Gary Marcus Urges Industry to Question the Engineering Behind AI Benchmarks

GaryMarcus · x · 2026-08-02

Reacting to recent impressive AI test results, Gary Marcus echoed an industry insider's deep skepticism: rather than just focusing on benchmark scores, there should be greater attention on the specific engineering processes achieving these breakthroughs. For instance, is it purely due to base model capability, or refined testing harnesses and post-training methods? The industry urgently needs more transparency regarding evaluation scaffolding and post-training details.

Related event: OpenAI Math Breakthrough Questioned by Gary Marcus and Others(38 posts)→

Original post →

More from Models

Models channel →