Leading Anthropic dev says 'a human has to take responsibility' with AI code projects, as some skills get lost

Live updates

1. 45m ago

   Anthropic developer urges human ownership as AI coding risks skill loss

   Anthropic software developer Addy Osmani says developers must retain responsibility for AI-assisted coding projects and keep exercising critical thinking as agents take over more work. His warning describes "cognitive surrender": a loss of engineering judgment when people stop engaging with the reasoning behind generated code. The discussion has circulated through TechRadar, Business Insider, Yahoo Tech and other outlets, while related analyses from Stanford and the World Economic Forum examine which mental work people should keep or delegate to AI. No concrete policy or product change has been announced.

   Why it matters

   Osmani is a Google Chrome team engineering manager known for web performance work, according to a Japanese freelance-job report. The concern is about skill erosion, not a claim that AI coding already produces defective software. Broader commentary on cognitive offloading gives the warning context by asking which tasks people should retain and which they can safely hand to AI.

   What is confirmed

   • Anthropic software developer Addy Osmani has said "a human has to take responsibility" with AI code projects.
   • Osmani warns that overreliance on AI can weaken engineers' critical thinking and create a risk of "cognitive surrender."
   • Osmani is a Google Chrome team engineering manager known for web performance work.
   • The World Economic Forum has published analysis on which thinking tasks people should keep and which they should delegate to AI.
   • Stanford Report has published an article asking what happens to learning when AI does the thinking for people.

   Still unconfirmed

   • A shattered.io summary presents three warning signs and advice for staying in control of AI-assisted work in 2026.
   • A Japanese report dates David Nield's TechRadar article to November 8, though the supplied material does not specify a year.

   What to watch next

   • Any Anthropic guidance or engineering standards that assign human review or ownership duties for AI-generated code.
   • Further reporting that identifies specific code-review failures, skill measurements or productivity data tied to AI-assisted development.
   • Public responses from Osmani or other Anthropic engineers detailing how they preserve critical thinking while using AI tools.
   Sources used for this update (11)

   1. TechRadar — Leading Anthropic dev says 'a human has to take responsibility' with AI code projects, as some skills get lost
   2. Business Insider — An Anthropic engineer warns about 'cognitive surrender' as AI agents take over his work
   3. Stanford Report — What happens to learning when AI does the thinking for us?
   4. World Economic Forum — AI Cognitive Offloading: Which Thinking to Keep vs Delegate
   5. Wukali — ARTIFICIAL INTELLIGENCE. THINK LESS—WE’VE GOT YOU COVERED
   6. Modern Ghana — Is AI Dulling Our Mi
   7. Yahoo Tech — An Anthropic Engineer Just Warned About the Hidden Cost of Letting AI Do Your Thinking
   8. tech.yahoo.com — Leading Anthropic dev says 'a human has to take ...
   9. shattered.io — Cognitive Surrender: Anthropic Engineer's 3 Warnings [2026]
   10. biztoc.com — Leading Anthropic dev says 'a human has to take ...
   11. ai.freelance-job.com — AIに考えさせ続けると、エンジニアの思考力は静かに消える — Googleエ...
   confidence 86%