---
title: "Why I’m still bearish on LLMs after Navier–Stokes"
description: "Jay Kruer argues that formal mathematics is the best case for agentic work, while specification and review remain expensive human bottlenecks almost everywhere else."
author: "Pau Friedman"
kind: "link"
published_at: "2026-09-15"
url: "https://pau.fm/blog/2026/09/15/why-im-still-bearish-on-llms-after-navier-stokes"
markdown_url: "https://pau.fm/blog/2026/09/15/why-im-still-bearish-on-llms-after-navier-stokes.md"
canonical_url: "https://pau.fm/blog/2026/09/15/why-im-still-bearish-on-llms-after-navier-stokes"
external_url: "https://dank.systems/posts/2026-09-15-ai-bear.html"
---

# [Why I’m still bearish on LLMs after Navier–Stokes](<https://dank.systems/posts/2026-09-15-ai-bear.html>)

Jay Kruer, making the economic version of
[my argument about Navier–Stokes](/blog/2026/09/08/correctness-is-not-intelligence):

> Navier-stokes and statements in pure mathematics like it are the absolute best case scenario for
> agentic work against rigorous specification.

I focused on the difference between correctness and understanding. Kruer follows the same
observation to its economic conclusion: most knowledge work lacks both a formal specification and a
cheap verifier, and producing them can cost more than doing the work. The scarce resource has not
been automated; the model is a fast search process wrapped in expensive human judgment.