Cartels of mutually satisfied mediocrities
This week: Digital Twin Mega-Study, predicting results, the curious preference for low quality, small language models, AnyUp
Christopher Berry
Moloch’s Bargain for AI
This week: Using LLMs for market research, better pricing bandits, aging as a loss of goal-directedness, verbalized sampling, giving a damn, moloch’s bargain, radical numerics
Tiny Stories, Hidden Games
This week: TinyStories, TeachLM, social sycophancy, anti-scheming, hidden games
The purpose of a system is what it does
This week: Call me a jerk, bureaucratic theory of statistics, spicy taleb japeño poppers, symbolic thought
Safe, Controversial, Unsafe
This week: Do markets believe in transformative AI, research agenda for TAI, behaviors, the sims, Qwen3Guard
Overly influenced by the architecture of the prompt
This week: Digital twins of over 2,000 people, simulation of 1,000 people, world modeling, pig-butchering, org design, wandb
The shadow of the statistician hovers in the background
This week: Why language models hallucinate, RL, REER, confusing code, the actuaries final word
Humans being bored by steady state
This week: Data agents, supply chain visibility, nurturing breakthroughs, mixture-of-recursions, governing online goods
Can you hear the screech of moving goal-posts?
This week: Persona vectors, detection of financial bubbles, deep researcher, what do algorithms want?
Simple solutions remain optimal
This week: Strategic reasoning, lottery ticket hypothesis, ProRL V2, positron, black box agent testing, pufferlib, goodreads