Woolly Mammoth, Dodo Get Another Shot as Startup Raises $200 Million

OpenAI researcher says they have an AI recursively self-improving in an "unhackable" box

MiniMax-Text-01 - A powerful new MoE language model with 456B total parameters (45.9 billion activated)

"MangaNinja: Line Art Colorization with Precise Reference Following", Liu et al 2025

"The 27-Year-Old Billionaire Whose Army Does AI’s Dirty Work" (Scale data-labeling failures: 27k bogus Q&A, many starting 'as an AI language model...')

The 27-Year-Old Billionaire Whose Army Does AI’s Dirty Work

"A Spymaster Sheikh Controls a $1.5 Trillion Fortune. He Wants to Use It to Dominate AI" (G42/Microsoft/Brad Smith/Huawei/Nvidia/Cerebras/...)

A Spymaster Sheikh Controls a $1.5 Trillion Fortune. He Wants to Use It to Dominate AI

[Article] "Types of Environmental Enrichments Offered for Cats and their Association with Housing Features and Cat Personality", Goncalves et al 2025 (T&F)

[Discussion] I trained an AI model to generate Pokemon

Cardinal-valued Secretary problem: set the threshold after √n candidates, not n/e

"Decisions under Risk Are Decisions under Complexity", Oprea 2024 (behavioral economics biases might be because people are dumb, not irrational)

"The Devil is in the Tails: Fine-grained Classification in the Wild", Van Horn & Perona 2017 (the Inception pretrained model didn't provide meaningful transfer)

"TSMC begins producing 4-nanometer chips in Arizona, [US Commerce Secretary] Raimondo says"

"When each proud fighter brags", Rheingans-Yoo (Max Chiswick, 1985–2025)

What’s going on with all these CEOs who drastically change their appearance over time?

[Article] "The Fiscal Impacts of Wildfires on California Municipalities", Liao & Kousky 2022 (UChicago)

Former OpenAI employee Miles Brundage: "o1 is just an LLM though, no reasoning infrastructure. The reasoning is in the chain of thought." Current OpenAI employee roon: "Miles literally knows what o1 does."

Redis creator antirez (Salvatore Sanfilippo) on Technical Blogging

"The tremendous gain of OpenAI's o3 may be overstated by ARC, because it's the first model able to operate on pixel grids of problem length that ARC happens to exist in" (humans underestimate the difficulty of 2D perception for LLMs, and it's this aspect of ARC-AGI that o3 scaling tackled well)

[Article] "Transplantable glucagonomas derived from pluripotent rat islet tumor tissue cause severe anorexia and adipsia", Madsen et al 1995

What happens to "high finance" as AI continues to advance?

"NeuralSVG: An Implicit Representation for Text-to-Vector Generation", Polaczek et al 2025

"20 Years of Bitext", Peter Brown & Bob Mercer 2013 (on early NMT, n-grams, finding & cleaning large linguistic corpora)

Waymo Picked Up 4 Million Rides in 2024, Tesla Had Zero

"Proactive Agents for Multi-Turn Text-to-Image Generation Under Uncertainty", Hahn et al 2024 {DM}

"High-resolution genomic history of early medieval Europe", Speidel et al 2025