Developer Reverse-Engineers 520 GPT-Image2 Cases Into an Atomic Prompt Skill Library
bigaiguy · x · 2026-08-24
A developer has reverse-engineered 520 community GPT-Image2 cases into a Claude Code skill, awesome-gpt-image-2. Viral GPT-Image2 outputs on Twitter — Apple-style science posters, infographic explainers, classical Chinese scrolls — are all downstream of prompt structure.
Most people copy prompts verbatim, tweak two words, and get broken output, because these prompts are prose: subject, lighting, layout, materials and copy are packed into one dense paragraph. The project organizes 520 cases across 13 categories (UI, infographics, posters, products, portraits, illustration, brand, architecture, characters, scenes, classical Chinese, documents, misc), compressing each into "Prompt as Code" with an atomic schema — prompts split into composable parts like subject, so changing one field only affects that dimension.
More from coding & agent
- Decision models match LLMs on routing and safety — and are 7-8x faster — randal_olson · 2026-10-16
- Perplexity open-sources SPACE sandbox, partners with CoreWeave, Akamai, Crusoe — AravSrinivas · 2026-10-16
- LLMs beat decision models on knowledge, but not on routing or safety — randal_olson · 2026-10-16
- Temporal Serverless Workers on Cloud Run: queue-driven autoscaling, scale to zero — rseroter · 2026-10-16
- Meta's Superintelligence Lab Runs 200+ Direct Reports—A Lesson for Flat Multi-Agent Systems — peterjliu · 2026-10-16
- LangChain CEO: the most interesting agent harness work lives at the harness-infra boundary — hwchase17 · 2026-10-16