OpenAI
OpenAI4d ago
Tech

GPT-6 Astra with Peter Gostev

3 min video2 key momentsWatch original
TL;DR

Peter Gostev says GPT-6 Astra handles complex 3D visual tasks and legacy code refactoring with minimal oversight, marking a step up in reliability from GPT-5.6.

Key Insights

1

No debugging neededAstra refactored 150,000 lines of legacy code end-to-end without requiring post-generation debugging — a sharp jump from GPT-5.6, which needed manual alignment work.

2

Pushes back naturallyThe model pushes back on feedback instead of reflexively agreeing, feeling more like a human collaborator than a people-pleaser.

Want this for every new video OpenAI posts? Brevyd summarizes each upload automatically, the morning it drops.

Deep Dive

Testing at the ceiling

Gostev explains his testing philosophy: medium-difficulty tasks are solved by too many models now, so he builds towards the hardest possible tasks to see where the new model breaks. He landed on 3D visual work because it lets him track capability improvements visually. His benchmark: a voxel representation of historic London spanning medieval, Tudor, and modern eras, all transforming within the same map — a task that requires sustained spatial reasoning across context shifts.

Astra's communication shift

With GPT-5.6, Gostev noticed the model would instantly validate every piece of feedback, reflexively agreeing with criticism in an artificial way. Astra behaves differently — it either pushes back on suggestions or genuinely acknowledges mistakes, mimicking natural human discourse. He calls this shift subtle but important. It's not about technical capability; it's about interaction quality.

The reliability breakthrough

Gostev had been trying to refactor an accumulating codebase — roughly 150,000 lines built up over multiple GPT versions — with GPT-5.6 working but requiring substantial debugging afterward. With Astra, the refactored code ran without intervention. He emphasizes this reduction in hand-holding as the real leap: models are finally mature enough he doesn't need to babysit every step. The constraint now isn't the model; it's his hardware. He moved workloads to a Linux box because CPU became the bottleneck.

Takeaways

  • Use Astra for stateful, multi-step tasks where you'd normally expect rework — the reliability gain justifies the API cost.
  • Stop evaluating models on easy benchmarks; test them on your actual hardest problem to see where they still fail.

Key moments

0:44Historic London 3D benchmark

I've built here is a voxel 3D representation of historic London, which goes across the different eras, medieval London, Tudor London, and it transforms all within the same map.

2:05Zero debugging needed

With Astra, it pretty much just worked and I didn't need to do anything. So for me, that jump in reliability that I don't need to babysit it with every single step, that's a big deal.

You just read one. Brevyd does this for every upload.

Follow OpenAI and every new video comes back as a summary like this, in your morning briefing. No watching required.