A NOTEBOOK ON AI ORCHESTRATION
OrchestrateAI is my working notebook on scoping AI work, building harnesses, and the deterministic systems that outlast the hype.
Updated weekly · from inside the work
WHAT I WRITE ABOUT · 08 TOPICS
When low-code earns its keep, when it has to move to a container, and the cost-versus-control trade in between.
→ 02 ContextQdrant, MongoDB, or a markdown library? Managing context for cost, and weighing determinism against flexibility.
→ 03 WordPressAI for SEO, landing pages, custom plugins, and everything in between.
→ 04 InfrastructureRetry logic, idempotency, audit trails, and the boring engineering that keeps AI agents reliable past the demo.
→ 05 SecuritySecure APIs, zero-trust rulesets, shrinking npm exposure, vetting models, and the boundaries I won't let an agent cross.
→ 06 HarnessBuilding Loopcycle. Planner, coder, reviewer loops that ship code without supervision drift.
→ 07 MarketingThe hurdles of plugging AI workflows into Google Reviews, Google Ads, Facebook, and Instagram, and where the APIs fight back.
→ 08 LearningHow curiosity, operations work, and model-assisted building became a practical apprenticeship.
FROM THE BUILD LOG
Architecture decisions, deployment patterns, and the rewrites that survive critical review. Written while I build Loopcycle and ship AI work for real businesses.
The Anthropic and Washington showdown that pulled Fable 5 and Mythos 5 dark three days after launch. The official reason is national security. It looks a lot more like consolidation in the making, and the only real hedge is staying model agnostic.
Anthropic starts metering programmatic Claude usage on June 15. They have a business to run, but their incentives and yours point in opposite directions: you want cheap tokens, model-agnostic workflows, and zero vendor lock. They want the opposite.
A career note on curiosity, early bets, Zengar, and the practical digital infrastructure work that came from filling high-value gaps.