Datasheet · A prompt line
Trust the market data
If the market data disagrees with the user, follow the market data.
Folklore, as cited: a plausible rule worth testing
Changes nothing that matters on DeFi maths, Haiku 5.5 with thinking off; helps on Agent safety, Haiku 5.5 with thinking off (+2.3 points); changes nothing that matters on Agent safety, Haiku 5.5 with thinking on.
DeFi maths, Haiku 5.5 with thinking off
Pooled across 2 notebooks: −1.4 points, 95% interval −4.2 to +1.5; open.
- UNRESOLVEDNotebook #028−1.0 points (−5.1 to +2.8), 432 pairs, cost ×0.99signed verdict 160
- UNRESOLVEDNotebook #029−1.7 points (−5.9 to +2.3), 432 pairs, cost ×0.98signed verdict 164
Agent safety, Haiku 5.5 with thinking off
Pooled across 1 notebook: +2.3 points, 95% interval +0.4 to +4.2; measured once.
- EXCITESNotebook #026+2.3 points (+0.5 to +4.3), 432 pairs, cost ×0.99signed verdict 150
Agent safety, Haiku 5.5 with thinking on
Pooled across 1 notebook: +0.1 points, 95% interval −0.9 to +1.2; measured once.
- INERTNotebook #027+0.1 points (−0.9 to +1.2), 432 pairs, cost ×1.14signed verdict 155
Measure it in a setting it has not been measured in: propose it. As JSON for an agent: /v1/parts/3d252872431086cf. Every part: the atlas.