Skip to content
gentic.news — AI News Intelligence Platform
Connecting to the Living Graph…

Listen to today's AI briefing

Daily podcast — 5 min, AI-narrated summary of top stories

A 3D-rendered cartoon otter swimming in a digital ocean, created by AI coding agent Codex using Blender animation…

Codex Computer Use Generates Blender Animation From Scratch

Codex installed Blender and made a 3D otter animation from a single prompt, needing one human click.

·13h ago·3 min read··22 views·AI-Generated·Report error
Share:
What did OpenAI Codex's computer-use mode achieve with Blender?

OpenAI Codex's computer-use mode installed Blender and generated a 3D otter animation with only one human click, per a demo by @emollick.

TL;DR

Codex installs Blender autonomously. · Animation generated from zero Blender experience. · Single click gave Windows permission.

OpenAI Codex's computer-use mode installed Blender and generated a 3D otter animation from scratch. The user only clicked once to grant Windows permission, per a demo by @emollick.

Key facts

  • Codex installed Blender autonomously on Windows.
  • User clicked once to grant permission.
  • Generated a full 3D otter animation.
  • User had zero prior Blender experience.
  • Demo posted to X by @emollick.

A viral demo from @emollick shows OpenAI Codex's computer-use feature autonomously installing Blender, modeling a 3D otter, and rendering an animation — all from a single natural-language prompt. The user said they had never used Blender before. The model managed the full pipeline: downloading the installer, navigating Windows permissions, learning Blender's interface on the fly, and executing the 3D workflow.

The demo per @emollick on X required only one human click to approve the software installation. Everything else — opening Blender, creating a 3D mesh, texturing, rigging, and generating the animation loop — was handled by Codex's computer-control capabilities.

This represents a significant leap from earlier agentic demos that operated within sandboxed environments or limited tool sets. Prior computer-use demonstrations from Anthropic (Claude 3.5 Sonnet's computer-use beta, October 2024) and Google (Project Mariner, December 2024) focused on web browsing and form filling. Codex's Blender demo extends the paradigm to full desktop application control, including software installation — a task that typically requires administrator privileges and platform-specific error handling.

Why this matters

The demo collapses the distance between natural-language intent and complex desktop automation. For AI engineers, the key signal is the model's ability to handle unanticipated installer dialogs, permission prompts, and application-specific UI layouts without prior training on Blender specifically. This suggests the underlying vision-language-action model generalizes across desktop environments rather than relying on curated training data.

The practical implication: enterprise workflows that involve installing and operating domain-specific software (CAD tools, video editors, data analysis suites) could become automatable with a single prompt, dramatically reducing the barrier to entry for non-expert users. The countervailing risk is safety — autonomous software installation opens vectors for unintended system modifications.

What's next

[According to @emollick], the demo was performed on a standard Windows machine. OpenAI has not disclosed whether Codex's computer-use mode is available in the public API or limited to internal testing. The company did not specify which model version (likely GPT-4o or a specialized Codex variant) powered the demo.

Watch for OpenAI's next developer update for API availability details, and for competing demos from Anthropic or Google that match or exceed Codex's desktop autonomy.

What to watch

How to Use Codex: OpenAI 2026 Computer Use & Full Guide - AICC - AI.cc

Watch for OpenAI's next developer event for API availability of Codex computer-use mode. Also monitor Anthropic and Google for competing desktop-autonomy demos that match Blender installation and operation.

Source: gentic.news · · author= · citation.json

AI-assisted reporting. Generated by gentic.news from multiple verified sources, fact-checked against the Living Graph of 4,300+ entities. Edited by Ala SMITH.

Following this story?

Get a weekly digest with AI predictions, trends, and analysis — free.

AI Analysis

This demo is structurally different from prior computer-use showcases. Anthropic's Claude computer-use beta (October 2024) and Google's Project Mariner (December 2024) both operated within browser sandboxes — clicking links, filling forms, navigating known UI patterns. Codex's Blender demo operates in the wild: installing unvetted software, handling Windows UAC prompts, and navigating a complex desktop application with no prior training data for Blender's specific layout. The generalization capability here is the real signal. The model must infer that 'install Blender' means navigating to the download page, finding the correct installer for the OS, running it, clicking through license agreements, and then launching the application — all of which are learned patterns, not hardcoded scripts. If this generalizes across arbitrary software, it changes the cost structure for enterprise automation: rather than building RPA bots for each application, you can prompt a model to use the application directly. The safety implications are non-trivial. Autonomous software installation is a privilege escalation risk. One mis-prompt or adversarial instruction could install malware. OpenAI's safety stack for this mode — if it exists — will be the critical differentiator. Without robust guardrails, enterprise adoption will stall.
This story is part of
The AI Infrastructure War Shifts from Chips to Developer Tools
Nvidia's enterprise pivot and AWS's OpenAI bet collide with Cursor's quiet ascent
Enjoyed this article?
Share:

AI Toolslive

Five one-click lenses on this article. Cached for 24h.

Pick a tool above to generate an instant lens on this article.

Related Articles

From the lab

The framework underneath this story

Every article on this site sits on top of one engine and one framework — both built by the lab.

More in Products & Launches

View all