Foundations & Architecture
How LLMs work, their strengths and limits, context windows, the agent loop, tool use, and the primitives — MCP, skills, plugins, subagents, memory, RAG — and when to reach for each.
→ open topic// learning journal
Designing AI adoption for the enterprise
A running record of what I'm learning about designing AI adoption inside an organization — written from the perspective of someone building on top of existing models: agents, skills, plugins, scripts, and marketplaces, rather than training models from scratch. Each topic below is filled in with examples, diagrams, and notes as I go.
The topics
Best Practices
Design patterns for shared skills, agents, and plugins that hold up across many users and teams, not just a single workflow.
→ open topicToken Efficiency
Getting great results at lower cost — context budgeting, caching, model selection, and measuring the tradeoffs.
→ open topicEvals
Offline evals before shipping, and online evals once a system is in production — what to measure and how to trust it.
→ open topicGovernance
Security, permissions, risk, privacy, and keeping meaningful human oversight of AI behavior in production systems.
→ open topicDistribution, Adoption & Framework
Marketplaces, reaching non-technical contributors, rollout strategy, operating model, and a single framework tying the previous topics together.
→ open topic