Magic Prompt? User Claims Saying "Vamos" to LLMs Boosts Performance
wavefnx · x · 2026-08-05
An X user shared a hilarious "magical" prompt engineering trick: he discovered that telling the large language model "vamos" significantly improved its performance.
Following up with further tests, he found that simply using encouraging words (like "con-") made the model perform excellently. This reflects the sometimes unpredictable and superstitious nature of interacting with LLMs, sparking amused resonance within the AI community.
Related event: Adding "Vamos" to Prompts Magically Boosts Model Performance(2 posts)→
More from Fun
- Claude Builds 3D Downhill Racing Game with Zero External Assets via Procedural Generation — prasenx · 2026-08-06
- Robots host a podcast: Reachy Mini bots discuss small models and argue whether a hot dog is a sandwich — Gradio · 2026-08-06
- Gradio x Modal Hackathon Winners: Browser AI Dueling Game Takes First Place — Gradio · 2026-08-06
- A date breaks a 160-day Codex streak, prompting jokes about poor planning — justalexoki · 2026-08-06
- Engineers in AI Startups Are Merely 'Wrappers' for Claude Code and Codex — burny_tech · 2026-08-06
- Google Codebase Anecdote: Jeff Dean Among the Few Top-Level OWNERS — peterjliu · 2026-08-06