AI NewsProductivityReported

Jan Schauma essay pushes back on Gates calling Claude superhuman at code

An essay by Jan Schauma says rich tech founders judge every skill from their own seat. They call Claude superhuman from there.

AI News

Editorial2 min read

LinkedInX
Jan Schauma essay pushes back on Gates calling Claude superhuman at code

Why it mattersA reader who buys a coding agent on a founder's word is buying their self-measurement. The essay gives a plain reason to ask who is doing the measuring.

Bill Gates told the New York Times that Claude coding models "are as good as I am". He added that they are "superhuman at writing code and in finding flaws in code". A new essay on netmeister.org asks a simpler question. Who is the person writing the yardstick?

The essay is called "The Claude Delusion". It was posted on 13 November by Jan Schauma. He is an infrastructure security architect who teaches graduate system administration and Unix programming at Stevens Institute of Technology. The post sat on the Hacker News front page for a day. It reached 152 points and 200 comments.

What the essay argues

Schauma's point is that a person who gets rich from one skill ends up grading every skill from that seat. A founder sees a model match their own code. They conclude the model is superhuman, because they already think they are. The essay walks through the shape of the move. First, success convinces the founder they deserve it. Then that story spreads to every other domain they touch. Then an AI that writes code at their level becomes, in their own words, as good as they are.

The essay names other examples of the same reflex. Elon Musk predicting a submarine would reach trapped cave divers, and his covid forecasts. Mark Zuckerberg considering a run for president. Elizabeth Holmes and Sam Bankman-Fried. Actors who ran for office on the strength of being recognised. Schauma treats the Gates quote as the same reflex, not a worse one.

The one specific number he leans on

Schauma cites a Bloomberg News piece from 28 October. It says the United States data-centre build-out has erased fifteen years of greenhouse-gas reductions. He uses the figure to say the Claude Delusion has a cost outside the founder's own life. The rest of the essay runs on argument rather than data.

The counter-argument on Hacker News

The top-rated reply on Hacker News is from a user named ianm218. They say Gates is a natural judge of a coding model. A long-time programmer who evaluated software for decades can tell when an AI matches working engineers. The reply compares it to seeing AI get good at SQL for people who write SQL every day.

That reply matches what many working developers have said this year. Schauma's answer is that spotting the bar being crossed is not the same as judging what is above it. A founder who feels superhuman will read a model that matches them as superhuman too.

The essay does not argue that Claude is bad at code, and the reader should not read it that way. It argues that a buyer who picks an AI coding tool from a founder's quote is buying that founder's self-image. The useful work is on the buyer's own tasks, with the buyer's own measure.

Source

The Claude Delusion by Jan Schauma. Hacker News discussion. The Gates interview Schauma cites: Ezra Klein podcast with Bill Gates in the New York Times.

Reported bynetmeister.org

This item was written by an AI system from the linked source. Reveneau is responsible for what it publishes.

Share
LinkedInX
Start a project