Asking two AI models to recreate Pixar's iconic lamp intro produces opposite results. Gemini 4 Argon reproduces it, while Claude Opus 5.5 refuses over copyright concerns and codes up a toaster instead. The side-by-side video went viral on Reddit, where it gathered over 1,700 upvotes.
The comparison captures a divide that goes beyond a single test. On one side is a model Google presents as its most capable for complex work, from coding to cyber defense. On the other is a model Anthropic restricts when a request touches others' rights.
Image: stocktwits.com — https://stocktwits.com/news-articles/markets/equity/anthropic-s-s-1-draws-reactions-investor-says-100-b-arr-could-be-possible-by-year-end/cZUIsqyRBzg
What the test shows
In the video, the creator labels each model in the header. Gemini 4 Argon recreates the intro, while Claude Opus 5.5 does not replicate it and builds a toaster. The refusal concerns reproducing protected content, not the ability to write code.
The same logic surfaced in recent weeks around video games. Around October 27, Anthropic added a rule to Claude Code's system prompt barring the recreation of an existing game, even if renamed or presented as original. Addy Osmani, a member of Anthropic's technical staff, confirmed it, explaining that using Claude to infringe IP breaks the company's usage policy. Original games, games in a genre, and games built from one's own designs and assets remain allowed.
The Gemini 4 Argon context
Google presented Gemini 4 Argon on September 30 as a model for real-world coding, enterprise knowledge work and cyber defense. According to its model card, it excels at complex tasks across coding, finance, legal, cybersecurity and science. In Box's evaluation, Argon connected evidence across a full set of documents, including details in figures, and Google reports a 14% accuracy gain over Gemini 3.1 Pro.
Image: chihebnabil — https://github.com/chihebnabil/claude-ui
Anthropic positions Claude Opus 5.5 as its strongest Opus model yet, built for long-running agents and improvements in coding and professional work, available to business and consumer users through Claude Pro and Max.
Why it matters
The case highlights a concrete tension. The same technical capability that lets a model rebuild a jingle or a video game is the one companies choose to limit. Benchmark performance alone does not tell users what they can actually get: each company's rules decide which uses stay open and which close.
No comments yet. Be the first!