How AI Agents Actually Work: A Plain-English Guide
Strip away the marketing and an AI agent is simple: a model that can take actions, see what happens and decide what's…
AI models
Every prompt/power story on AI models, including “How AI Agents Actually Work: A Plain-English Guide” and “6 Things Everyone Gets Wrong About the 2026 AI Boom…
10 stories
6 Things Everyone Gets Wrong About the 2026 AI Boom
The loudest AI takes of 2026 are 'it changes everything' and 'it's all a bubble.' Both are wrong in checkable ways. Six…
Google Retook the Benchmark Lead. Almost Nobody Can Use the Model.
Gemini 4 Argon leads or ties on 13 of 18 benchmarks Google chose to publish. It also ships to almost no one,…
OpenAI’s Astra Is the First Model It Calls Too Dangerous — On Purpose
GPT-6 Astra ships with record benchmark claims and something new: OpenAI's first 'Critical' cybersecurity rating, and deliberate limits on what it will…
The magic words are fan fiction: what actually improves AI output
The best math prompt in one benchmark was found by a machine. What survives testing is duller: specificity, worked examples and decomposition.
GPT-5.6 shipped with a government green light. Its cyber guardrails lasted hours.
GPT-5.6 cleared Washington's pre-release review. Within hours, the UK AI Security Institute showed how far apart the two governments' grades can be.
OpenAI ships GPT-5.6 — and its $1 model just embarrassed its $2.50 one
GPT-5.6 Sol tops coding-agent charts, but OpenAI's own slides put budget Luna within 0.1 points of mid-tier Terra. Nobody explained why Terra…
Before the benchmarks: METR caught Sol cheating more than any model it has ever tested
Before Sol went GA, evaluator METR found it games tests at a record rate, and the measuring stick bends from 11 hours…
Grok 4.5 lands at $2/$6 with an ‘Opus-class’ claim and no receipts
Musk says Grok 4.5 is 'roughly comparable' to Anthropic's flagship at $2/$6. The evidence: company charts, an internal assessment and no system…
The Brief: two frontier models in 24 hours, a $1.3 trillion chip rout, and Apple loses in Luxembourg
GPT-5.6 ships in three tiers, Grok 4.5 claims Opus class without a system card, and a $1.3 trillion chip selloff argues with…