# Verification, And What Only The Founder Can Decide **version** v0.3.27 · **date** 23 August 2026 --- ## Acceptance criteria The second book is not accepted because it reads well. It is accepted because these hold, and each of them is checked by a build rather than by an opinion. ### At every phase | # | Criterion | Checked by | |---|---|---| | A1 | The frozen first edition still hashes to its manifest | gate 14 | | A2 | Every node below level 1 has exactly one parent, and every declared child exists | gate 15 | | A3 | Every claim node carries an evidence state; every fact has a resolving evidence edge | gate 16 | | A4 | Every unit carries a resolving provenance block with a verdict | gate 20 | | A5 | No sentence in prose quotes a computed number as a literal | gate 21 | | A6 | Every internal link resolves; blocks balance; no banned edge; no key-shaped strings | existing gates | ### At the universe (end of phase 2) | # | Criterion | Checked by | |---|---|---| | U1 | Every candidate plot line traces end to end through the universe with no gap | the stopping rule, gate 2 | | U2 | Every node carries its family, and every claim its evidence state | gate 16 | | U3 | Every audit hypothesis from file 02 has been re-tested, and the ones that moved are recorded | a report, not a gate | ### At the spine (end of phase 3) | # | Criterion | Checked by | |---|---|---| | B1 | Every spine statement has at least three independent evidence paths | gate 17 | | B2 | No more than one of those paths runs through another spine node | gate 17 | | B3 | Any statement failing B1 is recorded in the graph as an opinion, not asserted as a theme | gate 17 | | B4 | The founder recognises the derived spine as the book he intended | an author's judgement, recorded in the register | | B5 | The pacing check passes on every plot line: no concept is used before it is introduced | gate 22, ADR-10 | ### At each altitude | # | Criterion | Checked by | |---|---|---| | C1 | The reading version survives anchor-stripping with no sentence losing its subject or object | gate 19 | | C2 | The edge set chapter matches the registry byte for byte | gate 18 | | C3 | The honesty table is generated from claim states, not written | gate 16 plus review | | C4 | Both PDFs build, the spine width matches the page count | existing gates | ### The one criterion that is not a gate **At the end of phase 4, level 3 should be measurably shorter than the first book's equivalent altitude, at equal or better coverage.** The first book's chapters 1 to 12 are roughly 15,600 words. Level 3 targets roughly 8,000. If level 3 arrives at 15,000 words, the audit's LIFT verdicts were a euphemism for CARRY and the rewrite did not happen. This cannot be a build gate because word count is not quality, but it should be reported on every build and looked at. --- ## What the rewrite does to the ten open decisions | Decision | What the rewrite does | |---|---| | **r001-D2** sign off the sequencing | **Superseded.** The sequencing question was evidence-first versus identity-first. The rewrite is both: the identity is the title of the second book and the evidence is its input. Needs restating, not answering. | | **r002-D2** pick the graph pilot chapter | **Dissolved.** Every unit is graph-first. There is no pilot. | | **r002-D3** the combined sequencing | **Superseded**, with r001-D2. | | **r003-D1** adopt the decoupling | **Answered by construction.** ADR-8: editions are joined by provenance edges, not equality. The gate flips as the third review asked. | | **r003-D2** which issues logic runs a review folder | **Untouched, still open.** Review tooling, not book structure. | | **r004-D1** does the ladder continue | **Answered yes, and promoted.** The ladder is not a section of the book; it is the book. `gen_altitudes.py` splits into three, which is what the decision was holding. | | **r004-D2** where the findings live | **Made urgent.** The second book's graph has a findings family from day one, so this needs answering in phase 1 rather than later. | | **r004-D3** which audience variant first | **Deferred cleanly.** The five altitudes are the source material for variants, per memo 1.14. The question reopens after phase 6. | | **r004-D4** confirm the taxonomy classes | **Absorbed into phase 1.** The registry is authored there; confirming it is part of that work. | | **r004-D5** the vendored dependency | **Untouched, still open**, and its premise is already corrected. | Four decisions are dissolved or answered by the rewrite. Two are made more urgent. Four are untouched. That is a real reduction, and it is a reason to start rather than to keep deciding. --- ## Open questions These are the founder's. Each one blocks something specific, named. ### 1. The freeze: copy or freeze in place? · **ANSWERED 23 August** **It is a move, to a `v1/` prefix.** Not a copy, and not a freeze in place. The second book takes its own copies into `v2/` as it needs them, so each edition owns everything it uses. ADR-1 in file 03 records the boundary, the URL cost and the gate. Shipped as v0.4.0. ### 2. Where does the second book live, and what happens to the front page? · **ANSWERED 23 August** **Each edition owns its prefix, and `/book/` is a pointer to the current edition.** The founder's decision, which is a third option neither listed here nor recommended: the second edition lives at `/v2/`, permanently, and everything about making it lives inside that tree too, including the project management and the scaffolding, in preparation for the day `/v2/` is archived and a `/v3/` begins. `/book/` at the root stops being an address that any edition owns and becomes the pointer to whichever edition is current, which from now on is the second. The root front page stays edition-neutral. Two supporting facts the founder recorded with it: the first edition was never released and nothing external links into it, so it is kept as history, source material and worked examples; and anything outside `/v1/` and `/v2/` must be deletable without either edition losing content, which is the boundary made testable. Executed at v0.4.4: the dev pack, the memos, the briefs and the review packs all moved into `/v2/`. *Phase 1 is no longer blocked.* ### 3. What are the five altitudes called, to a reader? L1 to L5 are engineering names. A reader needs words. Something like *the claim*, *the case*, *the method*, *the argument*, *the reference*, but these are a proposal and naming is the author's job. *Blocks:* phase 3, and everything the navigation says. ### 4. Wardley maps: level 4, or dropped? · **ANSWERED 23 August** **Neither. They are a spine candidate.** The founder gave the reasoning the first verdict was missing: a map is a graph that has gained position, movement and time; a map of a map of a map is the fractal claim at its most legible; the evolution axis is not fixed to genesis-to-commodity; and maps are what make a graph actionable. You cannot have maps without graphs, but maps are where graphs gain strategy and gameplay. One consequence to carry into phase 2: this is the **only** spine candidate whose evidence has not been carried onto this site. Nothing in `/v1/docs/` covers Wardley maps. The Send corpus holds eight briefs on them and none is carried. That gap is work for phase 2. ### 5. Are the external evidence weights right? ADR-4 proposes `carried_sources × 2` and `vault_demonstrations × 3` added to the peak formula. Those numbers are a proposal and nothing has argued with them. A published, recomputable formula that nobody has disputed is a proposal wearing the clothes of a measurement, which is the sharpest tension in the retrospective and it applies here first. *Blocks:* phase 2's result, though not its mechanism. ### 6. Does level 5 have a bound? Levels 1 to 4 have word targets. Level 5 is the reference layer and could be unbounded. *Blocks:* phase 6's gate, and it can wait. ### 6b. Is the economic and feedback-loop thread a spine candidate, or a plot line? · **NEW** The memo names something the current material barely carries: **the feedback loops and economic models that make a graph self-maintaining, even when the people in it do not care, do not want to, or when maintaining it runs against their own interest.** The founder wants to be able to ask the graph for the economic arguments and the feedback-loop arguments and get an answer. Two of the twenty-one carried sources touch it (*Paying The Fact Creator* most directly, *A Fact Does Not Exist In A Vacuum* second), and neither is about self-maintenance under indifference. That is a real gap, and the question is what shape the answer takes: a spine statement, a plot line that runs under several statements, or a hypothesis the book states and does not resolve. *Blocks:* nothing yet. It should be answered during phase 2, when the universe shows how much evidence actually exists behind it. ### 7. Who reviews the second book, and when? The review workflow is the site's strongest process and it produced four rounds on the first book. The second book should be reviewed from level 2, not at the end. *Blocks:* nothing yet. Worth deciding before phase 3 ships. ### 8. Is the graph RAG position sized correctly? Memo 1.17 says it "should not be a big thing", and file 04 proposes one glossary entry and one passage. If it should be a section at level 4, say so now, because it changes what level 4 argues. *Blocks:* phase 5. --- ## What would tell us this plan was wrong Named in advance, because a plan that cannot be falsified is a wish. 1. **Phase 2 passes trivially.** If all five spine statements sail through the converging evidence check first time, the check is too weak, not the spine too strong. Expect at least one failure. If there is none, tighten the check before writing. 2. **Phase 3 takes longer than phase 4.** Level 1 is 150 words. If those 150 words take longer than the 8,000 of level 3, the top-down premise is not working as claimed and the difficulty is in the compression rather than in the writing, which would be worth knowing early. 3. **All five spine candidates survive phase 3 unchanged.** The spine is supposed to be *derived* from where the plot lines converge, not confirmed. If the derivation returns exactly the five statements written down three days before the universe existed, the derivation is decorative. Expect at least one to move and at least one to be replaced. 4. **The provenance blocks are all CARRY.** If the second book is mostly carried text with new front matter, the rewrite was a reorganisation and the honest thing is to say so and go back to the refactor that was cancelled. 5. **Nobody disagrees with anything.** The method's whole claim is that it makes disagreement visible and specific. A rewrite that produces no arguments has not produced the thing it was for. --- This document is released under the Creative Commons Attribution 4.0 International licence (CC BY 4.0).