Asset Zero · The development archive

How the framework was made

In fourteen weeks an Australian and New Zealand community went from a question — where does AI end and human judgement begin? — to a unanimously ratified framework. This page is the record: what was asked, what came back, what was decided, and the receipts for all of it.

public webinars, June–July 2026

2,796

recorded community responses

volunteer review cycles in August

162 / 162

final-pass verdicts — zero objections

The timeline

Versioned in the open, end to end: v0.1 skeleton → v0.5 full draft → v0.6 (cycle 1) → v0.7 (cycle 2) → v1.0-rc (final pass) → v1.0, frozen 31 August 2026.

25 Jun – 30 Jul 2026
The webinar series asks the question
Four public webinars walk the asset lifecycle — knowing the asset; condition, performance and risk; planning and investment; work and operations — each asking where AI belongs in that part of the work. 2,796 consented responses become the evidence base.
See it →
2 Aug 2026
The skeleton — 18 sections, promised in public
An AI-native project (codename Asset Zero) drafts the framework skeleton from the evidence and publishes the plan, the open-questions register and its own pre-registered AI boundaries. The deadline is set: v1.0 freezes 31 August.
See it →
10–14 Aug 2026 · cycle 1
The principles pass through the working group
A volunteer working group judges every principle, writes break-cases against them, and restructures the set (Structure B, 6·4·1 with dissent recorded): records and accountability become separate principles, and the set closes at seven.
17–21 Aug 2026 · cycle 2
The instruments face practitioners
Eleven of eleven members complete all three instruments — 132 reasoned matrix calls hold every drafted level (one row closes with recorded dissent); the boundary test stops a vendor optimiser eleven times for eleven different reasons; the envelope anatomy is rebuilt in the group's own words.
24–27 Aug 2026 · cycle 3
The whole document, one last call
All 18 sections go to a whole-document last call with a reason required on every answer: 162 verdicts, every one a yes. A named coverage question decides the eighth-principle candidate, and a priority vote sets the published order of the recommendations.
31 Aug 2026
Version 1.0 — frozen, signed, citable
The Project Chair answers every comment in writing, closes all 18 sections (two with dissent recorded, shipped in the text), and the freeze makes the snapshot the citable record — on the exact day promised at kickoff.
See it →

Three cycles, every input answered

Each cycle published its packet in advance, took structured written input, and answered every comment in a public disposition log before a section closed. The packets ship in full on the progress page; the written answers publish on each section of the framework.

Cycle 1 — The principles

10–14 Aug 2026

What was asked

Judge all seven draft principles (keep · sharpen · reject, with reasons), break them with real cases from your own experience, and answer the structure question: should records and accountability stand as separate principles?

What came back

Eleven responses judged every principle — none was majority-rejected; eleven break-cases were written (nine opted into the public lessons library); ten members answered the missing-principle question, six proposing an eighth.

What was decided

Structure B adopted on the 14 Aug call (6 · 4 · 1, the four recorded); the former principle 1 became the framework's test of purpose; “named person” became the defined accountable person; every comment received a written disposition before the sections closed.

Cycle 2 — The instruments

17–21 Aug 2026

What was asked

Review all twelve rows of the Decision Rights Matrix at their drafted AI levels, run the nine-question boundary test against a live vendor-optimiser scenario, and set the envelope's numbers — including the volume budget.

What came back

Eleven of eleven members completed all three instruments — 132 of 132 matrix calls carried reasons; every drafted level held; the boundary test stopped the scenario eleven times for eleven different reasons; the room's first number (7.5 machine-created work orders per planner per week) entered the envelope.

What was decided

The matrix ratified with row 8 (risk acceptance) closed on recorded dissent — the arguments ship with the framework; two double-barrelled test questions were rewritten; envelopes became organisation-specific, split by work-order type, with the stop-list in the group's own words.

Cycle 3 — The final pass

24–27 Aug 2026

What was asked

Read the complete release candidate — all 18 sections — and answer, section by section: can you live with this? Plus the eighth-principle coverage question and a ranking of the nine recommendations.

What came back

Nine members of record completed all three instruments with zero late entries: 162 last-call verdicts, all yes; one member's full consistency audit surfaced fifteen cross-reference defects, every one fixed and logged in the framework's own errata list.

What was decided

The coverage question voted 6–3 to keep the candidate eighth principle — and the Project Chair withdrew his own candidate anyway, in deference to the dissent's reasoning; v1.0 ships the seven principles the group built together, with the full record inside §B2. The priority vote set the published order of the recommendations, tie recorded.

What the rooms said — the eight findings

The evidence the framework stands on, from 2,796 recorded responses — every claim traceable to the room that said it, on the evidence archive and in the full synthesis.

The ladder holds at Recommend, until the room watches bounded action work

the community's default is Recommend; demonstrated, bounded, stoppable, ledgered action moves it one step — to approval-gated action, not autonomy. The ladder's load-bearing rung is Act with approval, and it is earned by showing the controls, not by asserting them.

Feeds framework:B1 B3

Read finding 1 in full →

Approval over interruption: the gap that widened for three webinars, and W4's re-price of the stop

approval is the community's currency, but the stop was systematically under-priced until the room saw a storm. The framework must price the stop explicitly (P4-B) — and the evidence says demonstration, not exhortation, is what re-prices it.

Feeds framework:B2 B5 C3

Read finding 2 in full →

Reversibility is worth 38 points; "never" only appears when the case is concrete

the Decision Rights Matrix's "Act (bounded)" column is really a reversibility column. Envelope anatomy (scope, magnitude caps, rate limits, reversibility window, owner, expiry) is what the community is actually pricing.

Feeds framework:B3 B5

Read finding 3 in full →

The community trusts records over promises: ledger shapes rate highest

wherever the framework must choose a mechanism, choose the recording mechanism: named, dated, signed, auditable. The community endorses accountability artefacts more strongly than access rights, oversight declarations, or process promises. (A consistent reading, offered as interpretation: W2's principle — strong content, no ledger clause — landed mid-pack at 3.76.)

Feeds framework:B2 C1

Read finding 4 in full →

The community set its first number: 7.5 machine-created work orders per planner per week

the framework's envelope section can ship a worked example with a community-sourced default rather than a blank. Status: draft parameter for validation in cycles (OQ2 in the open-questions register), not a recommendation yet.

Feeds framework:B5 D1

Read finding 5 in full →

Money is where values hide: weights, defaults and the signed objective

P3-A (only a named human sets and re-signs the weights) is evidenced not by what the room endorsed (3.73) but by what it did: refusing default weights 85%, rejecting the optimiser's silent trade 64%, and reporting that in most organisations the weights and the check have no named owner. The matrix rows for prioritisation/valuation must carry the named-accountability column as their load-bearing cell.

Feeds framework:B2 B3 C1

Read finding 6 in full →

The is/ought gap: rooms demand gates their organisations don't run

the framework's job is to close an is/ought gap the community itself measured. That is the "getting started → maturing" path's evidence base ([W5B] §7.1, §7.6): most member organisations are at the left edge (no auto-creation, no drilled stop, unwritten weights), while their practitioners already demand the right-edge controls. Both facts are in the data; the framework must serve both.

Feeds framework:A1 A2 C5

Read finding 7 in full →

The human core is stable across the lifecycle — and it sharpens under pressure

the framework's human-role set (define purpose · validate evidence · approve the decision · own the value judgement · accept & record residual risk · hold stop authority · override · interpret context · govern the knowledge · remain accountable by name) is fully evidenced across four rooms — with measured guidance on which rights need defending in the text (the ones rooms under-price cold: stop, risk acceptance, context) versus which defend themselves (approval, validation).

Feeds framework:B1 B2 B4

Read finding 8 in full →

Ratified 9–0, frozen on the day promised

Every section of Version 1.0 passed its last call unanimously. Two sections carry recorded dissent from earlier cycles — the risk-acceptance row's level and the eighth-principle question — shipped inside the text, not smoothed over. The freeze itself is audited: all 18 sections flipped to Frozen v1.0 on 31 August 2026, and each one's comment-disposition log published beside it.

The recorded dissent, in place →

The project ran itself under the framework it was writing

The AI drafted; humans judged and signed. The project's AI operated at Recommend/draft-only against a pre-registered boundary profile, every action and refusal went to an append-only public ledger — the first entry in the chain is the gate refusing an unauthenticated request — and the whole record is summarised in the framework's own provenance appendix (§D4). To our knowledge it is the first project to pre-register and publish per-function human–AI boundary levels with enforcement telemetry.

The public action ledger → · the method →

Experience it yourself

▶️

Replay the webinar series

The interactive simulations the rooms voted through — the work-order storm, the green health score, the optimiser’s silent trade — self-paced, nothing recorded.

🔒

The full working record — AMC members · volunteers

The cycle close-outs — minutes, call recordings, and every volunteer answer of record — are shared within the working group as consented: cycle 1 · cycle 2 · cycle 3. Named credit lives on the credit page.