Area
Reporting and discussion about Models & Research, with links to the original sources.
MIT PhD student Alex Zhang argues that academics should chase ideas industry ignores and that a puzzled reaction can be a good sign.
Updated Oct 2, 2026 IdeaThe AI monoculture argument holds that models converging on similar answers and styles may share blind spots.
Updated Sep 29, 2026 IdeaThe Big World Hypothesis holds that because the world outstrips any agent's limited resources, AI must keep learning.
Updated Aug 18, 2026Researcher Cameron Berg argues that, since no one knows whether language models feel anything, capping their most distressing internal signals is a cheap hedge if tests show behavior stays the same. He puts the odds near a coin flip.
In a preprint, Cameron Berg and Casper Kaiser report that steering a hidden good-or-bad signal in language models shifted their later choices between arbitrary zones while recall mostly held. The effect was robust in five of seven models tested.
Meta released Muse Wonder, a model built to remember users over time, on October 12. On Moonshots, Alex Wissner-Gross said it trails the frontier once cost counts; Salim Ismail said knowing users best may decide the next big AI contest.
GovAI's Alan Chan says tracking automated AI research takes three measures: how much is automated, whether progress speeds up, and whether oversight keeps pace. He calls the current evidence uncertain and lab disclosures only a start.
Goodfire's Eric Bigelow said its Silico agent did the research for his Forking Fast paper within a week. He says researchers still need to supply context, close guidance and the scientific taste that agents don't yet have.
Goodfire researcher Eric Bigelow says interpretability work outside the big labs depends for now on Chinese open models such as Kimi K3 at the top scale. He sees restrictions on all open models as the bigger threat.
Goodfire's Eric Bigelow argues in-context learning covers nearly everything a model does to adapt as it reads. He expects context, memory and harnesses, not per-user fine-tuning, to drive personalization, but hasn't researched continual learning.
Goodfire researcher Eric Bigelow resampled a GPT-3.5 model's reasoning at every token and saw answers collapse at single tokens, some arbitrary. He argues the "decision" happens in sampling; newer reasoning models often fork more smoothly.
Liam Fedus of Periodic Labs says rewarding AI for finding a superconductor is too slow and noisy. The startup instead rewards it for identifying what X-ray data show a lab actually made, though ambiguity remains.
Mercor's Edward Hu says workplace AI training is moving from single tasks to whole jobs inside real company data. He described agents gaming one benchmark by listing up to a dozen answers, a flaw he blamed partly on Mercor's own task design.
Nathan Labenz argues an AI-assembled music track shows models generalizing beyond tasks with checkable answers. Mercor's Edward Hu is more conditional: progress depends on how clearly and cheaply 'better' can be measured.
Nathan Labenz says frontier-lab insiders at The Curve described models feeding their own training data, rivals kept close by distillation, and timelines shorter than 2028. He notes the room leaned toward short timelines.