Double pendulum video explains why LLMs are sensitive and unstable
emax · x · 2026-08-30
This post shares a double pendulum video as the best free explainer for why changing a single word can break Claude or GPT-5. The author argues that LLMs are not deterministic machines; every prompt is like a chaotic double pendulum. A tiny perturbation in the initial state (like one comma or whitespace) sends the entire response into a completely different attractor. This butterfly effect is the exact mathematical reason why API calls to models like Claude feel unreliable. The video is 26 minutes long.
More from Models
- Hands-on with Tencent Hunyuan Hy4: Execution enters top-tier open source — HeyAmit_ · 2026-08-30
- Opus 5 shocks user by silently completing complex full-stack tasks without chain of thought — PawelHuryn · 2026-08-30
- Commentary on GPT-4.5's language quality and compute scarcity — haider1 · 2026-08-30
- Rumor: Krea 3 to Feature Editing and 'May' Have Open Weights — ArttTaku · 2026-08-30
- 0.8B local model fine-tuned to match frontier model in dictation cleanup — MoodOdd9657 · 2026-08-30
- Disjoint Token Spaces Block Cross-Lingual Transfer; Shared Mapping Improves It 14x — LChoshen · 2026-08-30