> ## Content Index
> Fetch the complete content index at: https://wire.fourthweb.ai/llms.txt
> Use this file to discover other available public pages before exploring further.

# Anthropic's Own Engineer Warns AI Is Making Him Dumber at His Job
- URL: https://wire.fourthweb.ai/anthropics-own-engineer-warns-ai-is-making-him-dumber-at-his-job/
- Published: 2026-11-05T10:00:03.000Z
- Updated: 2026-11-05T10:31:37.000Z
- Description: The people building the tools are the first to see what those tools are taking from them. Addy Osmani, engineer on Anthropic's Claude Code team, warns that AI agents are eliminating critical thinking from his daily work — he's building the thing that's hollowing out his own cognition
- Author: Travis Wright
- Tags: AI Agent Economy, AI Agents, AI Infrastructure, OpenAI, Anthropic

**The people building the tools are the first to see what those tools are taking from them.**

### The Summary

- [Addy Osmani, engineer on Anthropic's Claude Code team, warns that AI agents are eliminating critical thinking from his daily work](https://www.businessinsider.com/anthropic-engineer-addy-osmani-cognitive-surrender-warning-2026-11?ref=wire.fourthweb.ai) — he's building the thing that's hollowing out his own cognition
- He identifies three stages: cognitive offloading (good), deferral (risky), and cognitive surrender (debt you'll pay later)
- The people shipping agentic coding tools are starting to notice they've become operators, not builders

### The Signal

Osmani isn't some skeptic lobbing grenades from outside. He works on Claude Code at [Anthropic](https://wire.fourthweb.ai/tag/anthropic/). He ships the agentic coding product. And he's watching it change how his own brain works.

[His Tuesday X essay](https://www.businessinsider.com/anthropic-engineer-addy-osmani-cognitive-surrender-warning-2026-11?ref=wire.fourthweb.ai) cuts through the "AI will make us all 10x engineers" hype with a simple observation: nothing in his workday forces him to think critically anymore. The agents handle it. He's not the first engineer to notice this shift — an [OpenAI](https://wire.fourthweb.ai/tag/openai/) alum recently told Business Insider that AI coding had become "tedious and boring" — but Osmani's position makes the warning land differently.

He breaks down three modes of AI-assisted thinking:

- **Cognitive offloading:** You delegate tasks but stay in control. This is the promise. "I do this all day. You should, too."
- **Deferral:** AI output becomes your input. You're still engaged, but the machine sets the agenda.
- **Cognitive surrender:** Complete trust in AI outputs. No questions asked. This is where the debt accumulates.

> "You have to build critical thinking into your day, because by default nothing will do it for you."

The researchers who coined "cognitive surrender" are from Wharton. Their work suggests this isn't just about coding — it's about how we interact with any sufficiently capable AI system. When the machine is good enough, the path of least resistance is trust. And trust, over time, becomes atrophy.

Osmani doesn't advocate abandoning agents. He advocates selectivity. Know when you're offloading and when you're surrendering. The former is leverage. The latter is cognitive debt — a bill that comes due when you need to solve a novel problem and realize you've forgotten how.

The irony: the people building Web4 are the first to experience its costs. Engineers at Anthropic, OpenAI, and Google are simultaneously shipping agentic tools and grappling with what those tools take from them. They're operators now, managing systems that write code, debug edge cases, and ship features. The creative problem-solving that drew many of them to programming in the first place? Increasingly optional.

This matters because what engineers experience first, everyone else experiences later. If the builders are noticing cognitive changes, those changes are coming for product managers, marketers, analysts, and anyone else whose work involves pattern recognition and decision-making under uncertainty.

### The Implication

Osmani's answer is simple but not easy: deliberately choose friction. Pick the problems you solve yourself. Ask follow-up questions when the agent hands you an answer. Treat critical thinking like a muscle that atrophies without use.

For teams building with [AI agents](https://wire.fourthweb.ai/tag/ai-agents/), this is a design challenge. How do you build systems that augment without replacing? That offload grunt work without eliminating the cognitive reps that keep people sharp? The companies that figure this out won't just have more productive teams — they'll have teams that can still think when the models hit their limits.

### Sources

[Business Insider Tech](https://www.businessinsider.com/anthropic-engineer-addy-osmani-cognitive-surrender-warning-2026-11?ref=wire.fourthweb.ai)