Built on the standards · PT4AI
A workbench for holding a claim to account.
The Precision Toolkit for AI carries precision, the discipline of getting the reference right and building on it without loss, into the work that usually goes without it: coordination, evaluation, and meaning. It is a workbench, not a method you adopt whole. You reach in for the tool that fits the problem in front of you, and the tools are built to combine.
What is actually new here
The rigor is borrowed. The edge is holding a claim to account in place.
The precision itself is not new, and the toolkit says so. It is drawn from where it was already worked to a high finish, metrology and the physical sciences, linguistics, forecasting, the study of how institutions actually hold, and reinstated where the material is softer. That shared source is why the tools fit together rather than sitting as a pile.
What the toolkit adds is a way to hold an arbitrary claim to account on the spot, with an observer-independence test as the crux: a claim stands only where an independent check could have found it wrong. That test is the through-line from the vocabulary check to the council, and it is what separates a claim that has been examined from one that has only been asserted well.
What is on the bench
Four instruments, built to chain.
Reaching for one tends to set up the next. Each carries a step-by-step guide, and each is built so a person can hand as much of it to a machine as the work allows.
The three-plate method
Three surfaces ground against each other make a true flat with no reference plate assumed. Applied to judgment: independent reads checked against each other, so no single read is trusted as the truth. This is the move the Precision Council runs at scale.
A sharper Occam's razor
Cut a claim to the form that still survives the checks it has to pass, not just the fewest parts. The general-register essay behind it, A Sharper Occam's Razor, is published separately.
The vocabulary check
Flag a word carrying more weight than it has earned and offer the plain replacement, so a claim cannot hide behind a soft term. This is the Frame Language discipline, and the one tool here that already ships as a callable MCP server.
Decomposition
Break a claim into the parts that each have to hold, so the whole cannot pass on the strength of one vivid piece while the rest goes unchecked.
Where it sits
The human front, with a callable layer underneath.
The toolkit is the workbench a person reaches into, with copy-and-run prompts at its web front door. The vocabulary check already exists as a callable MCP server, the Frame Language server, so an AI assistant can run that discipline directly rather than a person carrying it by hand. Published under Creative Commons Attribution.