joshua
-
OpenAI’s own safety report shows regressions in newly shipped GPT-6 Sol and Luna models
OpenAI’s October transparency update shows the GPT-6 models now powering ChatGPT score worse on some self-harm…
-
OpenAI’s AI Proved a 20-Year Math Conjecture. SpaceX Renamed Its AI So It Wouldn’t Say ‘AI.’
A math conjecture fell, a Nobel went partly to neural nets, and a billionaire rebranded his…
-
Attention Check: What Those ‘Rogue AI’ Wikipedia ‘Break-Ins’ Actually Were
Wikipedia’s host called OpenAI’s bots ‘rogue.’ Here’s what they actually did to Wikipedia’s servers — and…
-
Google Locked In 890 Megawatts of Nuclear Power. A Newsletter Called AI ‘The Largest Con in History.’
Real nuclear deals, real math proofs, real open-weight models — plus the week’s loudest, least-earned doom…
-
OpenAI Cut Its Flagship Price 80%. One Headline Called It Skynet.
This week: a muscle-powered swimming robot, a cheaper frontier AI model, and a ‘Terminator’ headline that…
-
Anthropic Commits $100 Million to Train 10,000 Enterprise ‘Claude Deployment’ Engineers
Anthropic’s new Claude Frontier Academy trains enterprise engineers to deploy Claude, betting talent — not model…
-
Meta Open-Sources Muse Gadgets, Letting Developers Build Custom Hardware for Its AI Agent
Meta released open-source firmware and an SDK so hobbyists can embed its Muse assistant into DIY…
-
Light-Powered AI Just Spotted Deepfakes at 98% Accuracy. A Blog Called a Chip Test ‘No Consent.’
Real this week: optical AI catches deepfakes, a helium qubit cuts errors 100x. Not real: a…
-
Attention Check: What Trump’s ‘Super Intelligence’ Renaming Actually Changes
A headline makes it sound like AI itself got renamed. In reality, it’s a terminology preference…
-
OpenAI Shelved a Model For Lying. Trump’s New AI Website Contradicted Him On Day One.
AMD paid $8.2B for the ‘godmother of AI,’ OpenAI killed a model for lying, and a…
