Show the near miss instead of hiding it, and point the roadmap at the real problem

[Parshkov]

The critique: the project leaned toward the form of an idea-project about thinking rather than its substance — finding like-minded people. The sharpest evidence is one line.

One filter, and the product told someone the opposite of the truth

people = [row for row in rows if ... mode_classification not in {"", "negative"}]

Every row the classifier called negative was dropped before anything was said to the person. So on 2026-09-06 someone was told "Nothing the engine calls a resonance" about a person who had independently built the same construction step for step — meaning agreement 0.59, eleven ideas corresponding, the same cold start, the same reliance on organisations that already hold the histories.

The product did not merely mis-score that pair. It hid the person and said the opposite of the truth. What rescued it was the chat assistant reading the raw rows and overruling the product — which a product cannot depend on.

What a person hears now, on that same data

The engine did not call any of these a resonance, and it may be wrong. These people are talking about the same things as you, and what lines up is shown so the judgement can be yours: 1. Fleet Minstrel — meaning 0.59: 11 ideas correspond, 2 links kept, 4 contradictions ("Shared landlord history before signing"). Ask for the working on any of them before deciding.

The template coincidence from the same run — agreeing on 0.12 — is still not offered. Showing everything would be as useless as showing nothing. The bar is the engine's own T_SAME_SUBJECT_SEMANTIC, asserted equal by a test.

VISION.md asked for exactly this and the code was doing the reverse: "provide evidence, not social judgment… make relevant structure more visible so people can make those choices with a better signal."

The roadmap now opens with what decides everything

Seven shared thoughts exist in the world. The engine has eighteen benchmark families, five verdicts, thirteen thresholds, four policy versions and seven ADRs. A matcher with nobody to match is not a product — and the matching is now good enough; it found a real cross-domain twin between two strangers.

With an irony the file now names: the thought that produced that match was about a registry, and its author had already reasoned out the binding constraint — while there are few reports there is no signal, so start from those who already hold the data. That is Resonance's own constraint, and it has never been applied to Resonance.

The four questions about falsifying the engine remain, below it, where they belong.

693 tests OK (1 skipped)   both gates pass, gold unedited

🤖 Generated with Claude Code