Anthropic engineer warns against 'cognitive surrender' from too much AI

An Anthropic engineer has written about the thinking linked to AI coding We risk 'cognitive surrender' if we don't keep engaging with AI, he says However, the developer isn't arguing against the use of AI coding agents The idea that our cognitive faculties might degrade through underuse in the age o

<![CDATA[ <article> <ul><li><strong>An Anthropic engineer has written about the thinking linked to AI coding</strong></li><li><strong>We risk 'cognitive surrender' if we don't keep engaging with AI, he says</strong></li><li><strong>However, the developer isn't arguing against the use of AI coding agents</strong></li></ul><p>The idea that our cognitive faculties might degrade through underuse in the age of AI isn't new, but it's been well summed up by Anthropic engineer Addy Osmani, who says that we risk losing key skills by letting bots do all our digital work for us.</p><p>"Nothing in my day-to-day forces critical thinking on me now that so much of my work runs through agents," <a href="https://x.com/addyosmani/status/2117512471930167377" target="_blank">writes Osmani</a> (via <a href="https://www.businessinsider.com/anthropic-engineer-addy-osmani-cognitive-surrender-warning-2026-11" target="_blank">Business Insider</a>). He adds that "skills that you stop exercising are going to go away, and probably faster than you might guess".</p><p>As you might expect given his job, the software developer isn't suggesting that AI agents shouldn't be used in coding projects — but he is saying that it's important to protect our critical thinking skills, despite the temptation to simply accept everything AI offers up.</p><p>Osmani also offers a counter to the idea that it doesn't matter if we lose the skills required for jobs that AI can do better anyway: "Regardless of how much AI helped, a human has to take responsibility," he says.</p><h2 id="maintaining-human-ownership">Maintaining human ownership</h2><div class="see-more see-more--clipped"><figure><blockquote class="twitter-tweet hawk-ignore" data-lang="en" cite="https://twitter.com/cantworkitout/status/2117512471930167377"><p lang="en" dir="ltr">https://t.co/2gyp2Erpxx<a href="https://twitter.com/cantworkitout/status/2117512471930167377">November 3, 2026</a></p></blockquote></figure><div class="see-more__filter"></div></div><p>Some level of cognitive offloading to AI is okay, Osmani argues, as long as human beings are still "maintaining ownership". It's when AI's output is accepted without checks and without any rigorous examination that we should start to be worried.</p><p>"You read the last answer well, so you skim the next one," explains Osmani. "You just scan the skeleton of the one after that. Then you get asked about a line and can't explain it, even with your agent's help... that's how cognitive debt accumulates."</p><p>The engineer has some suggestions about avoiding cognitive surrender. They include practising core skills, formulating ideas before asking AI, paying attention to what's generated, and making sure you understand the decisions AI has made.</p><p>Ultimately, while some skills will be lost as AI improves, we need to be deliberate in terms of those we're willing to let go, Osmani says — and not just let it happen without thinking. It's well worth considering the next time <a href="https://www.techradar.com/ai-platforms-assistants/gemini/gemini-can-redesign-entire-presentations-with-one-prompt-so-i-gave-it-my-sons-fun-art-lesson-and-it-turned-it-into-a-corporate-monstrosity">you ask for AI's help</a> with something.</p> </article> ]]>