It’s been several months since I posted about Finkelstein’s Immaturity Model and I think it’s worth revisiting some of the generalizations I made at the time regarding the model’s take on “autoprogramming”."

What is true as of Spring 2026:

• Bad product decisions lead to bad products
• LLMs optimize local maxima
• LLMs revert to the mean

What has changed since August 2025:

• The bad products are especially beautiful and well-designed
• I spend more time scrutinizing apps and systems and WAY less time scrutinizing code

I probably didn’t to scrutinize code in mid-2025, either, but to do so felt genuinely negligent, given the volume of very public bad outcomes being reported at the time.