visitor@andthattoo:~$ls posts █
>
[2026-06-10]IronText: A Long-Horizon Test for LLMs with an Exact Oracle - Small text POMDPs graded by an exact Bayesian filter. Opus 4.8 loses 63% of corridor episodes to one bookkeeping bug, a stamp that adds no information removes it, and the four carry regimes read as four verbs with one missing.
[2026-04-25]Structured CoT: Shorter Reasoning with a Grammar File
[2026-04-07]Emergent Computation via Cellular Automata
[2026-04-04]Scaffold: An External DSL for Autonomous Harness Optimization