ai morning #37 — gpt-5.6 sol rewrote its own infrastructure and cut openai's costs by 20%

ai morning by thehype · Jul 30, 2026 · 9m

A deployed model rewrote its own production infrastructure and cut costs by 20%. Not a benchmark. Not a demo. Production GPU kernels, rewritten in the wild, at scale — then a second self-optimization on speculative decoding on top of that.

Marcus walks through what happened, why Simon Willison called the savings "presumably billions of dollars a month," and what it means for builders who still haven't looked at harness engineering.

In this episode:

00:00 Intro

01:27 Sol rewrote its own GPU kernels in prod — GPT-5.6 Sol cut OpenAI's serving costs 20% and boosted token efficiency 15%+ by…