Latest addition
Inside GPT-2
A small model, understood. Follow tokens through attention, the MLP, and training, then see what GPT-OSS changes. Includes an illustrated PDF and transcript.
01 / Inside the model
Model internals
Architecture, training, and what changes when you change the weights.
How Language Models Learn and Change
Probability, representations, gradients, optimization, generalization, adaptation, and forgetting—a guide to how models learn.
Listen on MidnightRun
Abliteration
Model internals, the geometry of refusal, intervention methods, measurement, and the tradeoffs in changing model behavior.
Listen on MidnightRun
02 / Write it clearly
Coding
Programming ideas you can take away from the screen.
Swift, Heard Clearly
Follow a small command-line health checker through types, optionals, protocols, concurrency, cancellation, and testing. Written for listening, with no code snippets required.
03 / Around the model
Agents
The systems that connect models to tools and real work.
Building AI Harnesses
A longer guide to the harness around an AI model: the structure that makes an agent useful, inspectable, and dependable.
Good Design for Agent Harnesses
A compact set of design notes for thinking through agent harnesses and the choices that shape their behavior.
Listen on MidnightRun
04 / Test the boundary
Security
Application behavior, protocols, and the assumptions between them.
HTTP Boundaries
A nine-part field guide to authorized SSRF, shared-cache, OAST, and HTTP boundary assessment. Includes a companion PDF.