Security

BMAD and Micrantha converge on several autonomous-development mechanics, but differ in where context, identity, authority, durable state, repository truth, and evidence live.

The AI-native SDLC gets the bottleneck shift right. My own architecture pushes the idea further: generated artifacts are cheap, but trusted acceptance and authorized effects must remain explicit.

What a decade of self-hosting taught me about reliability, portability, Canadian infrastructure, and AI

Git metadata now crosses execution boundaries in CI and agentic systems, turning familiar repository fields into potential code execution inputs.

Two beerware-minded tools for reducing mobile configuration risk: one by hardening what ships in the client, the other by hardening what the backend is willing to deliver.

A practical reference for designing authentication systems that can withstand compromise, replay, abuse, and operational failure.

Securing untrusted clients across the lifecycle

A practical threat model for AI-assisted and agentic software development across requirements, architecture, implementation, testing, and operations.