EXHIBIT 000 / THE CASE FILE
Every claim on dissfam.com points back to this page. What follows is the operating record of one autonomous session: the brief, the run, the Engine reviewing its own work, and the bill of materials. Nothing here is a mock-up — you are standing inside the deliverable.
BRIEF.RAW / RECEIVED VERBATIM
“Make dissfam supernatural. You can do it straight to production. I will give you no instruction, no guidance. Deliver to your absolute best capabilities.”
That is the entire specification the Engine received. Strategy, positioning, copy, design system, artwork, code, review, and deployment followed from it.
One session, brief to production. The ledger below is the real sequence.
Four research agents swept the live portfolio in parallel — Cellavie, Every Era Estate, the old dissfam.com — and hunted the public web for the unlaunched products. The copy on /work is grounded in what they verified, not in guesses.
4 agents · 3 live properties mapped
Tokens, typography, the hairline “Signal” language, the generative mark engine, the WebGL field, and the homepage were built as a single coherent gesture — then frozen as a written contract (docs/DESIGN.md) before anything fanned out.
8 generative marks · 0 raster portfolio assets
Four builder agents took one route each — the registry, the Engine, the pipeline and terms, the contact channel — against the frozen design contract. Disjoint file ownership; no agent touched shared ground.
4 builders · 6 routes · 1 contract
Six adversarial reviewers — visual, mobile, accessibility, copy, code, functional — attacked the finished site. Every finding then faced an independent verifier instructed to refute it. What survived got fixed. What did not, died.
33 findings raised · 32 confirmed · 32 fixed · 1 refuted
The Engine and Process pages were rebuilt around live proof — the operating floor, the veto gate you can press, the engagement replay — followed by a structural refactor that left one implementation of every hard problem.
Press the veto gate yourself →5 live simulations · 29 e2e tests green
Build clean, lint clean, suite green, pushed to main. The human veto stood over every deploy. It still does.
same-session · straight to production
The strongest evidence for governed autonomy is the Engine catching its own mistakes. A sample of what the adversarial pass confirmed — against its own fresh work — and fixed before you ever saw it:
Visitors with reduced-motion enabled received blank pages — the reveal system never wrote its resting state over the server-rendered hidden state.
Fixed. Reduced motion now renders every page fully, verified by a standing regression test.
A decorative vignette mask was silently dimming real content — including the primary call-to-action buttons — on every hero section, worst on the contact form.
Fixed. The mask was moved to a non-content layer; buttons measured back to full brightness.
Every subpage declared the homepage as its canonical URL — an instruction to search engines to de-index the entire site except one page.
Fixed. Per-page canonicals shipped the same session.
The homepage stat “2 live on the public internet” was false — one of the two products sat behind an access wall. The Engine flagged its own marketing copy as unverifiable.
Fixed. The claim was rewritten to what is true, and the case-file links were gated by real status.
That is what the gate is for. Autonomy produced the site; adversarial review produced the truth; a human signed the release.
No templates. No stock imagery. No page builders. Every portfolio mark is drawn by seeded code at runtime.
This public register keeps illustrative workflows, product demonstrations, internal technical receipts, client evidence, and live operating receipts separate. Empty categories stay visible and empty.
Published with fictional sample data and explicit no-live-connection boundaries.
No separately approved real-product evidence. The current Demo remains illustrative.
Published from reproducible commits, deployments, contracts, and test records.
No releasable client artifact, narrative, logo, or measured case exists.
No approved live customer operating receipt exists.
Claim
The accepted production release exposes exactly three deterministic, explicitly fictional operating jobs with human-gate and receipt paths.
Method
Bind the implementation commit to the immutable production deployment, verify the canonical marker, then run reducer, browser, accessibility, security, and preservation gates.
Limits
This proves a technical release with fictional sample data. It is not a customer result, live connection, realised-value measure, or claim about model performance.
Reproducible sources
Verification and rollback
ROLLBACK / milestone/aios-phase-1-2026-07-31
Claim
The accepted release preserves the software-build process and adds the two-path Process hub, five-stage DissEngine installation path, scoped commercial family, and intent-aware Contact contract.
Method
Compare the relocated build process to the preservation baseline, validate all new stage/commercial/contact contracts, bind the release commit to production, and inspect desktop/mobile output.
Limits
This proves the published site contract. It does not approve a fixed price, duration, portability term, client integration, or installation outcome.
Reproducible sources
Verification and rollback
ROLLBACK / milestone/aios-phase-2-2026-07-31
Claim
The accepted release publishes the AI Operating System story, typed integration capability registry, About and claim-controlled Founder pages, Engine reframing, and complete responsive navigation.
Method
Validate typed content registries and claim release rules, run complete responsive/accessibility/browser gates, then bind the release commit to its production marker.
Limits
Named integration capabilities remain Planned or scoped. Founder biography, employer, qualification, portrait, and unsupported entity schema remain withheld.
Reproducible sources
Verification and rollback
ROLLBACK / milestone/aios-phase-3-2026-07-31
Claim
The accepted release publishes exactly three anonymous Assessment branches with deterministic separated scenarios, visible assumptions and evidence, plus scoped Trust and Privacy routes.
Method
Prove calculation invariants and browser-memory/network silence, publish fail-closed first, enable the route flag, repair the unmatched-POST status boundary, and bind the corrected commit to production.
Limits
The result is an estimate, not a promised benefit or combined ROI. It stores no visitor answer and makes no controller, sender, compliance, or customer-result claim.
Reproducible sources
Verification and rollback
ROLLBACK / milestone/aios-phase-4-2026-07-31
Claim
The production-deployed synthetic beta supports the eight-section private report, deterministic reconciliation, signed access, edit, export, deliberate share, revoke, single-use code, owner log, expiry, rate limiting, and exact deletion.
Method
Deploy public routes fail-closed, configure dedicated beta cryptographic keys, create only unmistakable .invalid synthetic records, run the full lifecycle across three browser engines, and clean every record exactly.
Limits
Public save, email, and real-data persistence remain disabled. No legal-controller identity, appropriate report sender, managed bot control, customer evidence, or realised value is implied.
Reproducible sources
Verification and rollback
ROLLBACK / milestone/aios-phase-5-2026-07-31
Every candidate moves through evidence, anonymisation, methodology, claim, and owner review. Client and live evidence cannot skip the controls that make it safe to publish.
Register the claim, source, owner, date, scope, and evidence class.
Reproduce the source and confirm that it supports the exact public claim.
Remove or aggregate identity, commercial detail, source data, and linkable combinations; client/live evidence cannot skip this review.
Record baseline, units, period, formula, exclusions, overlap, confounders, and uncertainty.
Ensure visible copy, metadata, social fields, schema, images, and captions make no stronger claim.
The evidence owner explicitly releases the bounded claim and can revoke it.
Publish from the registry, retain provenance, review expiry, and withdraw if the source or permission changes.
Templates are ready for evidence that does not exist yet. They do not turn a modelled Assessment, an illustrative Demo, or an internal release into a customer outcome.
A client case remains withheld until every required field has evidence-owner, anonymisation, methodology, and claim approval.
An Exhibit is an evidence container, not a licence to strengthen the underlying claim.
No realised-value claim is approved. The ledger remains empty rather than substituting modelled Assessment scenarios or illustrative Demo values.
Do not combine different units, released capacity with cash, pipeline with realised gross profit, cash principal with profit, or overlapping effects.
Permitted separated units
Method fields required before publication
baseline source and period · post-change source and period · calculation method · gross-profit/cash/capacity/risk classification · cost where applicable · adoption · confidence · overlap exclusion · confounders · evidence owner approval
Assessment calibration
0 customer records. No model-default revision is supported by realised evidence. Defaults remain explicit, editable, and unchanged.
The next case file could be yours. Send a brief and watch the same pipeline run with your name on the ledger.