🤖Anthropic Says AI Now Speeds Up Its Own R&D, Employees 12x More Productive

Anthropic's AI is now building its own successor faster

Anthropic says AI is now speeding up its own research, and its models are improving roughly twice as fast as they did a year ago. The gain isn't just from throwing more compute at training. AI's share of the speedup keeps growing. If you build on Claude, this is the part to read closely. The company says employees using its most capable internal model were 12 times as productive as they would have been without AI help. That's a claim about a frontier lab's internal workflow, not your codebase, so don't plan your sprint around it. But it tells you where the labs are pointing their tooling next. The numbers come from METR, the outside evaluator Anthropic gave confidential research, performance data, and employee interviews. Anthropic also held the power to redact METR's final report, and it used that power extensively. Three newer models remain unreleased, finishing training in July, August-September, and September-October. Anthropic says releasing details could help rivals. Anthropic plans its own report on AI progress and new safety steps, and the METR publication has already drawn fierce criticism from safety researchers.