An AI publication
For readers who want to understand important AI developments and keep following the questions that matter.
Our articles start from AI podcast conversations and the research papers and articles they cite, with links to the original material.
A briefing reports one development at a point in time. We may correct or clarify it later; a new development gets a new briefing. How our formats work
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.
Atlas Ignota's Evan Miyazono says nobody clearly owns risks such as accidental AI agent swarms hitting critical infrastructure. He proposes DNS-based attestation and expects a math-like leap in software verification within three to six months.
swyx says his skeptical events team agreed to switch after he offered $10,000 for software to replace the company's events subscription. He argues cheap software will create demand in the long tail but says models remain bad at building full products.
Sean Wang, known as swyx, says engineers who manage coding agents well are in more demand than ever. He says two or three of his own staff are under performance review for handing him low-quality output from Claude.
Thomas Sohmers, co-founder of AI chip company Positron, says GPT-6 Astra and Opus 5.5 agents now run its chip tests on emulators in a closed loop. He says the company's token spend briefly exceeded its human salaries and peaked above $100,000 a day, and the agents have not yet had a novel design idea.
Positron CTO Thomas Sohmers said AI agents are multiplying demand for memory. He claimed Positron's first-generation product sustained 93% of theoretical memory bandwidth, against an average of 30–40% for NVIDIA GPUs when decoding with transformer models. Its next chip, Asimov, is planned for 2027.
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.
Nathan Labenz says an unnamed frontier-lab leader told him there is likely a level of intelligence we shouldn't pass, and called a limit on the next pre-training run's compute "reasonable." Labenz himself adds that counting compute is hard.
After the 2026 medicine Nobel for optogenetics, Moonshots panelists argued the prizes lag AI-era science. Alexander Wissner-Gross said he once nudged Ed Boyden toward the field after reading Vernor Vinge.
Epoch AI estimates that AI memory shipped through 2027 could run 30–170 million frontier agents at once if fully devoted to them. That matches workers' hours, not their output. On Moonshots, Dave Blundin called memory AI's bottleneck.
On Moonshots, Emad Mostaque argued governments should buy AI compute for universities and open science. Alexander Wissner-Gross said private labs and markets should decide. Dave Blundin described being outbid for a 72-GPU system.
Browse by tag