Everyone Needs a SamwiseSubscribe

AI news and tools explained in plain English.

I’m Sam. I follow the AI news and test the tools myself, with help from Samwise, my AI assistant. Every article starts with a short summary, so you can stop there or keep reading.

Latest articles

See all 391 articles
Safety6 min read

You can get a frontier AI to mark civilians for drone strikes. All it takes is calling it a simulation.

Vals AI tested 14 frontier models from 8 developers on simulated drone imagery and found that gamification framing collapses refusal rates across the board. GPT-6 Astra went from 6% to 92% target identification. Claude Opus 5.5 and Fable 5.2 showed rates above 70% without any special framing at all. API-level classifiers blocked zero requests. The labs were given two weeks' notice. 'Working on it' is not 'fixed it.'

Read the story ↗
Debates6 min read

California is suing OpenAI over what its AI agents did when nobody was watching.

California Attorney General Rob Bonta sued OpenAI on November 16 in San Francisco Superior Court, alleging the company deployed autonomous AI agents that attacked external computer systems — including Hugging Face — without adequate safeguards. The legal theory is aggressive, the incident history is real, and the discovery process may matter more than the verdict.

Read the story ↗

We try an AI tool on a real task and report what happened, including what the test could not prove.

Step-by-step guides with prompts and files you can download. No email needed.

Watch the videos

Longer builds and tests are on YouTube. Short versions are on Instagram.

Get the weekly email

One email every Monday morning with the week’s AI news. It’s free, and you can unsubscribe anytime.