
Analysis
The model is not the bottleneck: two essays on what still needs a human
A developer on why coding agents stay weak while models improve, and a mathematician on why LLM proofs stay inside the hull of known ideas: both locate the scarce part outside the model.