The Oolong Lab
Writing & Experiments
Field reports, benchmarks, and research notes — published in the open as we build.
Research
30 July 2026
13 min
Hard facts, soft prose: Blended Generation with Recursive Language Models
Autoregressive models sample every token, including the ones that must be exact. We split the output into two regimes: computed where variance is failure, generated where variance is fine, and let the interpreter act as the external critic that models provably cannot be for themselves.
Research
2 October 2024
4 min
RAG systems have a recall problem, not a hallucination one
Stopping hallucinations are only half of the target of information retrieval systems
