What People Actually Built With GPT-6 Astra
Vibe Coding

What People Actually Built With GPT-6 Astra

Within a week of launch, testers had Astra driving Blender, Unreal Engine, Ableton and Final Cut Pro. The builds are the clearest evidence yet of what a computer-use agent can hold in its head.

By Marcus Chia 2026-09-18 8 min read
Developer workstation with code on screen — building with GPT-6 Astra computer use agents

I have spent two years telling designers and non-developers that AI can build them a working app. That claim is now boring and mostly true. What changed with GPT-6 Astra is different: the model does not just write the code, it drives the software. Blender. Unreal Engine. Ableton Live. Final Cut Pro. A browser, for hours, without losing the plot.

The early-access builds circulating since 3 September are worth studying not because they are impressive demos — plenty of demos are impressive — but because of what they reveal about where the capability ceiling now sits, and where the bill lands.

1. The builds worth paying attention to

A partial list from the first week, filtered down to the ones that demonstrate something rather than just looking good:

  • Pokemon FireRed, finished in 18 hours. Astra played the game to completion using screenshots for state tracking. That is sustained navigation, decision-making and error recovery across a session far longer than any agent benchmark measures.
  • A GTA-style driving game in roughly 90 minutes. Explorable city, streets, vehicles, pedestrians, mission interface. Not shippable. Genuinely playable.
  • A seaside villa reconstructed in Blender from listing photographs, including structural detail, furniture and ambient objects — then turned into a promotional video. The model inferred geometry that was never in the reference images.
  • A 3D iPod interface built by Pietro Schirano for browsing Codex threads, running as a Mac interface.
  • A ship in a bottle in Three.js — a 17th-century sailing ship moving with the waves, circling seagulls, a miniature harbour and a coral reef, with physics.
  • An Unreal Engine survival simulation populated with autonomous agents that had to cooperate to stay alive, with conversation emerging from survival incentives rather than scripted dialogue.
  • A full music arrangement inside Ableton — synthesiser design, multiple instrument parts, complete arrangement — produced by operating the DAW, not by generating audio.
  • A five-minute teaching video on T cell development, script through animated visuals through narration.
  • A Manhattan cityscape built progressively over a week of sessions, which is the interesting bit: the work persisted across many context windows.

OpenAI has published its own guide to building games with Astra, which tells you where the company thinks the demand is.

2. Why the long ones matter more than the fast ones

The 90-minute driving game is the one that gets shared. The 18-hour Pokemon run and the week-long Manhattan build are the ones that matter.

Agents have historically fallen apart on long tasks for a structural reason: when the context window fills, the session gets compacted into a summary, and the summary loses the specifics. Requirement three from hour one quietly disappears. Astra changes the mechanism — instead of compacting prior work into lossy summaries, it maintains searchable notes it can query for earlier requirements and test results.

That sounds like an implementation detail. In practice it is the difference between an agent that can work a ticket for twenty minutes and one that can work a project for a week. It is also why the Codex changes shipped alongside Astra matter more to working teams than any benchmark score.

3. What this actually costs

Now the uncomfortable part. One complex early-access build consumed roughly 8 million tokens — about US$17 for a single task. Astra bills US$10 per million input and US$50 per million output, and agentic work is output-heavy by nature.

Run the arithmetic for a Malaysian product team. Five engineers, three long agent sessions each per working day, US$12–20 per session. That is roughly RM 3,500–6,000 a month in tokens alone, before subscriptions and before anyone measures whether the output shipped.

Three things keep it defensible. Cache aggressively — cached input drops from US$10 to US$1 per million, and an agent re-reading the same codebase is the ideal caching case. Cap session length deliberately rather than letting an agent run until it stops. And treat max reasoning effort as a decision, not a default.

4. Where Astra still fails

Four failure modes to plan around

Four failure modes to plan around 1Cost blows up silently
Agentic sessions are output-heavy. One complex build hit ~8M tokens (~US$17). Nothing warns you mid-run.
2Polish is not production
A playable prototype in 90 minutes is not a shippable game. Asset quality, performance and edge cases still need humans.
3Verification is on you
The model reports success confidently. On long runs it will declare a task done that a human reviewer would reject.
4Tooling drift
Driving desktop software through screenshots breaks when UI changes. Pin versions and expect re-runs after app updates.

Four failure modes I have hit or seen reported consistently:

5. What this means if you are not a developer

The people getting the most from Astra right now are not the strongest programmers. They are the ones who can specify precisely — describe a scene, a game mechanic, a spreadsheet transformation, a video edit — and then evaluate whether the result is right. That is a design skill and a domain skill, not a coding skill.

If you are a marketer, an architect, a video editor, a finance analyst: the software you already know how to drive is now software an agent can drive with you. The bottleneck moved from "can you build it" to "can you say exactly what you want and tell when it is wrong." That is the whole premise of our Vibe Coding programme, and this release makes it more true, not less.

For teams already shipping with agents, the practical next step is agent orchestration rather than raw prompting — which we cover in the MCP explainer and in our comparison of AI coding agents.

💡
HRDC SBL-KHAS claimable

AITraining2U's Vibe Coding and AI Engineering programmes teach agent workflows, computer-use patterns and cost control on frontier models — HRD Corp-registered and claimable under SBL-KHAS. See how to claim.

Eighteen months ago I would have called an agent finishing Pokemon a party trick. It still is. But the same mechanism that let it hold a game state for 18 hours is the one that lets it hold a client project for a week, and that is not a party trick at all.

About the author

Marcus Chia →

12+ yrs Product Design · Vibe Coding Specialist · ASEAN-scale Products

Marcus has 12+ years in product design and front-end engineering, having shipped consumer and SaaS products used by millions across ASEAN. He specialises in vibe-coding workflows that turn Figma concepts into deployable apps using Claude Code, Antigravity, and Cursor — and teaches non-developers to ship polished, user-centric interfaces in days rather than sprints.

Frequently Asked Questions

Early-access builds include a full Pokemon FireRed playthrough completed in 18 hours using screenshots for state tracking, a GTA-style driving game built in about 90 minutes, a seaside villa reconstructed in Blender from listing photographs, a 3D iPod interface for browsing Codex threads, an animated ship-in-a-bottle scene in Three.js with physics, an Unreal Engine survival simulation with autonomous cooperating agents, a complete music arrangement produced inside Ableton Live, and a Manhattan cityscape built progressively across a week of sessions.

Yes — that is the headline capability of this release. Astra supports a computer-use tool alongside hosted shell, apply patch, code interpreter and MCP, and testers have driven Blender, Unreal Engine, Ableton Live and Final Cut Pro with it. On OSWorld 2.0, a computer-and-browser-use benchmark, Astra scored 72.6% while taking roughly 47% less time per task than GPT-5.6 Sol. Reliability is still well short of a competent human operator.

One complex early-access build reported roughly 8 million tokens consumed — about US$17 for a single task at Astra's US$10 input / US$50 output per million pricing. Agentic work is output-heavy, so costs scale faster than they do for chat. A five-person team running three long sessions daily can reach RM 3,500–6,000 monthly in tokens alone. Prompt caching, which cuts input to US$1 per million, is the main mitigation.

Context preservation. Previously, when an agent session filled its context window, prior work was compacted into a lossy summary and specific requirements got dropped. Astra instead maintains searchable notes it can query for earlier requirements and test results. Practically, this is what allows multi-hour and multi-session work — an agent that can hold a project across a week rather than a ticket across twenty minutes.

No, and the early evidence suggests the strongest results come from people who can specify precisely and evaluate results critically, rather than from the strongest programmers. If you already know how to drive Blender, Excel, Premiere or a CAD tool, Astra can drive it alongside you. The skills that matter are clear specification and the judgement to tell when the output is wrong — both teachable, and both the focus of vibe-coding training rather than traditional programming courses.

Want to apply this in your organisation?

AITraining2U runs HRDC-claimable corporate AI training for Malaysian organisations — from leadership awareness to hands-on builder workshops. Talk to us about a programme tailored to your team.