TAS Research 01 — QA1 Obligations Extraction (Checkpoint 1)¶

Purpose. First increment of the TAS tile research pass (Legislation-to-Tile method, move one). Captures the settled extractions, sector testimony, primary-source exhibits, and margin register from the 2026-07-08 session. Everything here is pre-ratification input to tas-00-obligations; nothing here is canon until Tim ratifies the obligations artefact.

Method note. Extractions are product-blind: what does the standard oblige, on whom, evidenced how, and where is it silent. Product implications are quarantined in the Margin Register (§9) and structural-ruling candidates (§10). Sector testimony is Tim's operator experience unless marked otherwise; secondhand web/AI-sourced claims are marked and were independently verified where load-bearing.

1. Founding premises (session-settled, for doctrine note)¶

1.1 The layer doctrine¶

The regulatory stack has four layers with different legal force:

Layer 1 — the Act. NVETR Act 2011 (Cth). Creates ASQA and the registration scheme; delegates standard-making to the Minister. ASQA makes no law.
Layer 2 — the legislative instruments. Ministerial, Federal-Register-registered, disallowable: Outcome Standards (F2025L00354), Compliance Standards (F2025L00355), Credential Policy (incorporated by reference in 3.2, 3.3, 1.5). This is where obligations live.
Layer 3 — ASQA's administrative layer. Practice guides, self-assurance guidance, audit approach. Zero legal force; describes how the enforcer intends to exercise the discretion Layer 2 gives it. Revisable by ASQA at will.
(Layer 2.5 — explanatory statements. Departmental, authored alongside the instruments; the material a tribunal uses to resolve interpretation.)

Doctrine: obligations bind to Layer 2; interpretation is informed by the ES; enforcement posture is read from Layer 3; where the layers diverge, the divergence is surfaced explicitly, never silently resolved. Findings must ultimately cite the instrument — a product bound to instrument text cannot be outflanked by a guidance revision.

1.2 The interpretive gap as moat¶

Outcome-based drafting deliberately creates interpretive space ("effectively", "sufficient", "appropriate" — the ES says the adverbs are deliberate). ASQA's interpretation governs in practice (contest routes: reconsideration → ART → courts; rarely used). Anchoring to Layer 2 is necessary but not sufficient; the RTO's strongest position is a reasoned, documented, clause-traced interpretation with evidence behind it. Every RTO faces the same ambiguity; most resolve it by copying a template and hoping. Encoding a defensible interpretation is the product's legal mechanism for "the difference is the depth not the scope."

1.3 The TAS demotion¶

Under the 2015 Standards the TAS was named in the instrument (Layer 2): a mandated document, and therefore itself the compliance object. Under F2025L00354 no standard names a TAS; the obligations attach to outcomes (1.1–1.5 etc.) and the strategy document survives only as the conventional evidence form (Layer 3). The document is now testimony, not compliance. Divergence between the paper and the practice is precisely what draws findings. The 2015 regime rewarded the document's existence over its truth and produced the sector pathologies documented below; the 2025 instrument is structurally kinder to a live record of designed-baseline-plus-reasoned-variance than to any static fiction.

1.4 Historical preamble — the system's original architecture¶

Australian VET descends from the master–apprentice model: competence forms in the workplace; the educational layer exists to warrant breadth (the domestic-installer apprentice is examined on three-phase because the job never showed it to him). The certificate is fundamentally a legalistic/licensing document asserting capability across the occupational spectrum. Marketisation (1990s competition policy — National Training Reform Agenda, user choice, contestable funding; reductio: VET FEE-HELP 2012–16) created a provider class whose rational behaviour is certification-at-minimum-cost inside a system whose integrity assumptions were built for actors with no incentive to hollow the credential. ASQA is the load-bearing patch for that market-design flaw: it polices documents because it cannot police motive. The 2025 outcome standards are the current patch generation — a bet that demonstrated genuine capability is harder to counterfeit than documents.

Product statement of the era (doctrine note): the bet pays off only where evidence of reality becomes cheaper to produce than fiction. RTOpacks makes truth cheaper than fiction for the honest operator.

1.5 The workplace-formation spectrum (Tim's taxonomy)¶

Core: licensed trades (electrical, plumbing, building). Apprenticeship forms competence; classroom rounds breadth; occupational licensing independently backstops the credential. System works.
Middle: placement-mediated occupations (childcare, aged care, agriculture). Integrity rides on fragile third-party evidence chains (workplace supervisor observation, logbooks).
Emergent edge: desk-based quals (project management, bookkeeping, business). No placement, no licence, no workplace formation. The certificate is the entire assertion, resting solely on the RTO's assessment system. This is where private RTOs operate at scale — and where every pathology below concentrates. The corpus expanded because it could (provider economics), not because occupational logic demanded the credential form.

Consequence: ASQA's risk topography is this spectrum drawn as a heat map. RTOpacks' customer density is highest where the integrity gap is widest — the sentinel posture is the guard against becoming a fiction-polisher there.

1.6 The regulator's institutional imagination (Tim's diagnosis)¶

The instrument's authors betray a TAFE-shaped mental model — publicly funded, infrastructure-rich, cohort-based — while the economics the government created are private, lean, online-tending. Where an indicator seems naive about commercial reality (industry engagement as an unpriced ongoing activity; placement hosts as willing partners; simulated environments as things one simply has), the TAFE-shaped assumption is usually the explanation. Obligations are uniform; capacity to discharge them was never evenly distributed.

2. Standard 1.1 — Training (extraction)¶

Outcome (1): training is engaging, well-structured, enables attainment consistent with the training product. Bites on: the organisation throughout (no indicator attaches to an individual trainer; trainer obligations live in Part 3). Exists vs demonstrated: mandates no document; every indicator is a "demonstrates." But (a)–(c) are practically impossible to demonstrate without a designed, documented delivery structure existing before delivery — 1.1 is the clause-level location of the TAS demotion: it implies the strategy without naming it.

Indicator	Obligation (anchored)	Notes
1.1(2)(a) within F2025L00354	Training consistent with the training product as registered (packaging rules, prerequisites, unit requirements per National Register)	Hardest-edged indicator; near-binary; machine-checkable against TGA
1.1(2)(b)	The chosen mode enables attainment — an obligation to have cohort-aware reasons for the mode	ES factors for the interpretive words: cohort; mode/resources/technology/facilities; industry expectations; breadth and complexity
1.1(2)(c)	Structure and pacing supporting progression, with sufficient time for instruction, practice, feedback and assessment (all four named)	"Sufficient" is not a number (ES). A schedule allocating contact hours but no practice/feedback time fails the indicator as written
1.1(2)(d)	Techniques, activities, resources engage and support understanding	Softest indicator; almost purely demonstrative
1.1(2)(e)	Conditional — bites only where the product requires placement/community learning: skills must be attainable in that environment	Obligation about the placement's quality, held by the RTO though the environment belongs to someone else

Silences (2015 habits not to reimport): no mandated volume of learning / nominal hours; no "amount of training" clause; no named TAS; no mode-approval process. Building nominal hours as a compliance requirement under 1.1 is inventing law; the obligation is reasoned sufficiency.

Sector testimony — the three authors of pacing (confirmed operator evidence, corroborating web/AI-sourced material verified in tone against Tim's direct experience): 1. Compliance author (consultant/compliance manager): unit-level fiction, nominal hours divided by weeks to make the arithmetic look right for ASQA. Rarely defines a week, never a day. 2. Coordinator: the true practical author of the timeline — driven by logistics (rooms, trainer availability, placement host constraints), not pedagogy. 3. Trainer: the actual pacing, by improvisation, captured nowhere (scribbled session notes at best). The three never reconcile; the audit-facing document describes none of what happened. Placement finding: in placement-heavy products, pacing individualises per student (host availability, checkpoint passes), tracked in panicked emails and logbooks — where 1.1(2)(e) bites, cohort-level pacing is a fiction squared.

Mode testimony: mode was chosen commercially and justified afterwards, universally. Equipment is the true mode constraint, not pedagogy — no-equipment quals drift to Zoom-weekly + one-on-one simulated tasks + video upload (mainstream: Tim's own CQU experience); equipment-heavy quals fall to whoever owns infrastructure (historically TAFE; "simulated work environment" doctrine = real-but-not-operating, the TAFE kitchen not the restaurant). Hence 1.1(2)(b) is structurally entangled with 1.8 — a mode claim is only as good as the resource claim beneath it.

Regulatory context (verified): ASQA Strategic Review of Online Learning, final report Jan 2023 (with NCVER; predates the instrument). 62% of providers who shifted online during COVID indicated they'd employ more online/blended going forward; "immature risk assessment and self-assurance systems" is ASQA's own language; headline finding: quality VET can be delivered online, no single feature is a categorical risk — risk varies with product, cohort, trainer capability, tooling. The review is plausibly why 1.1(2)(b) is drafted as mode-neutral demonstrate-fit rather than mode-restrictive. Review = first-class research source; commissioned deep-dives on CHC33015 (Individual Support) and SIT30816 (Commercial Cookery) — exactly the placement-heavy and equipment-heavy archetypes.

Compliant shape (for model, not law): designed baseline carrying the RTO's sufficiency reasoning + delivery record capturing reality + divergence surfaced as information ("planned X, delivered Y, because Z"). Drift is not the sin; unexplained drift is. Drift reference = the RTO's own documented rationale (ES factors), never nominal hours.

3. Standard 1.2 — Industry engagement (extraction)¶

Outcome (1): engagement with industry/employer/community representatives effectively informs industry relevance.

Indicator	Obligation	Character
1.2(2)(a)	Demonstrates how it identifies representatives and seeks meaningful advice/feedback	Process-shaped — the only QA1 indicator drafted as "demonstrates how it… seeks." Documented outreach into a silent industry genuinely discharges (a)
1.2(2)(b)	Advice/feedback informs changes to training and assessment strategies and practices	Closed loop — must close on whatever advice arrives, however thin; reasoned no-change is valid closure (nothing obliges acting on advice, only using it)
1.2(2)(c)	Training reflects current industry practice	State-of-the-world claim — outreach alone cannot discharge it; evidenced cross-module (trainer industry currency per 3.2/3.3, product currency per TGA, sector intelligence). First cross-module obligation in the map: no single tile can own (c)

Sector testimony: engagement mostly fabricated or minuscule — the yearly friendly-employer letter, conference coffee-chat canvassing, loosely documented, grey; auditors focused on student outcomes rather than the loop. Structural truth — the circular loop: small RTOs choose scope from the skills they already have; the founder's industry network is their social network; trainer, industry expert, and scope decision are the same person and history. The instrument imagines outreach to a distinct external constituency that doesn't exist at small scale. Industry apathy is unpriced: there's no money in responding, so silence is the default ("a fantasy of TAFE land").

Interpretive anchor (ours, defensible): 1.2 at RTO level is localisation engagement, not product-validation engagement. The JSC layer (paid, contracted national consultation) answers what the national product contains; the RTO layer answers how this product is delivered here, now, for this cohort (electives, mode, local skills needs, resource currency — the practice guide's own examples). This reading makes the obligation tractable and survives scrutiny; the instrument's silence on the division of labour is the gap we fill with reason.

Sample-skew note: small-sample advice is only a risk if treated as mandate; disposition-tracked closure (adopted / adapted / rejected-with-reasons) is the best evidence of a functioning system — judgement, not compliance theatre.

4. Standard 1.3 — Assessment system (extraction)¶

Outcome (1): the assessment system is fit-for-purpose and consistent with the training product. Structure: three-step chain — conformance (a) → pre-use review against the 1.4 tests (b) → review outcomes flow into changes (c). The instrument's signature closed-loop drafting again.

Indicator	Obligation
1.3(2)(a)	Assessment consistent with the training product's requirements
1.3(2)(b)	Tools reviewed prior to use to ensure assessment can be conducted consistently with the principles of assessment and rules of evidence (Standard 1.4)
1.3(2)(c)	Review outcomes inform any necessary changes to tools

(b) implies a reviewer role distinct from tool-author (foreshadows 1.5's separation rules). The honest (b) review is substantially a contextualisation check — the practice guide lists contextualising tools to cohort and industry/work context as the compliance activity; "a bit generic" is what a (b) failure is called in a finding.

Sector testimony — the procurement pathway: - New entrants buy kits (vendors incl. Catapult, smallprint): sold per unit or as quals — core plus the vendor's on-file electives, not the package's full elective menu. - Small print shifts the obligation back: "up to you to contextualise; all care, no responsibility." Legally correct — 1.3 is non-delegable; the vendor claim is marketing, not evidence. - Most deliver out-of-the-box + bolted-in internet PDFs. The purchase substitutes for the (b) gate: faith in the vendor = the review. Systemic, sector-wide, invisible until audit. - The shared-unit tell: two quals sharing a core unit with identical un-contextualised assessment — the absence of any per-qualification review is visible on the face of the documents (e.g. events vs travel sharing a customer-service core). Sharpest single audit-intelligence item of the session. - Vendor-elective constraint contaminates three standards: electives delivered = vendor's catalogue, undermining 1.1(2)(a) packaging position and the 1.2 story simultaneously. - Not a small-RTO disease: CQU diploma of project management — woeful third-party purchased material, scanned handouts, uneditable fishbone-diagram PDFs, delivered via Moodle. - Contextualisation, when done: owner or trainer/assessor bringing their knowledge to "a structure that's already approved"; the mutated version never version-controlled against the vendor original (evidence: a 2017 kit of a superseded unit still in circulation → Section 14 transition hygiene and 1.3(a) conformance rot silently).

The transposition defence (the genre's legal theory, named): if the artefact restates the unit's own language, it cannot be non-conformant — hide behind the tree, be the tree, claim the tree's anchoring; coverage-as-camouflage, collapsing "fit-for-purpose" into "mentions the right words." Why it fails at Layer 2: restating the criterion satisfies (a) but is incapable of satisfying (b). The eight 1.4 tests all live between the words and the judgement: validity (principles) requires practical application components — a pasted PC has none; reliability requires comparable outcomes across assessors — a criterion with no benchmarks/evidence spec/marking guidance is maximally unreliable by construction; sufficiency (rules) requires specified adequacy — "e.g. references" specifies nothing. The genre's blankness is itself the non-conformance. The adversarial-auditor test: "walk me through how two of your assessors reach the same judgement from this tool" is unanswerable from a pasted bullet list. Tim's formulation (doctrine note, verbatim-worthy): an unsophisticated posture on something that could be so much better by the addition of sophistication — perceivable complexity, but not actually complex if you know what you're doing. The sophistication 1.3/1.4 demand is the interpretive layer between training product and judgement (contextualised scenarios, evidence specifications, benchmarks, marking guidance) — skipped for unit economics, not intellectual difficulty.

The benchmark defence and the platform lock (from Exhibit B): the e-learning generation answers reliability with "Answer guide for trainer" model responses — a real mechanism, sector-standard for the reliability principle, but the benchmark is the same transposition prose, so assessors converge reliably on a generic target (precision without accuracy; the parrot outscores the operationally-true answer). And the benchmark is uneditable in the platform (confirmed): contextualisation is architecturally foreclosed. Platform-locked non-compliance: 1.3(c) is structurally impossible for the buyer; the SaaS generation is a regression from the Word-document era (which at least permitted uncontrolled mutation). The delivery infrastructure, not the content, becomes the compliance defect — the fix cannot be better content in the same platforms.

Judgement mechanics observed: justification = resemblance-to-benchmark, recorded as a reply in a thread; the assessor's reasoning (why this response satisfies this criterion for this cohort) exists nowhere. Against 1.4(2)(b) the trail supports a weak validity claim and is silent on sufficiency, authenticity, currency.

5. Standard 1.4 — Principles and rules (extraction)¶

Two distinct structures, two lenses (ES: systems satisfy the principles; judgements satisfy the rules):

Principles of assessment — 1.4(2)(a), govern the system's design and use: - (i) fairness — accommodates needs, reasonable adjustments, reassessment - (ii) flexibility — appropriate to context, product, student; assesses relevant skills regardless of how/where acquired - (iii) validity — includes practical application components enabling demonstration in a practical setting - (iv) reliability — evidence interpreted consistently; outcomes comparable across assessors

Rules of evidence — 1.4(2)(b), govern each assessor judgement: - (i) validity — evidence adequate to reasonably assure possession of the skills/knowledge - (ii) sufficiency — quality, quantity, relevance enable an informed judgement - (iii) authenticity — assurance the evidence is the student's original, genuine work - (iv) currency — evidence demonstrates current skills and knowledge

Sector testimony — authenticity: field state = signed declaration + a timer nobody reads + difficulty-as-friction (god-awful Word templates as a security layer). Tim's CQU admission as datum: an honest, time-poor adult ran an entire diploma through GPT — GPT reading Moodle's rendered HTML and answering the multiple choice. If the rule can't survive an honest busy person, it has no chance against a motivated dishonest one. Detector arms race (AI writing / AI detecting / AI laundering) is unwinnable; ASQA cannot lean on it.

Session finding — authenticity is a property of the evidence type, not a bolt-on: a typed answer in an unproctored text box is unauthenticatable now, permanently and structurally. What remains authenticatable is evidence with the person in the evidence: observed performance, live questioning, video of the candidate performing, workplace supervisor attestation. The instrument already points here (validity principle demands practical application; authenticity rule demands assurance of genuine work). AI inverted the economics of trust: the typed answer is now worthless as competence evidence, and systems weighted toward it are invalid-by-drift though nothing in them changed.

Interpretive position (ours, novel, Layer-2-derived): post-2022, an assessment system weighted toward unproctored text responses cannot satisfy the authenticity rule; an honest 1.3(2)(b) pre-use review today must weigh evidence types by their authenticability. No guidance says this yet; it follows from instrument text applied to current facts — exactly the class of reasoned interpretation the doctrine encodes.

Demand-side corollary (Tim): most people don't want to game the system; the time-poor mechanic gaming text boxes responds rationally to content that respects neither his time nor his existing competence. Engaging, contextualised material (1.1(d)) and honest RPL (1.6) shrink what needs gaming at all.

6. Standard 1.5 — Validation (extraction)¶

Law: - Risk-based: which components of the assessment system and what sample size are determined by risk (1.5(2)(c)); the remembered fixed five-year ritual is 2015 folklore, not the 2025 obligation. - People rules, three tiers: CP §3A — at least one of the people undertaking validation (collectively) holds a validator credential (TAE Cert IV variants; assessor skill sets; diploma+ in adult/VET education). 1.5(2)(e)–(f) — persons who designed/delivered may participate but must not solely determine outcomes. 1.5(2)(d) — TAE products only: fully independent validation (external person, not employed/subcontracted, no interest), after first cohort completes. - Closing the loop: validation outcomes must inform changes (1.5(2)(g)); validation that doesn't drive change doesn't satisfy the standard. - Correction to sector folklore: "you can't validate within your own organisation" is the TAE special case generalised; internal validation is lawful for non-TAE products provided the author isn't the sole determiner. Whether auditor-style independence (Tim's ATO/ASIC superannuation analogy) spreads to all products is a horizon watch, not law.

Sector testimony: in small RTOs validation mostly didn't happen — low throughput, no credentialed second person, and the "Hey Jono" problem: the only credentialed peers are competitors; validation means handing a competitor the proprietary assessment system. The instrument prices external collegiality at zero; the market prices it as suicide. Observed substitute: pre-audit fabrication — a stack of recent assignments and invented findings written the week before audit. The 1.5 loop performed as theatre.

The engine-validator question, answered at Layer 2: the Credential Policy's operative words are "the person(s) conducting the validation" and its credentials are held only by humans — AI as validator-of-record is ruled out by construction (same construction that rules out machine assessors; no agentic-ruling policy needed). Nothing constrains the validator's instruments: engine as apparatus (risk-based sampling, rules-of-evidence audit of judgements, tool-to-product mapping checks, benchmark-drift flags, full analytical file preparation) with a credentialed human determining outcomes and signing is fully compliant — engine raises hands, never signs, applied to a new surface. Baseline reversal: the human baseline is not rigorous panels but fabricated notes; engine-prepared validation would exceed actual (not nominal) sector practice from day one.

7. Exhibits register (primary sources, session-captured)¶

Exhibit A — smallprint Australia trainer/assessor manual, BSBMKG507 (print era). PDF, 112 pp, created May 2017 (pdftk/3-Heights), watermarked to United Colleges of Australia. Unit now superseded; artefact still in circulation. Construction: sections open "This section addresses the following performance criteria:" + PCs pasted near-verbatim from TGA + generic prose written at each bullet ("a retail business, for example…"). Mapping matrix maps tasks to PC/PE/KE line items — coverage demonstrated, competence design absent. Authorisation page performs the liability shift in writing: smallprint's designers conducted internal validation; "The RTO must conduct their own validation and mapping to verify that the assessment tools used actually enable the collection of evidence that complies with the Principles of Assessment and the Rules of Evidence, can be used by different trainer/assessors, can be consistently applied…" — the 1.3(2)(b) obligation handed back to the buyer on the final pages. File: uploaded to session 2026-07-08.

Exhibit B — learnlogin e-learning unit, CHCDIV001 "Work with diverse people" (current era). ucca.learnlogin.com enrolment 907115, captured 2026-07-08 (page text + screenshots). Same construction method as Exhibit A: four topics = the unit's four elements verbatim; "In this topic you will learn to:" = Element 1's PCs verbatim; content = dictionary-definition prose with (wikipedia.org) cited as the source for the definition of "belief" — in nationally accredited material. Assessment: knowledge questions with "Answer guide for trainer" benchmark text (same transposition prose); benchmark uneditable by the RTO (confirmed) → platform-locked non-compliance. RPL variant ("Skills recognition" per element): PCs pasted as the candidate-facing ask + identical four-bullet evidence prompt on every element (previous training courses / workplace documents / unpaid work / references) + "No content found" — the transposition genre applied to recognition; the aftermarket charges most (RPL editions sell at premium) where the method fails hardest. Significance of A+B together: the genre is invariant across vendor, decade, and format; the electronic era changed the container, not the method. CHCDIV001 is core across dozens of CHC quals — the shared-unit tell at maximum scale, sold into all of them.

8. Cross-standard findings¶

1.2 evidence accretes into 1.5 inputs at unit level. Unit-pinned industry feedback (dated, attributed, disposition-tracked) pre-loads validation; the system grows more defensible over time — opposite of the folder-of-letters model.
1.1(2)(b) ⇄ 1.8. Mode capability is substantially a facilities question; a mode claim is only as good as the resource claim beneath it.
1.2(2)(c) is cross-module. Discharged jointly by trainer currency (People), product currency (TGA/KN), and the outreach record — no single tile owns it.
Procurement contaminates 1.1(a) + 1.2 + 1.3 at once (vendor-elective constraint).
1.4 authenticity re-weights 1.3(b). The pre-use review must now weigh evidence types by authenticability.
TAS ⇄ 1.6/1.7. A strategy honest about a product states how RPL is offered against it; the claim binds to the RPL tile's substrate (reference, never re-bind — the LLND discipline).
The instrument's drafting signature: closed loops everywhere (1.2(b), 1.3(c), 1.5(g)) — review/advice/validation without visible consequence fails; reasoned no-change is valid closure.

9. Margin register (product implications — parked, not law)¶

M1. Claims-surface TAS. The tile is a claims surface bound to substrate: trainer claims → People canDeliver state; assessment claims → Studio tools (version-true); validation claims → Record evidence. The printable TAS is a generated view that cannot say what the substrate contradicts. Two faces: auditor strategy view + trainer-facing "trainer delivery brief" (working label only, ours, not official — assigned units, current sequence, version-true resources/tools, own canDeliver state; first read = ledgered induction event serving 3.1).
M2. Teacher's desk test (Studio output contract). Delivery-ready artefacts a human uses that day are the terminal specification; Studio's pipeline is derived from the outputs, not vice versa. At least two first-class output families per unit: delivery kit + RPL kit, generated from the same anchored substrate, format-identical across the catalogue. Cohort-differentiated rendering (school-leaver / re-skiller / time-poor licensed tradie) is what generation makes affordable — the 1.1(d) engagement obligation and the demand-side authenticity fix.
M3. Contextualisation-by-construction. Qualification context as a generation input: the shared unit renders differently per qual from the same substrate — the shared-unit tell becomes structurally impossible. The 1.3(b) review becomes a first-class ledgered lifecycle event (tool → review against the eight 1.4 tests → disposition → release).
M4. Consultation-loop mechanism (1.2). Unit-anchored question generation (same engine read that builds T&A material; third output), artefact-family membership (questions travel with the course, regenerate on package version), RTO-triggered dispatch, video-first capture (60-second phone-video ask beats typeform gauntlets; authentic, faced, dated), disposition-tracked closure (adopted/adapted/rejected-with-reasons). Generation record itself discharges (a) almost incidentally. Caution: seeking advice, never outsourcing judgement. Pooled/shared engagement substrate (many RTOs, one ask) = network-effect thought for much later.
M5. Pacing substrate. Designed baseline (carrying sufficiency reasoning against ES factors) + delivery record + reasoned-variance surface; three personas (compliance author / coordinator / trainer) as three views over one substrate. Pacing cohort-level by default, student-level where placement individualises (1.1(2)(e)) — connects to shared candidate store / intake pipeline substrate.
M6. Authenticity weighting. Evidence types carry an authenticity weight as a first-class assessment-design property in Studio; delivery surfaces make person-present evidence (video, observation, live questioning, third-party attestation) as cheap as text boxes were. Evidence-triage frontier (machine screens uploads against the rules of evidence; assessor judges) is the answer to the authenticity crisis, not just a time-saver. Hard line: machine judgement is non-compliant by construction (credentialed humans hold the pen); design the triage rung, do not architect toward machine judgement.
M7. Engine-prepared validation (1.5). Dissolves the "Hey Jono" problem: no competitor sees anything; analytical heavy-lift by the system; one credentialed human determiner (not the sole author). Downstream: pool of credentialed validators working from engine-prepared files (Marketplace shape). Accreted 1.2 feedback is already in the file when the validator opens it.
M8. Conditional surfacing. 1.1(2)(e) obligations surface only when the product requires placement; third-party evidence chain (supervisor observation, RTO-issued forms, insurance/workers-comp letters for student-found placements) is Record-shaped.
M9. Marketing derivative (post-window). The cautionary tale: the objection to "just use ChatGPT" and the objection to the incumbent vendors are the same objection — unanchored content without provenance; the sector's paid product already has the disease people fear AI will bring. Opening image: the Wikipedia citation. Metaphor: the silk flower fools the eye until you ask it to grow. RPL-only enrolments as high-margin capability (revenue without delivery hours) — a selling line and eventual Marketplace shape.

10. Structural-ruling candidates (for tas-01-model — not yet argued)¶

R1. TAS: peer tile or connective tissue between Studio/People/Record? (Every cross-standard finding feeds this.)
R2. Accreting Review or not?
R3. Outcome source — what is the sealed artefact, if any?
R4. canDeliver gate — unlike LLND, 3.2/3.3 bite directly on strategy claims; presumption yes, to be argued.
R5. One artefact, two faces (auditor view / trainer delivery brief) — same substrate, two renders?
R6. Pacing resolution rule: cohort-level default, student-level where placement individualises.
R7. Where the consultation-loop mechanism lives (Studio substrate, 1.2 surface — connective tissue data point).

11. Session deliverables status (Fable window, closes Sunday)¶

D1 — obligations map: Divisions 1–2 (1.1–1.5) extracted with testimony, exhibits, law — this artefact. Remaining: 1.6/1.7 delta vs rpl arc; 1.8; the 3.1/3.2 workforce threads and 2.x information obligations a strategy touches.
D2 — structural rulings: candidates registered (§10); argument sessions pending.
D3 — doctrine note: premises settled (§1) + transposition-defence rebuttal + truth-cheaper-than-fiction line + Tim's sophistication formulation; drafting is post-checkpoint work, Claude solo, Tim redline.
Verified external source to ingest: ASQA Strategic Review of Online Learning final report (Jan 2023) + the two commissioned qual deep-dives.

Checkpoint ends. Nothing above is canon until tas-00-obligations ratifies.

title: TAS Research 01 — Obligations Extraction (Checkpoint 2) artefact_family: tas method_move: 1 (research) feeding 2 (obligations) status: CHECKPOINT — session capture, pre-ratification session: 2026-07-08/09 (Fable window, session 2) author: Claude (RTOpacks-side), from conversation with Tim instrument: F2025L00354 (Outcome Standards for NVR RTOs) companion_sources: F2025L00355 (Compliance Standards — verified for readiness provisions); NVETR Act 2011 registration gate (ss 16–17); ASQA General Direction — resourcing requirements; ASQA CEO statutory declaration (initial registration); National Code 2018 Std 6 (CRICOS sliver); NCVER private-provider research (2003) covers: 1.6/1.7 (delta vs ratified rpl arc); 1.8 + registration-gate mechanics; Part 3 strategy threads; 2.x strategy threads — D1 COMPLETE appends_to: tas-research-01 (Checkpoint 1 — founding premises, Standards 1.1–1.5) outstanding: D2 (structural-ruling arguments R1–R8); D3 (doctrine note); verification riders §13

TAS Research 01 — Obligations Extraction (Checkpoint 2)¶

Purpose. Second increment of the TAS tile research pass. Completes D1 (the obligations map): the 1.6/1.7 strategy delta against the ratified RPL arc, Standard 1.8 with its enforcement mechanics, the registration-gate findings, and the Part 3 / Part 2 threads a training strategy touches. Everything here is pre-ratification input to tas-00-obligations; nothing is canon until Tim ratifies the obligations artefact.

Method note. As Checkpoint 1: extractions are product-blind; product implications quarantined in the Margin Register (§11); testimony is Tim's operator experience unless marked; AI-/web-sourced claims are marked and independently verified where load-bearing.

0. Provenance declaration — the testimony vantage (method, first-class)¶

Sector testimony in this arc is given from a small-end operator's vantage, declared by Tim as pointed, pragmatic and non-bipartisan — honest from where it stands. The small end is the sector's numerical majority: NCVER's research on private providers found roughly half had fewer than 100 students, 84% had ten or fewer full-time staff, 60% ran on one to five (2003 data — structural picture only; current-year figure owed, §13). The record is not single-voiced — the CQU datum is top-end university VET, the Strategic Review and practice guides carry the regulator's voice, Exhibits A/B speak for the vendor layer — but TAFE-insider and large-private testimony is absent and is named here as an evidence gap, not silently absorbed. The vantage aligns with the product's customer density (Checkpoint 1 §1.5), so it is the relevant one; the declaration exists so that field-mechanics claims whose truth might vary with scale are read as calibrated, not universal.

1. Standards 1.6 / 1.7 — the strategy delta (vs the ratified RPL arc)¶

The RPL arc (rpl-00/01/02) owns the process: offer, evidence, judgement, record, mill posture. None of it re-derived. The strategy carries six claims the tile does not discharge alone:

Δ	Strategy claim	Anchor
Δ1	The product's recognition posture. Per-product licensing/regulatory constraints on CT (the 1.7(2)(b) parenthetical) and RPL are design-level facts; the strategy states "for this product, RPL via X, CT available/restricted because Y"	1.6(2)(a), 1.7(2)(a)–(b) / F2025L00354
Δ2	One system, not an annex. The s4 definition of assessment system names RPL inside it; a strategy describing RPL as a parallel scheme misdescribes the system the instrument defines. Consequence: the 1.3(2)(b) pre-use review and 1.5 validation perimeter reach RPL tools by definition — Exhibit B's "Skills recognition" screens fail the same review the delivery kit fails	s4 + 1.6(2)(b) / F2025L00354
Δ3	Validation scope claim. The strategy's validation section cannot draw its perimeter around ordinary assessment only (rpl-00 1.6-14 holds the RPL side)	s4 (validation) + 1.5 / F2025L00354
Δ4	Credit-adjusted pacing. A granted RPL/CT outcome individualises progression; the 1.1(2)(c) pacing baseline needs a stated rule for credit-adjusted entry. RPL/CT is the second individualiser of pacing after placement	1.1(2)(c) × Div 3 / F2025L00354
Δ5	The intake sensor. 1.6(1) targets students with prior skills; the instrument's designed sensor is the 2.2(2)(a) pre-enrolment review — review is the sensor, offer is the actuator; the strategy states how the product's intake connects them (seam into LLND's home standard)	1.6(1) × 2.2(2)(a) / F2025L00354
Δ6	Aggregate integrity. "Maintains the integrity of the training product" (1.6(2)(c)/1.7(2)(c)) is visible at product level as granting patterns; the strategy is where the pattern reconciles with the stated posture or indicts it	1.6(2)(c), 1.7(2)(c) / F2025L00354

Delta silences: no obligation to grant; no uptake quantum; no mandated RPL kit/form; CT's evidence class is closed (AQF certification documentation or authenticated VET transcript — nothing lesser), and 1.7(2)(b) carries no "in accordance with the assessment system" language, confirming rpl-00's CT-is-not-assessment reading from the strategy side.

Checkpoint 1 cross-finding 6 stands: recognition claims bind to the RPL tile's substrate — reference, never re-bind.

2. Standard 1.8 — Facilities, resources and equipment (extraction)¶

Outcome (1): facilities, resources and equipment for each training product are fit-for-purpose, safe, accessible and sufficient. Per-product framing; four adjectives each doing distinct work (fit-for-purpose = matches product requirements; safe = WHS echo; accessible = disability and availability senses both open; sufficient = quantity against cohort, un-numbered — same drafting habit as 1.1(2)(c)'s "sufficient time").

Indicator	Obligation (anchored)	Character
1.8(2)(a)	Demonstrates how it identifies what the product requires, incl. which parts come from third parties	Process-shaped ("demonstrates how", cf. 1.2(2)(a)); design-time enumeration
1.8(2)(b)(i)	Where provided by the org or third parties: "are, and will continue to be, suitable and safe"	Continuing-state obligation, not point-in-time; non-delegable ("how the organisation ensures"), same shape as 1.3
1.8(2)(b)(ii)	Students have access to what they need to participate	The access leg; bites hardest where the mode assumes resources the RTO doesn't control
1.8(2)(c)	Documented strategies and procedures to identify and manage risks in WIL / placements / community-based learning	Conditional on placement — and the only document mandate in QA1

The sharpest law finding: 1.8(2)(c) is the only indicator in all of QA1 that mandates a document. The instrument that demoted the TAS names exactly one required document in Part 1 — placement risk procedures. The drafters knew how to mandate a document when they wanted one; everywhere else the silence is deliberate. It also shows where the instrument thinks the danger is physical: a student on someone else's premises using someone else's equipment.

The identification floor is machine-readable. Units of competency carry Assessment Conditions specifying required resources, equipment and environment — 1.8(2)(a)'s identification has a TGA-anchored per-unit seed, exactly like 1.1(2)(a)'s packaging check. Identification below the assessment-conditions floor is non-conformance visible on the face of the documents; identification is the RTO reasoning up from the floor to cohort quantities and mode.

The three-legged resource base. Provision splits three ways, governed differently:

RTO-provided — 1.8(2)(b) assurance, full strength.
Third-party-provided — same 1.8(2)(b) assurance, non-delegable.
Student-provided — not governed by 1.8(2)(b); handled as disclosure under 2.1(2)(c)(iv) (obligations to acquire materials, equipment or IT).

Interpretive position (ours, Layer-2-derived). When a mode assumes student-provided equipment and the assumption fails, 1.8(2)(b)(ii)'s access obligation does not vanish: a mode built on student-owned resources needs either a designed fallback or a per-student test of the assumption at intake — and 2.2(2)(a) already names digital literacy in the suitability review, so equipment access is the natural adjacent question. No guidance says this; it follows from instrument text. Same class of reasoned interpretation as the Checkpoint 1 authenticity position.

Entanglements. Cross-finding 2 sharpens to a three-way entanglement: 1.1(2)(b) ⇄ 1.8 ⇄ 1.4(2)(a)(iii) — the validity principle's "practical setting" is a 1.8 resource claim; the simulated-work-environment doctrine anchors at this junction; a mode claim is only as strong as the resource claim beneath it and the validity claim above it. Separately, the placement conditional surfaces in three standards — 1.1(2)(e) attainability, 1.8(2)(c) risk documentation, 2.1(2)(c)(iv) placement disclosure (sharpens M8).

Silences: no ownership requirement (lease/hire/borrow all fine); no physical-campus requirement — drafting is mode-neutral, so for online products the "facility" is the platform and (b)(i)'s continuing-suitability duty applies to the LMS as to a workshop (a light that makes Exhibit B's platform-locked defects a facility defect); no ratios; no inspection cadence; no equipment register mandated.

3. Standard 1.8 — field mechanics (sector testimony, small-end vantage)¶

Enforcement is front-loaded. 1.8 is hot at initial registration and scope change, background thereafter. Layer-2 anchor for why: s5(2) requires the regulator to consider whether an applicant "is complying, or will comply" — a pre-delivery RTO's resource base is nearly its only inspectable reality, so the audit surface follows the available evidence.
Identification is audited as list-existence; assurance is presumed from it. The equipment checklist (derived from Assessment Conditions — pot sizes, boards, knives) demonstrates (a); (b) is then presumed from the identification's quality ("you've listed it so obviously you thought about it") — nobody counts knives. The continuing-state clause in (b)(i) is essentially never audited as written: point-in-time inspection against a continuing-state obligation. Divergence surfaced, not resolved.
The evidence ladder. Walkthrough → photos → hire arrangement → letter from a facility owner (verification depth indeterminate — the auditor might phone the restaurateur). Beneath all: the commercial-premises presumption — a classroom in a commercial building draws a glance at fire exits because building class imports a baseline safety standard; (b)(i) "safe" discharged by premises category.
Inspection capability is a gradient that mirrors the workplace-formation spectrum. Kitchens get looked at; a tourism GDS shown on a screen is uninspectable by a non-domain auditor. Scrutiny is weakest exactly where the sector concentrates — the pathology pools where the auditor cannot see (consistent with every standard so far).
The third-party leg runs on an insurance chain — written agreement → RTO's own public liability policy → certificate of currency listing every delivery location → locations on the public record. The certificate expires and renews annually, making it the only artefact in 1.8 practice with a clock; the continuing-assurance obligation is discharged, to the extent it is at all, by an insurance cycle that exists for liability reasons. (Feeds M10.)
Simulated environment is a fidelity-scaling doctrine. The bar tracks the occupational infrastructure intensity: mechanics = garage with hoist; travel agency = three desks, posters, brochures. Assessment Conditions set the floor, the occupation sets the ceiling. Corollary: at the desk-based edge simulation is nearly free to stage — part of why private scope drifts there.
The danger-zone doctrine (regulatory adjacency). ASQA discharges "safe and suitable" by importing whichever adjacent regulatory regime natively owns the setting — building class for classrooms, health department for kitchens, by extension electrical/WHS regimes. A simulated environment in a danger-zone occupation is held to the real setting's regime (the home kitchen claiming simulated-commercial status invites the health-department question). Where no adjacent regime covers the setting, suspicion is the default. Fidelity isn't aesthetic — it's regulatory adjacency.

4. The registration gate — day-one-ready, enforcement by attestation, gate/road¶

Day-one-ready is real, and codified — but not in Layer 2. The test ("if I approve you today, could you open the door tomorrow") lives in: the NVETR Act's registration gate (s16 approved-form application; s17 discretion on "complies or will comply"); ASQA's General Direction — resourcing requirements for applicants (a legally enforceable ASQA-made instrument — a layer between the ministerial instruments and the practice guides that the layer doctrine must name: the one place ASQA makes binding rules rather than describing posture); and the CEO statutory declaration (under the Statutory Declarations Act 1959) attesting readiness. Verified: F2025L00355 carries no readiness provision (its Parts: information/transparency, training-product integrity, accountability; Schedules: fit-and-proper, logo, repeals). Grok-sourced input on this point recorded as AI-sourced, partially verified: the stat dec's existence and currency confirmed independently; its framing error ("clear legal requirement") not imported — no Layer-2 provision mandates dormant-operational readiness; the readiness test is ASQA's administered exercise of the Act's discretion, made personally binding by attestation.

Enforcement by attestation (doctrine-note category). The regulator cannot inspect readiness into existence, so it converts the gap into personal criminal exposure on a named human: the false declaration, not the unreadiness, is the offence. The same institutional move as document-policing (Checkpoint 1 §1.4), taken to its endpoint — where even the document can't be verified, make the signature the liability. It sorts operators by their relationship to the truth of their own paperwork. The honest applicant's holding costs (18+ months dormant-operational, Tim's case) are the price of a true signature; the dishonest applicant signs anyway and prices the risk.

Dated on primary evidence: the stat dec was in the gate by January 2017 (Exhibit C — "UCCA Statutory declaration by CEO.pdf" in INTVET0001795). The ~2019 "increased scrutiny" push tightened an existing attestation regime; it did not introduce it. (Prior hypothesis that it was post-2019 — refuted.)

Gate/road, proven on one RTO's whole record: two regulatory activities in ~9.5 years of registration, both inside the first 30 months (initial audit 7/08/2017; post-initial 10/09/2019); nothing since — through COVID, the online explosion, and the entire 2025 standards transition. The front-loaded profile generalises across standards (1.8 §3.1 above; 3.1(a) §8 below). Post-2019 absence is also the self-assurance era visible as absence: risk-triggered engagement watching data exhaust (AVETMISS, complaints, tip-offs), not premises. Consequences: the pre-audit fabrication ritual (Checkpoint 1 §6) assumed a scheduled audit — under risk-triggered engagement the theatre must be maintained permanently or the record must be true; and a mature RTO's practical audit surface is its data signature, not its kitchen.

Gate economics: day-one-ready + front-loaded 1.8 = the capital barrier is entirely at the gate, near-zero on the road. The holdable scope while awaiting audit is the scope with nothing to hold — a second, independent driver of desk-based drift (alongside inspection-gradient weakness, §3.4).

5. The mode genealogy (how online entered the sector)¶

2017: initial registration, online not contemplated ("they didn't really know you could teach it online"). 2019: online added not as a mode but as another delivery location on an existing scope — riding approval already granted for face-to-face. Online entered the sector through location-amendment laundering and was never gate-checked as a mode by anyone. COVID then killed the classroom and universalised the never-audited delivery form overnight. The 2023 Strategic Review is the regulator studying, after the fact, the thing it never gated; 1.1(2)(b)'s mode-neutral demonstrate-fit drafting is the 2025 instrument catching up — the first time the mode question is asked as law rather than smuggled through a location field. The instrument's answer was not to make auditors inspect LMSs but to make the RTO carry the demonstration burden — which loops to 1.8's "will continue to be": a gate that checks the start state, a road nobody patrols, and now a clause drafted against exactly that gap.

The "online was a dirty word" datum is pinned: AUDREC0009779, post-initial audit, 10/09/2019 — 2015-standards era, pre-review. Historical posture, not current divergence. (The 2019-era auditors declined to inspect the one facility of an online product — the platform generation's defects sit in precisely the facility field practice refused to walk through.)

6. Exhibit C — asqanet registration record (primary source, session-captured)¶

Exhibit C — asqanet application history + regulatory activities, United Central Colleges of Australia Pty Ltd. Two captures, 2026-07-09 (screenshots; underlying documents not retrievable — the fingerprint is the datum).

Application history: initial registration INTVET0001795 submitted 30/01/2017, approved. Document list includes: Application for initial RTO registration; Financial_Viability_Risk_Assessment_Pack; UCCA Fit and Proper Person; UCCA Statutory declaration by CEO.pdf — the full three-legged gate (money, character, sworn readiness) assembled pre-2025. Also: three trading-college registrations (Hospitality Training College of Australia; Integrated Marketing College of Australia; tourism&travelcollege australia) — corroborating the testimony scope mix (the kitchen, the BSBMKG unit of Exhibit A, the GDS on a screen). Minor changes 2018 ×5, 2024 ×1.
Regulatory activities list: AUDREC0007370 "Application - Initial", closed complete 7/08/2017; AUDREC0009779 "Post Initial", closed complete 10/09/2019. Nothing since.

Fence qualification (for the record): everything in Exhibit C is United Central Colleges of Australia Pty Ltd — home side, the RTOpacks operator. "UCCA" in the 2017 filenames predates and is unrelated to UCCA Inc.

Significance: primary-source proof of (a) enforcement-by-attestation in the gate by Jan 2017; (b) the gate/road audit cadence (§4); (c) the era genealogy dates (§5).

7. ⚠️ Corpus redline — people-00-obligations §3.1 anchors to phantom PIs¶

The authorised Standard 3.1 has two performance indicators: (a) "how it ensures the number of trainers, assessors and other staff are appropriate for the delivery of the services it offers"; (b) CPD access. people-00's §3.1 table cites four PIs (workforce planning / induction-PD-performance-management / currency support / records per trainer) and a different outcome sentence — policy-draft wording, the same defective-corpus disease named in the STANDARDS-ANCHOR-BLAST-RADIUS-01 supersession note. The 2026-07-07 correction fixed the 3.2/3.3 inversion but left the §3.1 rows uncorrected. PIs (c) and (d) do not exist; the authorised instrument contains no staff-induction obligation and no records-keeping PI anywhere (the word "induction" appears once, in the s4 services definition, meaning student induction). The rows' substance mostly survives under other anchors (records-of-credentials is implicit in demonstrating 3.2/3.3; possibly explicit in F2025L00355's Accountability division — unverified), but the anchors as written cite law that isn't there. Redline owed to the People arc — blast-radius family, needs filing; not TAS scope to fix, but TAS does not consume the rows as they stand.

Downstream amendment — M1 re-anchored. M1's "first read = ledgered induction event serving 3.1" lost its anchor. Re-anchored, arguably better: 4.2(a) (staff supported to understand the instrument components relevant to their role) + 4.2(d) (roles and responsibilities documented, accountable decision-making). A ledgered first-read of the trainer delivery brief is close to a literal discharge of 4.2(a) at trainer level.

8. Part 3 — the strategy threads (3.1, 3.2/3.3 delta vs the People arc)¶

Standard 3.1 (authorised text)¶

3.1(a) is a cross-module obligation in the same class as 1.2(2)(c): staffing adequacy = People roster × strategy load model. People knows who exists and what they can deliver; only the strategy knows the load — cohort × mode × pacing × product mix. The four 1.1(2)(c) activities (instruction, practice, feedback, assessment) are literally the demand function; a pacing schedule allocating feedback time nobody is rostered to give fails 3.1(a) and 1.1(2)(c) simultaneously. Neither tile can demonstrate 3.1(a) alone.
Perimeter split: the legal staffing scope is services (defined term — includes support and recruitment/enrolment staff); the strategy carries only the T&A staffing story. Noted so the tile doesn't over-claim.
Silences: no ratios, no named roles, no induction, no performance management, no record-keeping PI. The number is not a number; it's an argument.

Field mechanics (testimony): 3.1(a) operates as the auditor's risk-mitigation backdoor at the gate — scope-throttling ("we'll approve you initially for four"), applied even against a founder holding every qualification personally. No formula exists; capital dissolves the objection. The gate/road pattern generalises (§4). Scale finding, extending Checkpoint 1 §1.6: the instrument is written to the top end of town — obligations drafted for organisations with org charts, enforced against sole-trainer RTOs through discretionary backdoors. The honest small-RTO answer to 3.1(a) is reasoned sufficiency: "me, all hats, and here's why that carries this cohort volume."

Standards 3.2/3.3 — strategy delta (People owns the substance)¶

Δ	Strategy claim	Anchor
Δ1	Assignments are claims. A strategy scheduling a named person into delivery asserts in writing that their People state supports it (credential for activity, unit-level vocational competency, currency). A strategy contradicted by its own trainer register is divergence-draws-findings at its sharpest. Feeds R4 (canDeliver gate — presumption yes, hardened)	3.2(2)(a) / F2025L00354
Δ2	Under-direction is a designed feature. Where delivery relies on a working-towards trainer, the direction structure (who directs; judgement-block on assessment staffing) is part of the delivery design the strategy describes	3.2(2)(b) / F2025L00354
Δ3	Expert engagement is strategy-shaped by construction. The justification the law demands — "by reference to the requirements of the training product or the specific VET student cohort … in response to a specific need" — is a strategy statement; People holds the engagement record that discharges it	3.3(2)(b) / F2025L00354
Δ4	The 1.2(2)(c) loop closes through here. Trainer industry currency (3.3(a)(ii), People-held) is one leg of "training reflects current industry practice" — reference, never re-bind	3.3(a)(ii) × 1.2(2)(c)

Field mechanics (testimony):

Trainer non-fungibility is the structural fact under all of Part 3. VET delivery capability is person × product (credential + unit-level vocational competency + industry currency) — no relief-teacher substitution exists. Hiring follows enrolment load, never anticipates it ("staff are expensive"): casuals absorbed into existing load, shadowing, informal, unrecorded.
Two induction modes, both evidenceable: the load-absorption casual (delivery brief as the induction artefact — M1) and the expansion co-developer (marked as Tim's theorisation, credible: the trainer hired to grow the business is hired to co-write, because writing the tasks is the buy-in — authorship is the induction, and in a Studio-shaped world that contact is itself ledgered 4.2(a) evidence).
The signature is the source of the authority (postscript). The 1.4(2)(b) judgement is personally the trainer's; authority accretes upstream from the person who signs. A trainer arrives holding domain expertise and system literacy (the TAE is partly a credential in knowing what the system looks for) and will reshape content toward better outcomes whether or not authorised. Generalises the three-authors finding from pacing to content: the trainer is the actual author of delivered reality; a system denying first-class authorship doesn't prevent the authorship — it loses the record of it. (M2's governance rationale.)
Support staff: domestically, no distinct role — the trainer absorbs it (2.3's access claims at small scale are one person's calendar). The stratum where support staffing gets real is CRICOS — verified: National Code 2018 Standard 6 requires "sufficient student support personnel, in addition to academic staff" (no numeric ratio in the Code; providers self-set, e.g. 1:100 observed), plus a designated 24-hour point of contact and a documented critical-incident policy (records 2 years past enrolment). Same reasoned-sufficiency drafting, Home Affairs teeth behind it. Out of TAS-arc and launch scope; parked as a Landscape note.
Experts: the 3.3(b)–(c) construct has not been observed in the wild (small-end vantage). Field reality is guest speakers — "look who we've got today," the monthly talk — never woven into delivery under direction with assess-alongside discipline. Interpretive line (ours): the instrument never says when a guest becomes an "expert delivering training"; the defensible line is one-off enrichment outside the assessed delivery (below 3.3's threshold) vs an SME scheduled into the delivery of unit content (inside it, 3.3(c) apparatus attaches). The strategy is where the line is drawn per product — it can make the monthly guest lawful by how it describes them. Interpretive-gap-as-moat in miniature.

9. Part 2 — the information obligations a strategy touches¶

2.1 — the strategy is the source of truth for the public information surface. The 2.1(2)(c)(i) accessible-information list (code, title, duration, modes, location, commencement dates, scheduling, requirements to commence/complete incl. assessment requirements, licensing implications, third-party arrangements) is nearly a rendered strategy summary. 2.1(2)(a)'s "clear, accurate and current" makes marketing-vs-strategy divergence a compliance breach in its own right — the website promising a duration the delivery design doesn't support fails 2.1 before anyone enrols. (Feeds M13.)
2.1(2)(e) — change propagation. Students informed "as soon as practicable" of changes incl. transition of superseded products — Exhibit A's stale-kit problem surfaces as a student communication duty, not just catalogue hygiene. A product version event has a comms leg.
2.2 — the enforcement mechanism of the strategy's cohort premise. Every mode-fit and pacing argument (1.1(2)(b)/(c), ES factors) assumes a cohort; the 2.2(2)(a) pre-enrolment review (skills, competencies, LLN, digital literacy) is the per-student gate keeping the assumption true; 2.2(2)(b) advice is the load-bearing discharge (LLND arc, consumed not re-derived). Strategy delta: the strategy must state its cohort premise concretely enough that the 2.2 review has something to test against. A strategy with no stated cohort makes the suitability review a form with no question.
2.3 — access claims are staffing claims, per mode. 2.3(2)(c) (students told how/when trainers are reachable) + 2.3(2)(d) (timely response) = the access model the strategy designs per mode; loops to the 3.1(a) load model — promising access nobody is rostered to provide fails 2.3 and 3.1 together.
2.6 — one strategy-adjacent sliver. Wellbeing needs identified "by reference to the training product content" (2.6(2)(a)) — the product itself drives the flag (trauma-adjacent CHC content the obvious case); design-time act, strategy-side; support arrangements operational.
Excluded from the strategy delta: 2.4, 2.5, 2.7, 2.8 — per-student or organisation-level; no strategy claim to carry (the assessment design's capacity for reasonable adjustment is already inside 1.4's fairness principle).

10. Cross-standard findings (additions — continue Checkpoint 1 §8)¶

The gate/road enforcement profile generalises. 1.8, 3.1(a), and the registration gate itself all bite at registration/scope-change and go dormant on the road (proven on Exhibit C's audit cadence). The 2025 instrument's continuing-state clauses (1.8(2)(b)(i)) read as drafted against exactly this gap.
Three-way entanglement: 1.1(2)(b) ⇄ 1.8 ⇄ 1.4(2)(a)(iii). Mode claim ← resource claim ← validity's practical setting. The simulated-environment doctrine anchors at the junction; fidelity is regulatory adjacency, not aesthetics.
The placement conditional surfaces in three standards — 1.1(2)(e) attainability, 1.8(2)(c) risk documentation (QA1's only document mandate), 2.1(2)(c)(iv) disclosure. One conditional, three obligations (sharpens M8).
Staffing adequacy = People roster × strategy load model (3.1(a) × 1.1(2)(c) × 2.3). No single tile can demonstrate it; the 1.1(2)(c) activities are the demand function.
The strategy is the source of truth for the public information surface (2.1(2)(c)(i)) — the information obligations make the TAS three-faced (auditor / trainer / student-public renders of one substrate).
Recognition is inside the assessment system by definition (s4) — the 1.3(2)(b) review and 1.5 validation perimeter reach RPL tools; the transposition rebuttal transfers to RPL kits without modification.

11. Margin register (additions and amendments — product implications, parked, not law)¶

M1 (amended). Trainer delivery brief: induction anchor corrected 3.1 → 4.2(a)/(d). Two induction modes: load-absorption casual (first-read of the brief, ledgered) and expansion co-developer (authorship as induction — the Studio collaboration record is itself the 4.2(a) evidence).
M2 (amended). Teacher's-desk test gains a governance rationale: the trainer reshapes delivered reality regardless; capture the reshaping as first-class authorship or lose the record of it.
M10 (new). Resource-register substrate. Per-product resource requirement seeded from unit Assessment Conditions (TGA/KN read); three-legged provision split (RTO / third-party / student-disclosed); continuing-assurance artefacts are Record-shaped expiring evidence with clocks (certificates of currency listing delivery locations — the People expiring-credential pattern applied to premises); feeds the strategy's mode reasoning (1.1(2)(b)) and the 2.1(2)(c)(iv) student-obligations disclosure.
M11 (new). Credit-adjusted pacing. RPL/CT outcomes write into the M5 pacing substrate as individualiser #2 (after placement). The designed baseline carries a credit-adjustment rule.
M12 (new). Unified intake moment. The 2.2 review, LLND screen, and 1.6/1.7 recognition offer are one intake moment over the shared candidate/intake spine (live as of 2026-07-08). Sensor (review) → actuators (advice; offer).
M13 (new). The third face. Student/public product-information render (2.1(2)(c)(i) list) generated from the same strategy substrate as the auditor view and trainer delivery brief — three renders, one substrate, structurally incapable of disagreeing. The shared-unit-tell logic applied to the information surface. Change events (2.1(2)(e)) get a comms leg.
M14 (new). Guest-speaker line. The strategy's per-product description of visiting SMEs is what places them inside or below the 3.3 threshold — a worked example of encoding a defensible interpretation.
M9 rider. The RPL-mill aggregate-signal point (Δ6) gives M9's marketing derivative a compliance-side twin: granting-pattern visibility as a trust feature.

12. Structural-ruling candidates (additions — continue Checkpoint 1 §10)¶

R4 (hardened, not yet argued). canDeliver gate: 3.2/3.3 bite on strategy claims directly — the strategy's person-product assignments are verifiable claims against People state. Presumption yes, now with Δ1's mechanism.
R8 (new). The strategy's recognition-posture section as a reference-render of RPL tile state (Δ1–Δ6; cross-finding 6 discipline).
R9 (new). Where resource claims live: M10's resource register — new substrate, Record-domain evidence, or strategy-tile-domain? (The insurance-chain clocks argue for Record-shaped custody; the Assessment-Conditions seed argues for a Studio/KN read. Argue at D2.)

13. Source dispositions and verification riders¶

Dispositions this session: - Grok document (RTO operational readiness) — AI-sourced; partially verified (stat dec existence/currency confirmed; F2025L00355 absence confirmed); framing error ("clear legal requirement") identified and not imported; General Direction omission noted. Used as scout, not source. - NCVER private-provider research — 2003 data; structural picture adopted, marked dated. - National Code 2018 Std 6 — verified via provider policy documents quoting the Code (asqa/deewr primary text not fetched); ratio absence and 24-hr contact confirmed across two independent sources.

Riders owed (carried to next session): 1. Chrome errand (single batch, asqa.gov.au blocks server fetch): current CEO stat dec wording · Guide to initial RTO registration v1.6 (Dec 2025) readiness language · resourcing General Direction post-1-July-2025 status. 2. people-00 §3.1 phantom-anchor redline — file to the People arc (blast-radius family). 3. Current RTO size-distribution figure (ASQA annual report or NCVER DataBuilder) to replace the 2003 structural citation. 4. ASQA Strategic Review of Online Learning + two commissioned deep-dives — ingest to project knowledge (carried from Checkpoint 1). 5. F2025L00355 Accountability division — check for the records-keeping obligation displaced from phantom 3.1 PI (d).

14. Session deliverables status (Fable window)¶

D1 — obligations map: COMPLETE. 1.1–1.5 (Checkpoint 1) + 1.6/1.7 delta, 1.8, Part 3 threads, 2.x threads (this checkpoint). All with law, testimony, and exhibits.
D2 — structural rulings: candidates now R1–R9; argument sessions next.
D3 — doctrine note: premises enriched this session — enforcement-by-attestation; the gate/road profile and self-assurance-as-absence; the mode genealogy (location laundering → COVID → review → 1.1(2)(b)); the General-Direction layer (ASQA's one binding-rule instrument class); written-to-the-top-end scale finding; trainer authorship/signature authority; the provenance-declaration method itself. Claude drafts solo, Tim redlines.

Checkpoint 2 ends. Nothing above is canon until tas-00-obligations ratifies.

title: TAS Research 01 — Structural Rulings Argued (Checkpoint 3) artefact_family: tas method_move: 3 (structural rulings) feeding tas-01-model status: CHECKPOINT — session capture; verdicts settled in-session (Tim concurrence 2026-07-09), canon only when tas-01-model ratifies session: 2026-07-09 (Fable window, session 3) author: Claude (RTOpacks-side), from argument with Tim instrument: F2025L00354 (Outcome Standards for NVR RTOs) appends_to: tas-research-01 (Checkpoint 1 — premises, 1.1–1.5; Checkpoint 2 — 1.6/1.7 delta, 1.8, gate, Parts 2–3) covers: D2 — structural-ruling candidates R1–R9, argued to verdict outstanding: D3 (doctrine note — unblocked, Claude solo draft); riders per Checkpoint 2 §13; FILING-03 owed once FILING-01/02 close

TAS Research 01 — Structural Rulings Argued (Checkpoint 3)¶

Purpose. Third increment of the TAS tile research pass; completes D2. The nine structural-ruling candidates registered at Checkpoints 1 §10 and 2 §12 are argued to verdict. R1 is the keystone; R2–R9 are argued as its cascade. Three rulings (R3, R4, R9) carry in-session refinements from Tim that materially improved the verdicts and are recorded as first-class. Everything here feeds tas-01-model; nothing is canon until that artefact ratifies.

Method note. Each verdict is argued strongest-counterargument-first where a real counterargument exists. Anchors cite the instrument; product mechanics cite the margin register (Checkpoints 1 §9, 2 §11).

R1 — Peer tile or connective tissue? Verdict: peer tile with a reference-only claim discipline. (Keystone)¶

The counterargument (pure connective tissue), stated at full strength. The instrument demoted the TAS: no standard names it; the document is testimony (Checkpoint 1 §1.3). The 2015 regime's central pathology was the document as compliance object. A "TAS tile" that stores its own copies of trainer assignments, tool lists and recognition posture rebuilds the fetish in digital costume — a second source of truth that diverges from substrate exactly as the paper TAS diverged from practice. On this view the TAS is a generated report over Studio/People/Record: a query, not a thing.

Why it fails: the demotion demoted the document, not the design. Strip away every claim the strategy makes about other tiles' domains and a residue remains that no existing module holds:

The cohort premise — which 2.2(2)(a) needs concretely stated or the suitability review is a form with no question (Checkpoint 2 §9).
Mode reasoning — 1.1(2)(b) is an obligation to have cohort-aware reasons, argued against the ES factors; reasons are authored content, not derivable from any substrate.
The pacing baseline with sufficiency reasoning — 1.1(2)(c)'s four named activities, plus the individualiser rules (placement, credit — M11).
The load model — cross-finding 11: only the strategy knows cohort × mode × pacing × product mix; the 1.1(2)(c) activities are the demand function 3.1(a) is tested against.
The access model per mode — 2.3(2)(c)–(d).
The authored interpretive lines — the guest-speaker line (M14), the 1.2 localisation reading, the student-equipment fallback position. The interpretive-gap-as-moat premise (Checkpoint 1 §1.2) makes reasoned, documented, clause-traced interpretation the product's legal mechanism; that reasoning needs a home with a lifecycle — drafts, ratification, version history when the reasoning changes. Queries don't have lifecycles.

The state-axis argument. Studio's unit of state is product content; People's is the person; Record's is the evidence item; RPL's is the recognition process. The strategy's unit of state is product × cohort × mode as delivered by this RTO — a distinct axis with a distinct lifecycle (it versions when a local decision changes, not when the training package does). Distinct axis, distinct lifecycle, original state: a tile.

Why it is not a plain peer tile: the claim discipline. Most of what a strategy asserts is grounded in other tiles' state — assignments in People canDeliver, tools in Studio versions, validation in Record evidence, recognition posture in RPL state. The discipline that makes the tile 2025-native rather than 2015-in-costume is cross-finding 6 generalised: every claim about another tile's domain is a live reference, never a copy. The tile owns assertions and reasons; it references grounds. M1's law then holds by construction — the printable TAS is a generated view that cannot say what the substrate contradicts, because it holds no copy to contradict with.

Formula: the strategy tile is the author of reasons and composer of claims — never a custodian of facts held elsewhere.

R2 — Accreting review? Verdict: yes.¶

The instrument's drafting signature is closed loops (cross-finding 7): 1.2(b) advice → change; 1.5(g) validation → change; the variance surface → revision. All three loops need somewhere to close. An accreting review register on the strategy tile — industry-advice dispositions (adopted/adapted/rejected-with-reasons), validation outcomes, variance patterns, intake-sensor signals — that periodically closes into a new baseline version is the 2025-native shape. R2 also supplies the tile's lifecycle: baseline vN → live record + accreting inputs → reasoned revision → vN+1. Without it the tile is a static document with better plumbing; with it, drift-is-information becomes operational.

R3 — The sealed artefact. Verdict: sealed baselines, timestamped renders. Signer named by the RTO.¶

The law mandates no document, but the divergence-as-information doctrine requires a pinned baseline — "planned X, delivered Y, because Z" is incoherent unless X is citable at a version. The sealed artefact is the designed baseline including its reasoning, ratified and versioned. Renders are never sealed — they are generated, pinned to (baseline version, substrate read time), and therefore reproducible as at a date. That reproducibility is what makes testimony defensible testimony.

Settled refinement (Tim, in-session): the RTO names its signer; the product records the naming. Ratification authority is not a fixed product role — a TAFE will name a department head, a two-person RTO names the director who is also trainer and author. The same mechanism sizes across the whole organisational spectrum with no configuration beyond naming a person. Bonus: the act of naming the signer is itself 4.2(d) evidence (roles and responsibilities documented, accountable decision-making) — the product generates compliance evidence from the accommodation.

R4 — canDeliver gate. Verdict: yes — ratify-block with the instrument's own exception path. The keyhole doctrine.¶

Δ1 (Checkpoint 2 §8) hardened the presumption: a strategy scheduling a named person asserts in writing that their People state supports it. Mechanism under R1: assignments are live references to People canDeliver, so the render shows assignment status truthfully and cannot assert "supported" when People says otherwise. But the gate is not a data-entry prohibition — the trainer-authorship finding generalises: denying the record of a real assignment doesn't prevent the assignment, it loses the record. So: reality is recordable; the contradiction is surfaced loudly; and a baseline cannot ratify with an unsupported assignment except through an explicit documented exception — which is exactly 3.2(2)(b): the under-direction design, with named director and assessment judgement-block. The gate's escape hatch is the instrument's own.

Settled refinement (Tim, in-session): the keyhole doctrine. The lock demands a key exist; it never judges the key. The system asserts structural truth (a supervision arrangement exists, is documented, named and dated), never substantive truth (that the arrangement is any good). Assessing the veracity of the arrangement is a judgement, and the machine never holds the pen on a judgement (M6 hard line, applied to governance). The veracity of the key lies with the keyholder — the RTO and its named signer — not with the lock. This is what keeps "the system doesn't lie" intact: the system only ever asserts things it can actually know.

The attestation import (doctrine-note feeder). A hollow key still costs the dishonest operator: before, an unsupported trainer was ambient fiction spread across rosters and timetables, deniable everywhere; now it is a discrete, signed, dated artefact — "supervision arrangement: [word]" — with the signer's name on it. That is enforcement-by-attestation (Checkpoint 2 §4) imported into the product: the lie is converted from fog into a falsifiable document with a name attached. The honest operator fills the keyhole honestly at near-zero cost; the dishonest one owns the lie in writing. Truth cheaper than fiction, again.

R5 — One artefact, N faces. Verdict: subsumed by R1.¶

Three renders (auditor / trainer delivery brief / student-public per M13), one substrate, structurally incapable of disagreeing — a property of the R1 verdict, not a separate ruling. Residue worth keeping: renders are generated views pinned to inputs (baseline version + substrate read time), so any historical render is reproducible. Feeds R3.

R6 — Pacing resolution. Verdict: the strategy owns the rule, not the instances.¶

Cohort-level baseline by default. The individualisers — placement (1.1(2)(e)) and credit (M11) — create per-student overlays living on the shared candidate/intake spine (M12), not in the strategy substrate. The strategy states the individualisation rules; the overlays are student-record state. Keeps the tile clean of per-student churn while making "fiction squared" (Checkpoint 1 §2 placement finding) impossible — the overlay is the record.

R7 — Consultation-loop home. Verdict: Studio home, strategy closure.¶

M4 stands: the outreach mechanism is unit-anchored Studio substrate (questions travel with the course, regenerate on package version). The strategy consumes dispositions as R2 review inputs and renders the closed loop. Connective-tissue data point, resolved consistently with R1.

R8 — Recognition posture. Verdict: reference-render, plus two original slivers.¶

The posture itself (RPL via X, CT constrained by Y) is RPL-tile state — rendered, never re-bound (cross-finding 6 discipline). The strategy's original contributions are only Δ4 (the credit-adjustment rule, feeding R6) and Δ5 (the statement of how this product's intake connects sensor to actuator). Straightforward under R1.

R9 — Resource claims home. Verdict: three-way split, no new substrate. North Star ratified with phasing rider.¶

The M10 register decomposes: the requirement floor is read from unit Assessment Conditions (KN/Studio read — derived, never authored); the provision design — which legs (RTO / third-party / student-disclosed), which premises, which fallback for the student-provided leg — is strategy-original substrate, because it is delivery design and feeds 1.1(2)(b) mode reasoning directly; the assurance evidence — certificates of currency, agreements, walkthrough records, the clocked artefacts — is Record custody, on the People expiring-credential pattern.

Debated in-session and resolved. Tim's pragmatic instinct — never make the operator leave the surface — is satisfied by the surface, not by custody: the register is a strategy-tile surface composing all three layers; the operator works it entirely in-surface; the rows show each evidence item and its clock. The custody question is separate, and Record wins it on the uniform-claim-discipline argument: an assignment references People state, a validation claim references Record evidence — a resource claim holding its own attachments would work differently for no principled reason, and the uniform discipline is the product's legal posture (M1). Weaving-room correction adopted: Studio is the content loom; the strategy tile is the delivery-design loom; Record is the evidence shelf.

Phasing rider (settled). Record is specced, not deployed. If the TAS tile reaches build first, the interim is in-tile custody with the schema already Record-shaped (evidence item, clock, ID — the People expiring-credential shape), so handoff is a migration, not a redesign. If even that proves wrong in the builder, plan B is ratified with reasons and we move on — reasoned change and reasoned no-change are both valid closures, our own doctrine applied to ourselves.

Doctrine-note feeders banked this session (for D3)¶

Structural truth vs substantive truth — the boundary of what the system may assert; the general form of the M6 hard line.
The keyhole doctrine — the lock demands a key, never judges it; veracity lies with the keyholder.
The attestation import — the product borrows the regulator's own endpoint move: where verification is impossible, make the signature the liability; fog → falsifiable artefact.
Reasoned-closure reflexivity — the product's own architecture decisions run on the instrument's closure logic (ratify plan B with reasons).

D2 status: COMPLETE (pre-ratification)¶

All nine candidates argued to verdict; R3/R4/R9 carry in-session refinements; concurrence given 2026-07-09. Canon lands when tas-01-model ratifies. D3 (doctrine note) is unblocked — premises per Checkpoint 1 §1, Checkpoint 2 §§4–5, and the four feeders above; Claude drafts solo, Tim redlines.

Filing: FILING-03 (append this checkpoint to workspace/apps/tas-research-01.md) is owed once FILING-01/02 close on digest match — strict sequence preserved.

Checkpoint 3 ends.