The Office of Method

There are projects that begin because a market exists.

Others begin because a problem refuses to leave.

Office of Method is for the second kind.

The work moves across mathematics, systems, intelligence, learning, writing, practice and ventures. Those subjects do not form a conventional specialty, and I do not intend to pretend that they do.

The common element is method.

A problem is defined. The existing account of it is examined. A model is built. Something is attempted. Reality disagrees somewhere. The disagreement is preserved rather than polished away. The model changes. Eventually there is an artifact—or a defensible reason not to make one.

Office of Method exists to preserve that record.

What is here

Mathematics

FRAG005 is a long-form mathematical manuscript developing recursive compressibility and MDL-style formalization, with executable reference semantics, finite-instance certification and formal verification in Lean.

Systems

Software, instrumentation, reliability, AI systems and release engineering. Current work includes LEDGER and TROUBLEDPOET: projects concerned, in different ways, with what a system may legitimately claim to know.

Intelligence

Evidence, requirements, collection, analysis, warning and decision support. This includes strategic-intelligence study and the development record surrounding real intelligence products.

CAPTUREDMIRAGE

A longitudinal experiment in serious chess improvement. The purpose is not merely to accumulate study hours or memorize exemplary moves. It is to investigate whether training can change the model from which strong decisions emerge.

Writing Room

Original fiction, research and revision. The finished manuscript matters. So does the machinery discarded while making it.

Practice

Personal practice systems and tools designed, where possible, to become less necessary as practice matures.

Venture Lab

Commercial experiments exposed to the useful cruelty of reality: costs, suppliers, customers, constraints and failure.

Workbench

The methods that move between all of them: evidence, measurement, research, stopping rules, tool use, AI-assisted work without surrendered judgment, and the difference between producing activity and producing an artifact.

What gets published

Not everything.

Some material is public. Some belongs to members because the useful object is the complete working paper, dossier, template or postmortem. Some remains commercial. Some is confidential. Some is embargoed because later publication rights matter.

The existence of an audience does not turn every private thing into content.

The rule

Publish work, not activity.

Hours are not findings. Busyness is not method. A useful record contains something that changed: a result, an error, a decision, an artifact, a model or a question made more precise by contact with reality.

There will be failed projects. There will be elegant systems that break under serious testing. There will be work that stops because stopping is the correct result.

Those are not interruptions to the record.

They are the record.

— J. Bentley Tar