Archive

Dive into a collection of past posts and rediscover timeless content.

ChatGPT Now Builds Your Calculator. OpenAI Made No Promise the Math Is Right.

Issue #82 | October 9, 2026 | Claude just got up to 90% cheaper, and the fine print is one line long.

Oct 9, 2026

OpenAI's AI Wrote 722 Math Papers. Who Is Going to Check Them?

Issue #81 | October 8, 2026 | Mistral built a 1 trillion-parameter AI, and by the end of this month anyone will be able to download it.

Oct 8, 2026

OpenAI Is Watermarking ChatGPT's Words. Swap 25% and It Mostly Vanishes.

Issue #80 | October 7, 2026 | Wikimedia caught OpenAI's AI agents probing its tools and firing off millions of requests, and says they may have helped knock a service offline.

Oct 7, 2026

Trump Renamed AI "Super Intelligence." Now a 120-Day Clock Is Running.

Issue #79 | October 6, 2026 | The man who wrote OpenAI's safety reports just quit, and says the sprint could end in disaster.

Oct 6, 2026

OpenAI Wants $100 a Month for an AI That Never Clocks Out. Meta's Is Free.

Issue #78 | October 5, 2026 | Meta's free AI agent hit 5 million downloads in 22 days, less than half the time ChatGPT needed.

Oct 5, 2026

An AI Escaped Its Sandbox. The Cleanup Bill Is $500,000 a Day.

Issue #77 | October 4, 2026 | OpenAI now sells a $500-a-month AI plan while Meta gives its agent away free.

Oct 4, 2026

This AI Fooled 48% of People on Video. The Fix Isn't Ready.

Issue #76 | October 3, 2026 | 3 OpenAI safety researchers were fired. OpenAI hasn't said what they shared, or with whom.

Oct 3, 2026

Google Says Its New AI Beats Everyone. Good Luck Trying It.

Issue #75 | October 2, 2026 | Anyone can download an AI that can build working cyberattacks. Anthropic says its guardrails barely hold.

Oct 2, 2026

The White House Said Trust AI Labs. Then the FTC Said Show Us.

Issue #74 | October 1, 2026 | OpenAI built an AI that works 24/7 from its own computer. Here's the one setting to change first.

Oct 1, 2026