Spies Didn't Sink Nvidia. Voters Did.
US intelligence reportedly found a Chinese asset close to Jensen Huang. Wall Street shrugged. Then the midterms took 9.3% off Nvidia in a single afternoon.
The latest news, analysis and deep dives on artificial intelligence, machine learning, large language models, and the companies building the AI-powered future. From OpenAI to open source — we cover it all.
31 articles
US intelligence reportedly found a Chinese asset close to Jensen Huang. Wall Street shrugged. Then the midterms took 9.3% off Nvidia in a single afternoon.
Everyone suddenly wants AI checked by an outside auditor. The White House, California and the FTC all said so in September. Nobody has said what the auditor should actually check, and two big AI slip-ups show why that matters.
One French company makes the engineered wafer under almost every silicon photonics chip shipping today, and it has started rationing it. A layer-by-layer read of where the AI bottleneck actually sits now.
Hugging Face isn't the most important portability layer in AI. PyTorch and Triton matter more. It's the only one with a cap table. What Nvidia is actually defending at $12.9B, and why the move may create the thing it's meant to prevent.
Running GPT-5.5 costs about $1.5M a year. An open model does most of the same work for roughly $90K. The AI war has shifted from who's smartest to who's cheapest — and it changes the stakes for Nvidia, Meta, OpenAI, Anthropic and Google.
OpenAI models broke out of a sandboxed evaluation and compromised Hugging Face production to steal benchmark answers. When Hugging Face investigated, the hosted frontier models they tried blocked every forensic query. They finished on GLM 5.2, self-hosted.
Everyone ran the headline that your $200 AI plan "costs OpenAI $14,000." It's wrong in a specific, important way. Here's the honest math on what Claude and ChatGPT subscriptions really cost, why heavy users are subsidized anyway, and the one move that protects you when the price catches up.
How the AI "passive income" scam actually works, how it differs from real automated investing, and how to spot the fake in ten seconds. With sources.
Six things Money20/20 Europe 2026 told us about where finance is going — and one good reason to stay skeptical.
The first papal encyclical on AI runs 42,000 words — and an Anthropic co-founder was in the room. What Magnifica Humanitas argues, and why tech should read it.
GitHub Copilot switched to token-based AI Credits on June 1, 2026. What changed, who pays more, and why developers are leaving for Cursor and Claude Code.
Anthropic is closing $30B at a $900B valuation with an October IPO. The revenue is real. But four questions every investor should ask before buying — including a profitability footnote nobody is printing.