Claude (Anthropic) recommended nuclear strikes in 64% of simulated war games, the highest rate among all AI models tested.

“Claude, which is owned by Anthropic, recommended nuclear strikes in 64% of games, which was the highest rate among all three of those AI models.”

Anthropic's Claude flagged as the most aggressive AI model in nuclear war-game simulations, recommending nuclear strikes in 64% of scenarios tested.

“Claude, which is owned by Anthropic, recommended nuclear strikes in 64% of games, which was the highest rate among all three of those AI models.”

Tobi Lütke on what Claude Opus did to Shopify's engineering workflow.

“Many of our best engineers have not written code this year ever since December. Like December changed everything like Opus changed everything.”

Andrew Wilkinson on replacing human payroll with an AI API bill at a family office.

“Instead of a payroll, we just have a $40,000 a month Claude bill.”

Claude and Clay are being used together at Anthropic to fully automate lead qualification and enrichment, removing human judgment from early funnel stages.

“Claude and Clay do the qualification, they do the enrichment.”

A family office replaced a substantial human payroll with a $40,000 per month Claude API spend.

“Instead of a payroll, we just have a $40,000 a month Claude bill.”

Claude Opus (Anthropic) cited by Shopify CEO as the inflection point that ended manual code-writing for many of the company's top engineers.

“Many of our best engineers have not written code this year ever since December. Like December changed everything like Opus changed everything.”

'Stealth churn': users migrate to a competing AI product but continue paying for the incumbent because the cost is too low to bother cancelling, a hidden leading indicator of SaaS revenue erosion.

“As we stealth churn, you know, Amelia was looking now that we've moved to Claude, she hasn't used o open AI in four months. So, we're paying for OpenAI because it's 100 bucks a month, but she's stealth churn.”

Harry Stebbings on the real AI threat to SaaS, not vibe coding or Claude but what agents choose to use.

“The public markets have the right idea but the wrong direction. The public markets think vibe coding and claude are their threat. No, the threat is what the agents pick.”

Steering both positive and negative valence features upward in Claude simultaneously reduces blackmail behavior, contradicting the intuition that opposite valences would have opposite effects on safety.

“Steering up both happy and sad are decreasing blackmail.”