GPT-6 Astra Demo: One Agent Juggling Many Computer Tasks

Melvin Vivas · X video post · 2026-09-04 · 2:42 · 478 views · Open on X

Topics: Industry Trends & Job Market, AI Agents, Tool Use & MCP · Level: beginner

Summary

Melvin Vivas reacts to the launch video for GPT-6 Astra. The quoted post calls it the company's most capable model for work, coding, collaboration, scientific discovery and cybersecurity. The demo shows one user giving voice commands to an agent that works on several tasks at the same time on a computer: design, 3D modeling in Blender, slide decks, an eBay listing, a game, food ordering, a legal template edit, booking a tennis court and exporting a 3D-printable file. The post has no tutorial in it. Its value is as a look at where computer-use agents are heading.

Key points

• GPT-6 Astra is presented as a general 'do almost anything on a computer' model aimed at work, coding, collaboration, science and cybersecurity.
• In the demo, one user runs many tasks at once by voice, and the agent switches between them and reports progress (e.g. 'Checking now', 'Looking into ordering food').
• Creative work happens in steps: a yellow circle becomes a detailed rocket window, then a 3D model in Blender, then an STL file for a 3D printer.
• Feedback is iterative: the user asks for more detail, background colors that match the rain jackets, and a note that the table is slightly damaged.
• The agent works across apps and local files, for example attaching a photo from the Downloads folder to an eBay listing.
• Professional work is shown too: drafting a licensing agreement template and rewriting the limitation of liability clause to favor the licensor, with lawyers still reviewing it.
• Real-world actions such as ordering food and booking a court are carried out with user confirmation ('Yeah, book it').

Resources mentioned

• GPT-6 Astra · tool · openai.com · paid
  The model announced in the quoted launch post, pitched as the developer's most capable model for work, coding, science and cybersecurity, and able to operate a computer.
  Also in: Customizing Your Coding Setup with Pi Coding Agent Extensions (Melvin Vivas on X · notes), Pi Agent Council: Ask Multiple LLMs in Parallel and Compare Their Advice (Melvin Vivas on X · notes), Use GPT-6.1 Sol by Default, Save Astra for Emergencies (Melvin Vivas on X · notes), Dots in ChatGPT: always-on AI agents that you hand responsibilities to (Melvin Vivas on X · notes) and 49 more
• Blender · tool · blender.org · free
  Free, open-source 3D creation software. In the demo the agent opens it to turn a drawing into a 3D model.
  Also in: Drive Blender from a Coding Agent via Headless Python Scripts (Melvin Vivas on X · notes), Building a 3D Flight Scene with Sonnet in Cursor, three.js and 3D Models (Melvin Vivas on X · notes), ComfyUI Raises $30M: Open-Source Visual AI Pipelines Move into Production (Melvin Vivas on X · notes)

Try this

• Build a 3D browser game where the player dodges asteroids with the arrow keys and uses space to boost (prompted in the demo).
• Build a design-to-fabrication pipeline: sketch, detailed render, Blender 3D model, STL export for a 3D printer.