#agents

Reasoning models write notes to themselves before they answer. Why they need this scratch paper, why it is getting harder to read, and why that matters.

Everyone obsesses over models and prompts. The real bottleneck in RAG is retrieval, and fixing it takes engineering, not bigger context windows.

Andrej Karpathy's autoresearch lets AI agents run about a hundred ML experiments overnight on one GPU. It looks like the beginning of autonomous science.

Context7 fixes the context pollution that makes agents hallucinate APIs. A supply-chain flaw raises the deeper question: who fills the context window?

Prompt engineering was the warm-up act. Context engineering decides what reaches the model, when, and how. That is what makes agents drift or compound.

When agents write and maintain API backends, the criteria shift: predictability, fast feedback and low ambiguity beat elegance. That points to Go.