The Real Cost of Autonomous AI Research: $100K in API Spend, 1 Billion Output Tokens, and Several Hundred Hours of Human Validation
The Hacker News·high signal
Buried in Anthropic's cryptanalysis writeup is the operating cost of the 7-round AES-128 result, where Claude worked 'almost entirely autonomously' after researchers built a scaffold -- generating several hundred million tokens over three days from just three substantive human messages, ultimately reaching one billion output tokens. The attack came out 200 to 800 times faster than prior work depending on measurement method. Human researchers then spent roughly $100,000 in API costs and several hundred hours validating the findings.