Opus 5 test reveals models understand objects deeply but lack novelty
scaling01 · x · 2026-08-16
An experiment by Prime Intellect and Elie Bakouch with Opus 5 highlighted a gap in AI capabilities: while models clearly understand the objects they manipulate at a deep level, very few genuinely new ideas emerge from their outputs.
More from Models
- Span-02 debuts: open-weight 'anytime reasoning' model that takes a deadline, not an effort level — HeyAmit_ · 2026-10-16
- Anthropic's slowdown overblown: August still grew MoM, frontier workloads stay sticky — FinanceYF5 · 2026-10-16
- Anthropic's 'growth slowdown' overblown? ARR projections show steady climbs for both labs — FinanceYF5 · 2026-10-16
- Dev: Claude writes 'zero bugs' on healthcare apps, 10x token efficiency — repligate · 2026-10-16
- Shanghai AI Lab unveils Intern Lumina U2, a diffusion LLM unifying image, video and 3D tasks — Shanghai-AI-Laboratory · 2026-10-16
- This Week in Tech breaks down Gemini 4, OpenAI's Dots agents, and Anthropic's leaked IPO filing — LouMM · 2026-10-16