Rick Hallett

Software Engineer

Leading AI-augmented development

I built a multi-model adversarial review pipeline because I understood that a single verification layer is not sufficient when your generation source is probabilistic.

Shipping production TypeScript, Python, and Go. Building The Pit and learning in public about what AI-augmented engineering requires.

earth. lets stop the slop.

writing / research / about

Recent

2026-03-18 Behavioural evaluation at machine speed: testing AI assistants before deployment
Pre/post Likert instruments, multi-turn dialogue scenarios, three-strike relent rules, and an eval harness that runs 8 behavioural tests in seconds.
2026-03-18 Personal AI infrastructure: fleet provisioning for non-technical users
How I built a system that provisions independent AI assistants for family members with filesystem-locked governance, personality calibration, and zero configuration burden on the user.
2026-03-18 The halos toolchain: 8 CLI modules for personal AI infrastructure
Memory graphs, work tracking with 13-state machines, agent telemetry with spin detection, and the CLIs that make fleet management auditable.
2026-03-10 The Agentic Engineering Bootcamp
A first-principles curriculum for SWEs who steer AI agents. Five bootcamps, 51 steps, 208-259 hours. From Linux primitives to evaluation pipelines.
2026-03-01 Build your house from paper and 0.37% of the time you'll get screwed 100% of the time
A standing order said 'read the file.' The file was 33,700 tokens. So we wrote 60 lines of JavaScript.

all posts →