The Verified Research Loop

Every chapter of The Verified Research Loop with its printed epigraph, what you can do afterwards, and the moment it is built for.

1. Introduction

   Fluency Is Not Evidence

   The most dangerous research output is the one that sounds finished.

   What you can do afterwards

   You will learn to separate the surface plausibility of an answer from the checks that make a particular claim usable, and you will leave with a concrete way to route one untraced claim.

   Use this chapter when

   You are about to paste an AI-generated market size, competitor claim, or regulatory fact into a document that other people will act on.

2. Chapter 1

   The Research Confidence Trap

   A decision can advance while the evidence behind it stands still.

   What you can do afterwards

   You will learn to separate felt confidence, source support, claim accuracy, topic coverage, and outcome probability before any of them is allowed to stand in for another.

   Use this chapter when

   A quick AI-assisted research session has produced an answer that feels ready to circulate.

3. Chapter 2

   TRACE Overview

   Five steps turn a research question into an inspectable decision record.

   What you can do afterwards

   You will understand the five TRACE steps, the artifact each step must produce, and the conditions that send a run backward, forward, or into `HOLD`.

   Use this chapter when

   A decision has generated a research question and you need to define the record before opening search results or asking an assistant.

4. Chapter 3

   Target the Question

   A research run needs a decision boundary before it needs a search box.

   What you can do afterwards

   You will produce a versioned research brief that defines the decision, the answerable propositions, the search boundary, and the reason a run may stop without pretending to be complete.

   Use this chapter when

   You are about to search, prompt an assistant, or dispatch an agent and cannot yet state which decision the research will inform.

5. Chapter 4

   Retrieve Across Sources

   A result is a pointer. A source body is an object. Evidentiary support is a claim-relative relationship.

   What you can do afterwards

   You will build a source registry that distinguishes discovery from body access, records provenance and transformations, and preserves enough context for claim-relative appraisal.

   Use this chapter when

   A versioned brief exists and you are ready to discover and obtain the bodies that may bear on its claim IDs.

6. Chapter 5

   Assess Sources and Tiers

   A source has a role. Evidentiary weight belongs to a specific claim-source relationship.

   What you can do afterwards

   You will turn a source registry into claim-relative appraisal records without letting publication type, prestige, polish, or a tier number settle what a source can bear.

   Use this chapter when

   Retrieve has produced identified bodies, versions, access states, and source IDs for a bounded brief.

7. Chapter 6

   Cross-Check Claims

   The ledger is useful when a second reviewer can reconstruct the proposition, the evidence relations, and the unresolved risk.

   What you can do afterwards

   You will build a claim ledger that preserves qualifiers and keeps source relations separate. Its status fields will not hide missing bodies, shared provenance, or an unsafe decision disposition.

   Use this chapter when

   Retrieve has produced stable source IDs and Assess has produced claim-relative appraisal records.

8. Chapter 7

   Export the Decision Memo

   The memo records what the evidence permits, what the author infers, and what the decision owner chooses.

   What you can do afterwards

   You will export a versioned decision memo that preserves claim IDs, separates evidence from inference and action, makes `HOLD` usable, and can be corrected when a source or decision changes.

   Use this chapter when

   The claim ledger contains the load-bearing propositions, one-source relation rows, currentness state, and dispositions needed for a decision handoff.

9. Chapter 8

   Research Types

   The five steps stay stable. The evidence contract, rights boundary, and stopping authority change with the domain.

   What you can do afterwards

   You will adapt TRACE to market sizing, competitor research, regulatory questions, and people diligence without carrying one source hierarchy or one confidence shortcut across all four.

   Use this chapter when

   A brief contains one of these domains, or several leaves that need separate evidence and safety contracts.

10. Chapter 9

    Agents and MCP for Research

    Assign bounded work to a dated capability, then test the assurance act that is supposed to catch its failures.

    What you can do afterwards

    You will define an agent contract, pin an MCP capability and trust boundary, test one bounded task, and reject a run that fails source, authorization, provenance, or safety checks even when the output looks complete.

    Use this chapter when

    A TRACE step may be assisted by a model, agent, tool server, browser, database, parser, or orchestration layer.

11. Chapter 10

    Triangulation and ReasonKit-think

    Agreement can be dependent, disagreement can be informative, and a valid argument can begin with a false premise.

    What you can do afterwards

    You will map evidence dependence, distinguish convergence from complementarity, externalize a contested argument, and create an executable plan for the premises that remain unresolved.

    Use this chapter when

    A decision-critical capsule is `Partial`, `Unresolved`, or `Conflicting`, or when a recommendation depends on several sources connected by an uncertain inference.

12. Chapter 11

    Failure Modes and Red Flags

    A completed loop can still be wrong. The safeguard is not ritual compliance but visible failure, controlled recovery, and a decision that can remain on hold.

    What you can do afterwards

    You will be able to detect five recurring failure patterns, classify their consequence, stop unsafe promotion, and run a recovery record through correction, review, and recipient notification.

    Use this chapter when

    Before a memo is distributed, whenever a critical source changes, when a reviewer challenges a load-bearing relation, or when a previously accepted claim is reopened.

13. Conclusion

    Conviction at Speed

    Conviction is not certainty. It is a decision made with an inspectable basis, visible uncertainty, explicit authority, and a route for change.

    What you can do afterwards

    You will assemble the complete TRACE artifact chain, choose an honest disposition, and install a review and update event around one authorized decision.

    Use this chapter when

    You have completed the preceding chapters and need to turn the method into a controlled operating practice.