Notes on building with agents, infrastructure, and the tooling in between.
SERIES · 6 PARTS · AUGUST 2026
A six-part series on letting an agent write its own orchestration: what dynamic workflows are, how to read and write the scripts, the patterns that make fan-outs converge, how to ship them to a team, how to compose them, and what they cost.
Your Agent Can Write Its Own Harness
What dynamic workflows are, how they differ from subagents, skills, and agent teams, and how to run your first one.
How to read and write a complete workflow script: typed data out of agents, parameters in, the sandbox rules, and how to iterate on a script across runs.
Pipelines, Barriers, and Verification: Orchestration Patterns That Actually Converge
How to structure multi-stage fan-outs, encode verification that actually filters, write loops that actually terminate, and know what a stopped run keeps.
From One-Off Script to Team Asset: Saving, Versioning, and Shipping Workflows in Plugins
How to turn a good run into a named, parameterized, code-reviewed workflow: in the repo for your team, or in a plugin for your org.
Conductors: Composing Predefined Workflows with Generated Orchestration
How to build the two-layer system: a library of leaf workflows composed at task time by a conductor script, hand-written or generated on demand.
The Economics and Governance of 100-Agent Runs
How workflow runs actually cost, what bounds them, how to watch them, and where the ROI is real. For the person deciding whether this belongs in their organization.