Figma Gave GPT-6 Astra a Moonshot. Here's what happened.

3 min video2 key momentsWatch original

⚡

TL;DR

Figma's design team used GPT-4o Astra to autonomously build a complete flight-control UI prototype in two hours, generating production-ready design systems and multi-screen journeys from a single prompt.

Key Insights

Two hours, fully autonomous — Astra spent two hours building a complete prototype, handling design systems, components, and multi-screen journeys without explicit step-by-step guidance — a level of autonomous reasoning Figma's team hadn't seen before.

Understood deep principles — The model understood deep design principles unprompted — it structured a brand system and enforced consistency across screens before being asked, suggesting reasoning beyond pattern matching.

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

Deep Dive

Astra's confidence and steering ability stood out

The Figma designer noted Astra felt different from prior models — it had what she called a level of confidence and its own opinion about how to approach tasks without requiring constant guidance. A key strength was its ability to pause and understand a new task thoroughly before executing, which led to better results downstream. The model also accepted mid-task direction changes smoothly, a capability she found especially valuable when prototyping.

Two-hour sprint produced a full design system

The team prompted Astra to design a flight-control system for a single-person helicopter, starting with hero flows and user journeys. Instead of incremental sketches, Astra generated complete prototype concepts and built them out over two hours. It handled design system setup, component consistency, and multi-screen experiences across different surfaces — all delivered with high fidelity. The designer called it unprecedented for a model to use that many Figma features correctly in one pass, from system setup through end-to-end flows.

Next: deep workflow integration at Figma

The designer plans to integrate Astra into Figma's own feature development pipeline. The goal is to compress the messy idea-exploration phase into rapid refinement and visual communication of concepts, then de-risk technical execution in parallel. She expects this workflow speedup to let the team prototype and ship features faster, with Astra handling both conceptual work and risk mitigation simultaneously.

Takeaways

✓Astra's ability to understand and enforce design systems autonomously — without being told — suggests models can now reason about structural principles, not just generate outputs.

Key moments

0:27Astra's confidence and autonomy

“There's a level of confidence that I could feel when using it that was a little bit different from anything that I had tried in the past. It felt like it had its own opinion about how it was going to go about doing this in a way that didn't really require a lot of guidance.”

1:15Model built production system in two hours

“It worked for like 2 hours and then built a new prototype. Getting a common brand system in place using components the right way to actually create consistency across multiple different screens.”

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.