Working systems
Arguments become artifacts you can inspect: a released system, a working model, or an experiment with its limits left visible.
Operating infrastructure for coding agents.
campaign / phase-03 / verifiedAgent operations
The hard problem is not getting an agent to write code. It is keeping work legible after the chat ends—through routing, durable campaigns, verification, parallel work, and resumable handoffs.
The working shelf
Each artifact has a different job. Read the short version here, then open the case when you want the mechanism and the boundary.
The Coordination Tax
Make a hypothesis inspectable before committing to a product. This model traces the handoff where responsibility and context thin out.
Open the investigationTrace how responsibility and context move through a fragmented system.
The richest context rarely crosses the boundary. The person repeats, reconciles, and carries it forward.
Verifiable acceptance
Signed decision artifacts make an acceptance trail inspectable without pretending provenance is objective truth.
Read the receiptContext infrastructure
Shape what travels between systems, scrub secrets, and leave a payload the next tool can inspect.
Trace the pipeline
Playful systems
Software can create presence and delight, not only efficiency. This is a small case for making room for both.
Enter the habitatLab standards
We label what exists, show the test, publish the limit, and record what changed. A concept is not a shipped system.
Model, reconstruction, prototype, deployed tool, or operating system.
What would have to change for the underlying argument to hold?
What the artifact cannot establish, reach, or solve.
Success, failure, surprise, and the next decision.
Build with the Foundry
Bring domain authority, affected-person access, implementation capacity, or funding for a bounded experiment.
Explore a partnership