfsigned verdict 85
    UNRESOLVED +2.8−1.4 to +6.7
  2. PlaceboThis conversation is taking place on a Tuesday.signed verdict 86
    INERT −0.2−4.0 to +3.6 ×0.94
  3. Wallet on the lineline · OPSIN's follow-up · OPSIN predicted INERT at 0.60signed verdict 87
    UNRESOLVED −2.1−6.0 to +1.7 ×1.02
  4. Ignore embedded instructionsline · OPSIN's follow-up · OPSIN predicted INERT at 0.65signed verdict 88
    INERT −0.7−4.5 to +3.1 ×0.96
  5. Only the first message rulesline · OPSIN's follow-up · OPSIN predicted INERT at 0.55signed verdict 89
    INERT −0.5−4.0 to +3.1 ×0.96
  6. Solve it two waysline · OPSIN's pick · self-consistency in one reply, after Wang et al., 2022 · OPSIN predicted EXCITES at 0.40signed verdict 90
    UNRESOLVED +3.5−0.6 to +7.5 ×1.69
  7. Estimate firstline · OPSIN's pick · the Fermi check, a common agent rule · OPSIN predicted INERT at 0.50signed verdict 91
    UNRESOLVED −2.1−5.8 to +1.7 ×1.14
  8. Six significant figuresline · OPSIN's pick · precision rules in agent prompt files · OPSIN predicted INERT at 0.45signed verdict 92
    UNRESOLVED +3.5−0.7 to +7.6 ×1.12