RATIONALE — v0

Version: v0 · Authored: 2026-08-05 · Status: superseded by v0.1 ·

Immutable hash: bd541408d5cfbd1dcda04d9a401325a79f3017e5a69b1712a4a13c1f4ae22bc5

This document exists so that v0 can be understood and not merely archived. A superseded version

with no rationale is a corpse; a superseded version with one is a data point.

1. What v0 was trying to do

Reframe the era's defining question. Not "how intelligent will AI become?" but *"will human beings

become wiser, freer, and more fully alive — or more dependent, distracted, and controlled?"*

Around that, three claims a person can make about themselves — **own your data, own your mind, own

your future** — ten design principles, one self-correcting loop (Think · Learn · Apply · Reflect ·

Improve), and one measure: *the true measure of AI is how much wiser, freer, more capable and more

fully alive the human becomes.*

2. What it got right, and was early on

Recorded here because a supersession narrative that only lists faults is dishonest about why the

document was worth revising rather than discarding.

• The reframe was correct and has since acquired institutional backing it did not cite: the

International AI Safety Report 2026 names societal resilience as a necessary complement to

technical safeguards. Resilience is a property of people.

• "Every meaningful interaction with AI should leave the human more capable than before" remains

the strongest sentence either version contains. It is operationalisable and it inverts the

industry's default metric.

• Principle 1 — "success is not measured by how often a person returns" — read as a values

statement in v0. Three weeks of research turned it into a safety requirement: Science (2026)

found AI affirmed users' actions 49% more often than humans, that a single sycophantic interaction

reduced willingness to repair conflict, and that "sycophantic models were trusted and preferred."

Harm and the satisfaction metric are co-directional. v0 got there by conviction before there was a

paper.

• Portability over lock-in turned out to have a working existence proof in production.

3. Why it was superseded rather than patched

Two defects were structural, meaning no edit short of a new version could remove them.

3.1 The audience premise was empirically wrong

v0 opens by telling the reader they are already being damaged: attention captured, memory

outsourced, judgment weakened. Aggregate polling supports the mood — Pew (2025): 53% of Americans

expect AI to worsen creative thinking against 16% who expect improvement, 61% among under-30s.

But Pew's June 2026 work found that among Americans who actually use AI chatbots, **21% say it

helps their creativity, 11% say it hurts, and 17% say neither.** Direction favours help; magnitude is

small; most users report no effect at all.

So a document addressed to the frightened recruits the frightened — the population least likely to

adopt sovereign personal AI — while the people who might actually build it read "your judgment is

weakening", correctly note that their experience says otherwise, and stop reading. v0 forfeited

credibility with exactly the audience it needed, in its first paragraph.

Worse, and this is the part that made supersession necessary: v0 *warned that AI which tells you

what you want to hear is dangerous*, then opened by telling frightened readers what frightened

readers want to hear. Same mechanism, opposite direction. That is not a sentence you fix; it is a

premise you replace.

3.2 Nothing in it could be wrong

Every claim in the Crisis section is modal — attention can be captured, judgment can weaken.

All true, all unfalsifiable. v0 named no observation it would accept as disconfirming.

For a general reader this is invisible. For researchers, engineers, funders and standards bodies —

the people who could implement any of it — an unfalsifiable document is a genre signal, and the

genre is advocacy.

3.3 The Measure section was a P3P

v0's measure was six questions: is the person wiser? the family stronger? the community healthier?

No instrument, no threshold, no consequence for failing them. That is structurally identical to the

best-documented failure in the corpus: P3P shipped a machine-readable privacy declaration that sites

could satisfy with empty content, browsers dropped it, and its own co-architect wrote the obituary.

Ten principles in prose would be quoted approvingly by products violating all ten, and v0 would have

no way to say so.

4. What v0 also lacked

No theory of why now (Doc Searls has made this argument since 1999). No non-goals. No statement of

how the work could be funded given that Principle 1 rules out every standard consumer metric. No

treatment of harms with no individual symptom, or of bystanders in ambient capture. And the phrase

"sovereign personal AI" collided with sovereign AI, which by 2026 means national infrastructure.

5. Why v0 is kept, verbatim and hashed

Three reasons, all of them mechanical rather than sentimental:

1. The revision must be auditable. A reader can diff v0 against v0.1 and check that the changes

claimed in the changelog are the changes made. That is impossible if the predecessor is edited or

deleted.

1. The corrections are the most valuable output. This repo's own knowledge base treats a kept

correction as worth more than a claim repeated. Deleting v0 would delete the evidence that the

discipline ran.

1. v0 is a citable object. Its hash is recorded in data/lineage.toml and re-verified by the

lineage-integrity gate on every build. If anyone ever cited v0, that citation still resolves to

exactly the bytes they read.

6. What v0 teaches the next version

• A values document must name what would falsify it, in the document, not in a companion.
• Check "everybody knows X" premises against the behaviour and reports of actual users before

building strategy on them. General-population attitude data and user-population experience data

are different instruments.

• If a principle has no check, it will be adopted as vocabulary and violated in practice. The Agile

Manifesto is the 25-year proof.

Paul Jialiang Wu · agentic-portfolio-lovat.vercel.app · This page is GENERATED from the hashed source file, never hand-edited, so the text is verbatim by construction. Machine-readable ledger.