Benchmarks
Thirty-nine models, fifty-five dragons, one bubble after another
Thirty-nine models drew a dragon blind, as SVG. Confidence correlates with quality negatively, and twelve of them invented tools they never held.
Articles
Long context
Fiction has no compiler: four LLM failures you only see where the build never breaks
A developer has a compiler to catch mistakes at once. Long-form fiction has none. Four findings about what a model does when nothing corrects it.
Long context
The model wrote my turn for me, then reached for the tool that ends the conversation
A model began calling the tool that ends the conversation on its own. When it happens, four refuted explanations, and what finally stopped the dangerous one.