Mv1 - The honesty fence is the pitch
How to read this page
Three ways to read this page. Precise is the document itself, exactly as it is written in the repository. Plain and Clear were written for this website to help you meet that document — they are about it. They are not it, and they are not evidence.
The Encyclopedia is the UNI method written out as a reference work: 39 pages, arranged in wings, setting out what the programme is attempting and why it is built the way it is. This is where the ideas are explained in order and in prose, rather than as code, as runbooks, or as dated receipts.
Every chapter is authored against two ledgers and never ahead of them. One records what UNI has built, and the evidence class of each claim. The other records nature's own regularities, kept separate on purpose. That way a fact about biology is never quietly reused as a fact about the software. Where a chapter and a ledger disagree, the chapter is the thing that is wrong. Every chapter closes with an invitation to falsify it, and a recorded negative is published beside the result it qualifies rather than after it.
Read "How to read this work" first. It is the evidence constitution: the classes, the four ledger states, and the rule that a finished chapter is not the same as a working system. Then the calibration ledger, which carries the figures every other chapter is required to use.
What it is not: a description of a person or of a mind. The programme calls itself a developmental active-inference simulation, a bounded peek into a toy world, and its own index prints how much of the developmental ladder has actually been earned — roughly two rungs out of eleven or more. It is also not a report of what is running today. For what ran, and when, go to the evidence record.
Your browser cannot switch reading levels, so the document itself is shown.
Precise — the source document
This is the document. Rendered from the repository at the commit above, with nothing rewritten for the web. A gate re-renders it on every deploy and fails the build if a single byte differs.
The honest position of this chapter is small and load-bearing: publishing our negatives, fencing our claims, and tracing every figure to a ledger row is a method, not a capability. It is a discipline we can show works (the press kit was built this way, and it holds together), and a posture we choose to lead with in public. It is not, and never becomes, evidence that the underlying science is proven. "We are honest about what we have not shown" is a true sentence and a useful one; it is not a substitute for "we proved it," and any reader who hears the second when we say the first has been misled. This chapter exists to make that confusion impossible. The whole program remains a developmental active-inference SIMULATION, a bounded peek into a toy world, never a person and never a mind, with the honest standing position printed in full: ~2 of 11+ developmental rungs earned.
The pitch, stated plainly: we put the things we have not proven on the front page, we attach a falsifier to every claim, and we invite you to break the work rather than fund it. The differentiator is not a result. It is the refusal to overclaim, made operational.
The posture, and the one thing it is
The reusable discipline is recorded as method claim M15 (Honesty-fence-as-the-pitch), evidence class method. It has two halves. First: publish the negatives front-and-center, in a boxed "what we have NOT proven," and treat the published negatives as the credibility rather than something to bury. Second, a hard, test-enforced vocabulary-leak guard: public copy never externalizes the internal framework vocabulary (the active-inference / expected-free-energy / free-energy names) and never features an internal channel handle; it says "count baselines," "prediction-error," "LOOP not LEAP." The call-to-action that the method mandates is exact, and it is the spine of this whole chapter: the CTA is "help us independently verify," not "fund the vision."
Because M15 is carded method, it asserts no capability and cannot be raised into one. Its falsifier is a violation, not a measurement: public copy that carries the fenced framework vocabulary, that features a channel handle, that states an inflated claim, or that asks the reader to fund the vision rather than to verify the work. The fence is the claim here. Living inside it is the entire content. The discipline is to make the fence louder under pressure, not wider: even under a partner-facing push to put a chip on the board soon, the package moved no claim line.
The proof that the method runs (with its travelling fence)
The method is not only asserted; it has been executed end-to-end once, and that execution is recorded as TA3 (no-API publishing proven end-to-end, press kit), evidence class E. The seven-document press kit, plus a regenerable PDF, was authored entirely on the Claude subscription / CLI with no autonomous server-side LLM in the authoring path, and every figure in it traces to a verified ledger row backed by a one-command reproduction. The kit leads with the one proven thing (the count-baseline language PASS), puts a boxed "What UNI has NOT proven" on page one (active inference still NEGATIVE; roughly 10 to 15 percent behind backprop LLMs on char-perplexity by a chosen design trade; the audit remediation; the Phase J caveat), and demotes forward vision to a clearly labelled "belief, not proof" box.
The fence that must travel with TA3 is its own evidence class. E means test-covered, not feature-working (a Class E pass never satisfies a Class A criterion). TA3's falsifiers are concrete and they are the discipline: a press-kit figure that cannot be traced to a ledger row, a one-command reproduction that fails, or a server-side content LLM discovered in the authoring path. The claim is precisely "the no-API model demonstrably produced this artifact, traceably," and nothing about the artifact's polish upgrades the science it reports. The press kit is a faithful messenger; it is not a result.
Negatives are content (cited as a provenance-flagged snapshot)
The credibility move rests on the program's standing rule, Section 0 (Negatives are content): a partial, a negative, or a "most pieces don't help" decomposition is a measurement that the design is incomplete, not an exit and not a failure to hide. The negatives are the credibility.
This is exactly where the fence has to be loudest, because the temptation is to convert the negatives into a single impressive number. The last recorded ledger snapshot is 882 rows = 350 PASS / 0 FAIL / 183 NEGATIVE / 349 PENDING, and it must be cited as a provenance-flagged snapshot, never as a stand-alone re-countable headline. The reason is recorded in the constitution itself: the ledger derives from a deduplicated merge of 615 extracted claims, the snapshot enumerates 882 rows, and the carded body exposes only on the order of 140 distinct rows, so a skeptic counting the published rows cannot independently derive 183. Per the calibration-down rule, we headline only the negatives the ledger body actually enumerates (World C is a count baseline, not comprehension and not beats-LLMs; Phase J carries J.attribution_caveat where overall NLL worsens; the comprehension-above-retrieval central wall is thrice-NEGATIVE; the metabolism organ helps foraging but building went minus 14 percent; the EDAIT trade is roughly 33 perplexity versus a backprop GPT's roughly 25), and we present the 183 figure only with its provenance flag attached. To feature "183 published negatives" as a bare credibility number, before the count is reconstructable from the published rows, would itself be an overclaim, and it is forbidden.
The CTA, and the lift we have not earned
The mandated call-to-action is "help us get this independently verified" (a third-party fresh-split challenge), and never "fund the vision." That phrasing is the near-zero-claim-risk posture: it asks the reader to falsify, which is something a true claim survives and a false one does not.
The fence on the outcome of this posture is recorded as TA25 (relationship-first engagement ladder), evidence class B, and it is unambiguous: engagement-lift is TARGETED, NOT YET PROVEN (Class B / pending), and must not be raised to Class A. The ladder (Discover-and-Elevate, Follow-Up, Bridge, then a CTA only at step four when invited), the Deep Tagged Replies, and the own-post defense are a designed mechanism whose origin is a recorded negative: 372 posts produced one comment in 30 days. TA25's falsifier is its registered measurement: over three disciplined weeks, if author-reply rate (25 percent or higher), reply-chain depth (1.8 or higher), and own-post impressions (plus 30 percent in four weeks) fail to move, we revise the playbook, not the volume. Until those targets are met on a held window, the honest copy is that the engagement model is a bet we are instrumenting, not a result we are reporting.
What is NOT claimed in Mv1 - The honesty fence is the pitch
- Ceiling. That "we are honest / we publish our negatives" is itself proof that the UNI science is correct, that the program has won its field, or that the engagement strategy works, is NOT shown. The most we claim is: an honesty-and-traceability method exists and is reusable (M15,
method); it was executed end-to-end to produce one fully ledger-traceable press kit with no server-side LLM (TA3, Class E, test-covered not feature-working); the program publishes its negatives as first-class content and cites the 882 / 183 figure as a provenance-flagged snapshot (Section 0); and the relationship-first engagement lift is targeted but unproven (TA25, Class B / pending). - Fences engaged. Red line 7 (never raise a claim above its source evidence class) governs TA3 at E and TA25 at B. Red line 12, the vocabulary-leak guard, forbids externalizing the active-inference / EFE / free-energy framework names and forbids featuring any internal channel handle in this public copy. Red line 10 (no PII; a link may exist but the handle is never featured) applies to the publishing surface. Red lines 1, 2, 4, and 5 apply by inheritance: the copy this chapter governs never claims AGI, consciousness, created life, or beats-LLMs. The Section 0 provenance fence forbids the bare "183 negatives" headline. The standing program framing applies throughout: developmental SIMULATION, bounded peek, toy world, Class U where not claimed, unrefereed working preprint.
- Negatives that travel with this claim (cite alongside, never strip). TA25's "engagement-lift is Class B / pending, do NOT raise to A" (origin: 372 posts to one comment in 30 days) travels with any mention of the engagement ladder. TA3's Class-E falsifiers (untraceable figure / failed one-command reproduction / a server-side content LLM in the path) travel with any "we proved the no-API model" statement. The Section 0 provenance flag (882 rows, only ~140 carded, 183 not re-derivable) travels with the negatives figure every time it appears, alongside the press kit's own front-page negatives (active inference still NEGATIVE on the reader, ~10 to 15 percent behind backprop LLMs by a chosen trade, the audit-driven down-calibrations).
- Parked / owed. No sign-to-park is owed by this chapter (M15, TA3, and TA25 are not parked-frontier rows). What is owed is empirical: the TA25 engagement-lift targets remain a pending Class-B measurement that can only be moved by observation, never argued up; and the CTA's promise is itself owed a discharge, namely an actual independent third-party fresh-split verification, which has not yet occurred.
- One-line honest summary a skeptic could not dispute. We have proven that we can publish a fully ledger-traceable, no-server-LLM press kit and that we lead with our negatives and a "verify us, do not fund us" CTA; we have not proven, and do not here claim, that this honesty makes the underlying science correct or that the engagement strategy works.
Falsify this
The lead falsifier is operable and it is TA3's: take the seven-document press kit and pick any figure in it; if that figure cannot be traced to a row in CLAIM-LEDGER.md, or its named one-command reproduction fails to regenerate it, or a server-side content LLM is found anywhere in the authoring path, then the "no-API publishing, every figure ledger-traced" claim of this chapter is false as stated. Secondarily: if any public copy governed by this chapter is shown carrying the fenced framework vocabulary, featuring a channel handle, presenting the bare "183 negatives" headline without its provenance flag, or running a "fund the vision" CTA, the honesty-fence method (M15) has been violated at its definition.
Sources
- Ledger:
CLAIM-LEDGER.md- M15 (Section 3, method claims); TA3 (Section 4.1, no-API content model); TA25 (Section 4.4, engagement); Section 0 (Constitution and Standing Fences, "Negatives are content," the 882 / 350 / 0 / 183 / 349 provenance-flagged snapshot). - Authoring spec:
MASTER-PLAN.md- Wing M, Mv1; PART I front matter FM-1 (Evidence Constitution), FM-2 (A-U rubric), FM-3 (red lines + vocabulary-leak guard + the SIGNED 2026-06-27 forbidden-phrasings list), FM-4 ("what is NOT claimed" template). - Narrative grounding (PII-redacted digests):
curated/uni-mind-digest.md("The honesty-fence AS the pitch," the press-kit template and the "help us get this independently verified" CTA, "urgency never licenses a claim");curated/website-digest.md(the no-API content model, the relationship-first / instrument-the-bet engagement posture, the audit-corrected-DOWN science landings, no paid ads, the handle "link may exist but never featured" refinement). - Archive pointers (do not open; PII): the
uni-mind,orchestrate-linkedin, andwebsitesource archives behind the digests above.
sha256 47f9448bd335ccad — of the original file, so what was ingested stays checkable.
Plain — written for this website, not the source document
Leading with your own failures is a method, not an achievement, and that is the whole of the argument here. Publishing the negatives, stating plainly what is not claimed, and tracing every figure to a row in the ledger — a list added to and never edited — is a discipline the program can show works. It is never evidence that the underlying science is right, and the thing underneath is a simulation: a toy world, not a person. Saying that we are honest about what we have not shown is a true sentence; it is not a substitute for having shown something, and the chapter exists so that nobody hears the second when the first was said. The call to action it mandates is exact: help us get this independently verified, never fund the vision. It also carries its own unfinished business. A relationship-first engagement approach is a targeted bet being instrumented, not a result, and the promised independent third-party check has not yet happened.
Plain · written 2026-08-01 by claude-opus-5 · not yet checked by a person · about the document whose sha256 is 47f9448bd335ccad
Clear — written for this website, not the source document
The chapter's honest position is small and load-bearing: the honesty posture is a method, recorded as one, and it can never be raised into a capability claim. The work it is a posture about is a simulation — a toy world, and not a person.
The method has two halves. First, publish the negatives at the front, in a visible box saying what has not been shown, and treat them as the credibility rather than as something to bury. Second, a test-enforced guard on public wording, so that public copy never externalises the internal framework vocabulary and never features an internal channel handle. The mandated call to action follows from both: ask the reader to help verify the work independently, never to fund a vision. Because the method asserts no capability, what would show it broken is a violation rather than a measurement, such as public copy carrying the ruled-out vocabulary, stating an inflated claim, or asking for funding instead of scrutiny.
The chapter then records the one time the method was run end to end. A multi-document press kit was authored with no autonomous server-side language model in the authoring path. Every figure in it traces to a row in the ledger, a record added to and never edited, and each row is backed by a one-command reproduction. That run is recorded as test-covered, which is not the same as feature-working, and what would overturn it is concrete: a figure that cannot be traced, a reproduction that fails, or a content model discovered in the authoring path. The kit is described as a faithful messenger, not a result.
A third section is where the limit is stated loudest. The recorded snapshot of ledger rows and negatives must be cited as a snapshot, with a note saying where it came from, because the negative count cannot be rebuilt from the rows on show. So the chapter headlines only the negatives the ledger body actually enumerates. Those are a reading result that is a count baseline rather than comprehension, a specialist language gain whose overall score worsens, a comprehension wall recorded three times over, and a metabolism organ that helped foraging while building went backwards.
The final section puts a limit around the outcome of the posture itself. The engagement ladder it describes is a designed mechanism whose origin was a recorded failure, and its lift is targeted and pending, not shown. Until registered targets move on a held window, the honest copy is that this is a bet being instrumented.
Clear · written 2026-08-01 by claude-opus-5 · not yet checked by a person · about the document whose sha256 is 47f9448bd335ccad