Writing Dispatches from the deep.

Technical posts and essays.

2026 · 07 · 14

RL Post-Training on Macs

Multi-turn RL of an 8B mixture-of-experts model with 14 Macs in four countries generating rollouts and one B200 training it. Held-out pass@1 on PaperSearchQA went from 29% to 63%.

technical 30 min Pluralis blog
2026 · 01 · 05

The Act of Creation

On moving from passive absorption to active curation, and finding the self through creation.

essay 12 min Substack Liminal
2025 · 08 · 14

Introducing q Evaluation Harness

The first open-source evaluation framework for LLMs on q/kdb+. At launch the best model, Grok 4, solved 43.4% of q-HumanEval on the first attempt, against 96.2% for top models on Python's HumanEval.

with Andrew Morrison
technical 7 min KX Systems on Medium