Despite annualized revenue jumping from $47B in May to $65B in July, Anthropic's flagship model is losing the consumer war to cheaper alternatives. The paradox: the lab is growing fast while its best product struggles for adoption.
A new category of AI-generated abuse imagery — slightly altered to evade detection — is flooding X. The subtle modifications make platform moderation and legal enforcement significantly harder to apply.
A six-month freeze on major electrical infrastructure is the latest local tactic to delay a massive data center built to support America's nuclear arsenal. It's a preview of how communities will fight the AI buildout on energy grounds.
Torvalds publicly acknowledged AI did the grunt work on an otherwise unsolvable bug — while also noting it repeatedly declared the problem impossible. A credible, unvarnished endorsement from the most skeptical of sources.
A16z's latest charts confirm that humans are now the minority user of AI infrastructure. Agents are consuming compute at a scale that dwarfs individual usage, and the gap is accelerating fast.
Bronze Age Arabian pottery showed consistent visual identity, material quality, and cultural reputation — the full toolkit of a modern brand. Marketing didn't start with Madison Avenue; it started with clay.
A16z dropped the stat of the week: AI agents now consume nearly five times the tokens that human users do, a figure that has grown 14x since February. This isn't a usage trend — it's a structural shift in who (or what) the internet's compute is actually serving. The infrastructure assumptions of cloud pricing, API design, and even security are built for humans. That era is ending.
Andrew Ng flagged the Marin project as a critical demonstration of openness in AI — open code, open data, open training recipes. With regulatory and competitive pressure pushing toward closed systems, Ng's public endorsement of Marin reads as a deliberate signal: the open-source coalition is paying attention and organizing.
Ethan Mollick called out a real methodological rot in AI research: papers measuring AI's impact using outdated models, without adequately flagging the limitation. In a field where a three-month-old benchmark can be irrelevant, publishing results from older models without clear caveats is closer to misinformation than science.