blog | Alex L. Zhang

September 26, 2026 · 12 min read

Language Model "Shape"

Some short musings on the shape of language models, e.g. what it means to design a language model around a harness, and not the other way around.

August 24, 2026 · 19 min read

Speculative Programmatic Tool Calling

Speculative programmatic tool calling pre-launches tool calls from a partially generated REPL program, while the harness is still producing tokens, instead of waiting for the whole generation to finish.

harnesses speculation RLM tool-calling

July 20, 2026 · 30 min read

Language model harnesses are compositional generalizers

Modern post-training brute-forces ever more environments and ever longer horizons, because frontier Transformers are still poor at compositional generalization. A good harness makes every model call locally in-distribution, so what it learns carries across lengths and domains.

harnesses compositional locally in-distribution

April 26, 2026 · 10 min read

A Mini Exercise on the Mismanaged Geniuses Hypothesis (RLMs on LongCoT)

We underestimate how good language models actually are, and they are inhibited by how we use them. On LongCoT, frontier models fall short of compositional reasoning tasks whose pieces they can already solve in isolation.

mismanaged longcot

April 09, 2026 · 10 min read

The Mismanaged Geniuses Hypothesis

AI models are already good enough for the next leap in capabilities. The mismanaged geniuses hypothesis says frontier language models are severely underutilized, because each individual call is used poorly.

mismanaged scaffolds