Across AI sources
What researchers, builders and critics are saying about AI.
Their arguments, examples and disagreements, with links to the original material.
RoboCurve reports that GPT-6 Astra placed a block in a bowl in 19 of 20 trials, up from 8 of 20 for its predecessor. But it completed only two puzzle insertions—the same as the older model—in a 120-trial evaluation that separates basic manipulation gains from reliable precision.
A briefing reports one development when it happens. We correct or clarify it later; a new development gets a new briefing. How our formats work
An Anthropic Institute working paper models an extreme scenario in which annual real GDP growth reaches 15% by 2030 while 17.9% of cognitive workers are unemployed. On Moonshots, Emad Mostaque questioned whether demand could survive the disruption. The disagreement turns on how workers find new income—and who owns the machines.
A controlled comparison of 2019–2025 datasets and training recipes reported compute-efficiency gains of 12-fold from data improvements versus 3.7-fold from model recipes. The Moonshots panel explored the business opportunity—and used BloombergGPT’s reportedly short-lived advantage to question whether owning unique data is enough.
Google DeepMind’s AlphaGenome Atlas, announced on 8 September 2026, makes predicted molecular effects for roughly nine billion single-letter DNA changes available for researchers to look up. The Moonshots panel sees an opportunity to widen access and help patient groups organize research—but the path from a prediction to a treatment still runs through outcome data and experiments.
Insilico Medicine has announced first dosing in a 52-week Phase 3 trial of rentosertib for idiopathic pulmonary fibrosis. A separate analysis of 42 earlier-trial participants found younger-looking blood-protein profiles, prompting a Moonshots debate about longevity—but not establishing rejuvenation or longer life.
DeepSeek says V4.1-Flash cuts the high-bandwidth memory needed for its cache to a quarter of its predecessor’s requirement. The Moonshots panel sees an opening for companies to build on efficient open-weight models—but a smaller cache is not the same as a small model.
OpenAI says roughly 10,000 coordinating agents produced a proposed proof of breakdown in a forced three-dimensional fluid flow, with humans consolidating the work and Lean checking the formal argument. On Moonshots, Sam Altman’s call to pace progress divided the panel: was this an anticipated scientific capability, or a surprise about how quickly and cheaply it could be reached?
Follow a thread