It’s the Lessons We Learned Along the Way. Or, Is It?
This title could be clearer and more informative.Try out Clickbait Shieldfor free (5 uses left this month).
A co-founder of Bauplan shares a personal experiment replacing a summer intern with ChatGPT and Claude Code for a research spike on querying branched data tables. The AI stack delivered tangible outputs — a working prototype and an ACM paper — in 48 hours versus weeks, but the author reflects on a deeper loss: no real understanding was gained. Unlike mentoring human interns, which sharpens thinking bidirectionally, AI-assisted work produced results without building mental models. Despite the experiment, Bauplan still hired two human interns for 2026, and the author wrestles with the grey area between 'humans think, LLMs fix syntax' and the risk of cognitive atrophy from outsourcing too much to AI.
