Newsletter
Engineering notes on building reliable AI agents (architecture, orchestration, evaluation) and running evidence-gated workflows (policies, prompts, procedures).
Also covers safety boundaries for tool-using systems (permissions, guardrails).
- Agent architecture patterns and execution control planes (LLM-led vs controller-led).
- Evaluation harnesses, verification workflows, and evidence-gated quality gates.
- Copy/paste templates: prompt templates, checklists, and operational procedures.
- Security boundaries for tool/retrieval use (prompt injection, routing, RBAC) where relevant.
