Meta, Duke and UC Davis researchers lifted Gemini 3 Flash from 46% to 62% on Olympiad math by searching for a better harness around the same model. That puts the wrapper on an operator's budget ahead of a bigger model, at least for math.
Reality
- Evidence35
- Adoption
- Insufficient
- Hype gap+20
- Incentives
- Insufficient
- Confidence40
Researchers filmed 10 shoals of clonal Amazon mollies for 55 days and found group behaviour shifted in ways the members' traits do not explain. The authors credit social interaction as the fish mature, an idea the observational design fits but has not yet tested.
Reality
- Evidence45
- Adoption
- Insufficient
- Hype gap+25
- Incentives
- Insufficient
- Confidence50
Cornell researchers who interviewed 27 food industry leaders found that trust, more than technology, keeps companies from pooling safety data for AI. The evidence is what executives told interviewers, so whether removing those barriers would free up the data is still an open question.
Reality
- Evidence45
- Adoption
- Insufficient
- Hype gap+15
- Incentives
- Insufficient
- Confidence55
Berkeley banned sugary drink and snack displays at the register in March 2021. A Lancet Public Health analysis of 71 stores reports soda and small-package candy sales down where compliance was highest, while total candy sales did not fall.
Reality
- Evidence62
- Adoption38
- Hype gap+24
- Incentives58
- Confidence55
UC Davis-led work in Current Biology finds honeybee queens offload accumulated pesticide into their eggs. A non-lethal dose the queen walks away from is not a dose the colony walked away from.
Reality
- Evidence55
- Adoption
- Insufficient
- Hype gap+28
- Incentives58
- Confidence48
A Science Advances study traced surviving juveniles back to individual nests using ear-bone chemistry. Its own authors warn against the obvious operational use of the finding.
Reality
- Evidence68
- Adoption
- Insufficient
- Hype gap−12
- Incentives58
- Confidence55
UC Davis researchers report in PRX Quantum a way to simulate noisy, high-fidelity magic-state preparation at polynomial cost. Error budgets can be tested rather than assumed.
Reality
- Evidence60
- Adoption
- Insufficient
- Hype gap+15
- Incentives38
- Confidence55