Datumline methods

Operating methods for people who run AI agents and make decisions with them: governance that classifies what an agent may do, receipts that refuse an unproven "done", independent verification, decorrelated ideation, and finance-grade decision frameworks.

Each product is a working kit, not a prompt pack: method documents, templates, and scripts whose tests fail when the method is broken.

Start free

Receipted Operator is a free, MIT-licensed Claude Code plugin. It adds receipts, a truthful status set and a verifier that catches false completion claims.

/plugin marketplace add datumline/receipted-operator
/plugin install receipted-operator@datumline

Receipted Operator Pro

governance and an independent verifier for AI agent work

Three tools in one download: the Receipted Operator plugin, Audit Trail (tamper-detecting receipts and an oracle that catches a verifier that stopped checking) and the Governance Gate Kit (a gate that classifies every agent action the same way every time). Audit Trail and the Gate Kit are not in the free repository.

Get Receipted Operator Pro

Methods

Agents Roster

four household AI agents with charters, never-do lists and their own scripts

Four working AI agents for running a household or small family office, each with a written charter and its own scripts.

Audit Trail

receipts, a ledger that refuses unreceipted done, an independent verifier and its oracle

Receipts for AI and human work: a ledger that refuses "done" without evidence, a verifier that checks independently, and an oracle that catches a broken verifier.

Decomposition and Cascade

split a prompt into atoms, change one, and let every dependent check itself

Split a long prompt or policy document into atoms, map what depends on what, and change one line without silently breaking the others.

Domain Expansion

dissect an idea to its root, import another field's theory, keep what survives

Take an idea down to its root mechanism, find where else that mechanism has been studied, borrow that field's theory, and keep only what survives your own facts.

Idea Factory Council

decorrelated multi-model arms, red team, blind judge and a no-stake governor

Run a question past several AI models at once, with assigned lenses, an adversarial red team, a blind judge and a no-stake governor, so you get decorrelated judgement instead of one model's confident average.

Deal Assessment

deal memo pipeline with red flags, scenarios, kill list and a decomposable confidence

A structured deal-memo pipeline: intake that refuses to assume, a red-flag checklist, a scenario table, a kill list, and a verdict with a confidence you can decompose.

Portfolio Estimates

sizing, comparables and path-to-target on an assumption ledger that re-marks

Compose a research-backed estimate for any space (market size, comparables, path to a target) where every number lives in an assumption ledger and every output is re-marked when an input changes.

Real Estate Calculations

every metric the inputs support, every missing one named, no defaults

A schema-driven property calculator that computes every metric its inputs support, names every metric it cannot compute and why, and never fills a gap with a default.