Neuruh · Convergence Graph · Design Specification

One graph.
Every system connected.

This is the convergence graph. It connects the organ registry, current gates, DeedSonar, commercial proof, runtime promotion and learning into one dependency system — not a set of separate diagrams.

Read it as dependencies, not decoration. If a step is missing, everything downstream of it is not proven. The graph exists in source. The live handoffs are what turn it from architecture into an organism.

Whole-system success predicate

One bounded real target travels the full chain without the founder manually carrying machine state between agents. That is the operational definition of the organism.

INTAKE→MISSION→AUTHORITY→EXECUTION→COURT→CUSTODY / PROMOTION→DELIVERY→RECEIPT→OUTCOME→LEARNING→PUBLIC-SAFE PROOF

Break any link — a missing receipt, a skipped court, an outcome never recorded — and the chain stops being evidence. It becomes a story.

13 + JDAGs A–I, K–N · plus §J boundary

11Steps in the success predicate

11Global stop conditions

1P0 seam named in DAG A

The DAG explorer

Select a DAG on the left. Play it through step by step, or click any card below to jump straight to it. Branch points, gates, and hard rules are kept in the step text — not simplified away.

All thirteen DAGs at a glance

The full dependency surface. Execution, product, commercial, and learning are not separate roadmaps — they feed each other.

Where the DAGs converge

Six hubs appear again and again. They are the joints of the graph. Lose one and multiple DAGs fail at once.

Receipt

The universal handoff. Execution, delivery, custody, and proof all require one. A receipt-chain break is a global stop.

AEHIK

Independent Court / Sentinel

Nothing promotes itself. Source, release, product, claim, and behavior changes all pass an independent check.

ABCKM

OutcomeRecord

The point where prediction meets reality. It closes execution, DeedSonar, commercial, and learning loops.

ACDFK

Governance / Promotion Gate

Merge, promote, deploy, and behavior updates require authority outside the worker that produced the work.

AFG

Proofstream

Internal receipts become public-safe proof only after redaction and a wording-ceiling check. Proof then drives distribution and sales.

EHM

Calibration

Forecasts are captured before outcomes, then scored. Learning, opportunity scoring, and cost decisions all depend on it.

DFN

Founder attention boundary

Section J is a design rule, not a preference. The founder is the authority at specific points — and explicitly not the transport layer everywhere else.

Founder input belongs here only

✓ Objective — what is being pursued and why
✓ Consequence — weighing what happens if it is wrong
✓ Joint business authority — decisions that bind the venture
✓ Ambiguous canonical conflict — when two sources of truth disagree
✓ Legal / financial commitment
✓ Irreversible / destructive action

The founder should not be

× Receipt courier — carrying receipts between agents
× Session memory — remembering state the system should hold
× PR scheduler — sequencing merges by hand
× Runtime watchdog — watching for failures the system should surface
× Handoff transport — moving work from one stage to the next

Why this matters: Every time the founder carries machine state, the graph has a hidden human edge. That edge does not scale, does not log, and does not survive absence.

Global stop conditions

Any one of these stops the line. They apply across all DAGs, not just the one where they were first noticed.

STOP · Identity conflict STOP · Unexpected dirty tree STOP · Different writer already active STOP · Missing authority STOP · Provider limit STOP · Semantic health failure STOP · Source / runtime mismatch STOP · Unproven public claim STOP · Scope expansion STOP · Receipt-chain break STOP · Duplicate-controller proposal

A stop is not a failure of the graph. It is the graph working — refusing to promote, publish, or proceed on unproven state.