GPT-6 Astra cracks 'impossible' image-to-animation task with a minimal prompt
BLUECOW009 · x · 2026-09-08
A user previously challenged on X that "no LLM can solve this task": given a first and last frame image, generate an animation whose arrows never intersect the black lines. Now another user reports GPT-6 Astra solved it—with the simplest possible prompt, just the two images and a short description of the goal, without any elaborate explanation of the drawing or expected approach.
More from Models
- Dhenu Vision 1.0 hits 94.4% precision, beating every frontier model tested — DevDminGod · 2026-09-08
- Developer: Astra Is the First Model That Feels Genuinely Opaque and Alien — MajmudarAdam · 2026-09-08
- Ethan Mollick: GPT-6 Pro Now Spots Real Flaws When Reviewing Academic Papers — emollick · 2026-09-08
- Graham Neubig Underwhelmed by Astra for Coding, Sees GUI Computer Use as the Real Leap — May_F1_ · 2026-09-08
- SemiAnalysis: Gemini 3.8 Flash and Muse Spark 1.3 look clearly benchmaxxed — firstadopter · 2026-09-08
- Flash 3.8 turns out to be surprisingly good at 'hearing' spectrograms — cgarciae88 · 2026-09-08