The Claude Delusion
An essay on the netmeister.org blog argues that AI-era awe reflects overconfidence among successful people more than the real power of the tools. The author writes that late in 2025 Anthropic's Claude coding models became as capable as his own programming, a skill he says he built from age 13 to 24. The essay then targets tech leaders who believe their success makes them uniquely fit to decide humanity's future, who call AI a digital Prometheus, and who promise it will cure all diseases while still holding military contracts.
The essay uses Bill Gates as its main case study. According to a summary published by daily.dev, the piece notes that Gates described an OpenAI model passing an AP Biology exam for him, then praised Anthropic's Claude coding models as 'super good'. The same summary says the essay also argues that 'agentic' AI language lets people disown responsibility when AI systems misbehave, and it links power with weaker empathy, including on climate change.
Gates's own account, shared in Gates Notes, describes a 2022 challenge he set OpenAI: pass an AP Biology exam with questions outside its training data. Per his account, the model answered 59 of 60 multiple-choice questions correctly in September 2022, and an outside expert gave its written answers a 5, the top score. That model was GPT, not Claude, so the two products in the essay's example are different. This brief relies on excerpts from the essay and on secondary summaries, so some of its full argument may not appear here.
Why it matters
For founders and operators, the essay is a warning about how AI products get described. Its target is the habit of stretching a model's narrow strengths, such as writing code or answering exam questions, into broad claims about judgment, science and society. Teams that sell to executives should expect buyers to repeat that kind of language, and they may need evidence to back it up.
The agentic angle matters more directly. If vendors present autonomous agents as trustworthy by default, questions about who answers for an agent's mistakes shift toward the company that sold the tool and the customer that deployed it. Investors reviewing AI coding or agent startups should ask how each product handles permissions and destructive actions, and whether its marketing claims more autonomy than the product can safely deliver.
Key points
- 1Netmeister.org published the essay The Claude Delusion, reported on November 14, 2026.
- 2The author says late-2025 Claude coding models matched his own programming skill, which he built from age 13 to 24.
- 3Bill Gates is the essay's main case study, cited for praising Claude's coding abilities as 'super good'.
- 4The essay argues that awe at AI reflects overconfidence among successful people more than AI's actual power.
- 5Gates's 2022 AP Biology test involved OpenAI's GPT, which answered 59 of 60 multiple-choice questions correctly.
