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 say two 5th-century BC Etruscan tombs were built as immersive sensory spaces, citing exceptionally long echoes in one of them. Their own account leaves open whether the sound was designed or simply came with the room.
Reality
- Evidence28
- Adoption
- Insufficient
- Hype gap+40
- Incentives60
- Confidence35
Researchers say two fifth-century BC Etruscan tombs were built for sensory effect, citing exceptionally long reverberation in the Tomb of the Blue Demons. Intent stays an interpretation until the team publishes reverberation times measured against comparable chambers.
Reality
- Evidence28
- Adoption
- Insufficient
- Hype gap+45
- Incentives60
- Confidence32
Duke's Sarah Cooley and colleagues measured a snow-and-ice cooling effect of -14.4 W/m2 over Northern Hemisphere lakes, against -8.2 over nearby land. Climate models often do not separate lakes from land, though the study as reported does not say how much that misstates the seasonal energy balance.
Reality
- Evidence55
- Adoption
- Insufficient
- Hype gap+10
- Incentives
- Insufficient
- Confidence50
Max Planck researchers followed 21 wild orangutans to independence and found that each infant's mix of watching and exploring tracked its eventual diet. Strong watchers who rarely explored still ate widely, though the diet data cover 13 animals and end at independence.
Reality
- Evidence60
- Adoption
- Insufficient
- Hype gap+20
- Incentives
- Insufficient
- Confidence58
Neurophos has made Kymeta founder Nathan Kundtz its chief technology officer, opened a Sunnyvale office and scheduled a low-rate Austin production line for later this year, with the chip due in 2028 and no benchmarks published.
Reality
- Evidence45
- Adoption6
- Hype gap+42
- Incentives70
- Confidence55
The award to Tempus funds product development and a prospective multicenter study of an autonomous version of its Olivia app, and ARPA-H wants an FDA authorization package filed within 24 months.
Reality
- Evidence48
- Adoption8
- Hype gap+12
- Incentives68
- Confidence58
Ditches dug decades ago for a farming venture that was abandoned leave the tract emitting 130,000 tons of carbon dioxide a year. The plan to plug them rests on credit buyers whose price nobody has published.
Reality
- Evidence57
- Adoption
- Insufficient
- Hype gap+30
- Incentives72
- Confidence47
A Duke-led study of nearly 200,000 résumés put the rate near 1%, and two founders found examples in their own queues last month. The record still stops short of showing that any injection changed an outcome.
Reality
- Evidence52
- Adoption38
- Hype gap+22
- Incentives46
- Confidence50
The scaffold carries astrocyte vesicle signals on its own surfaces, and thinning the neutrophil-rich cells it recruited cut new vessel growth, though the Duke summary gives no number for how far motor function moved.
Reality
- Evidence42
- Adoption8
- Hype gap+26
- Incentives70
- Confidence40
That result sits inside a growing set of studies where generative AI raises the rated quality of a single piece of work and lowers the spread between pieces, with 400,000 indexed papers among the samples.
Reality
- Evidence58
- Adoption54
- Hype gap+14
- Incentives42
- Confidence57
A subcutaneous myocyte graft gave mice muscle, bone and immune gains. The mice were also free to roam their cages, which is the one condition the intended patients do not share.
Reality
- Evidence60
- Adoption8
- Hype gap+32
- Incentives55
- Confidence66
Chief economist Ronnie Chatterji says the brief "is changing a lot." If the lab with the best view of the roadmap writes its research plan in pencil, no hiring manager should be writing theirs in ink.
Reality
- Evidence34
- Adoption
- Insufficient
- Hype gap+28
- Incentives70
- Confidence48
A new grading method scores models on satellite radiation trends rather than past surface temperatures, and the ones previously discounted come out on top.
Reality
- Evidence48
- Adoption
- Insufficient
- Hype gap+18
- Incentives
- Insufficient
- Confidence42