Micca Music Publishing Platform
Prepared for Alberto & Consuelo · 1 October 2026
Agreement document extraction now reads the uploaded file. The current focus is testing one writer through contacts, agreements and songs, with permanent storage still to build.
01 / Where the project stands
How much of the product can you see and click?
How close is this to something you can safely sell?
The hosted prototype supports guided testing. Agreement import now uses real AI extraction; durable records, real logins and safe financial processing remain unfinished.
“The closest answer to how much visible functionality has been represented? remains 71.1%. The closest answer to how close is this to a safe commercial product? is 24–27%. Quoting either number without its lens would be misleading.” — Deep-Dive Re-Audit, 21 July 2026, §1.1
02 / What is built
The ranges below have not been rescored since 21 July. Dated corrections identify later work; they do not establish a new production-readiness percentage.
| Area | Can you demo it? | Can a customer rely on it? | Audit finding and later corrections |
|---|---|---|---|
| Contacts | 92–96% | 15–25% | Transient contact state1 Oct review: contacts persist in this browser. Shared storage and real permissions remain unfinished. |
| Songs / catalog | 90–95% | 15–25% | 27 Sep: contact-to-song links repaired; clean testing mode available. Song edits still disappear on reload. |
| Agreements | 91–95% | 15–25% | 1 Oct: real document extraction, editable royalty allocation and territory input. Agreements still disappear on reload; imported royalty rates need manual review. |
| Accounting | 80–88% | 10–20% | Visually mature; no journal or transaction boundary |
| Royalties / statements | 80–90% | 10–20% | Not safe to pay anyone from. Allocation and period semantics unsafe |
| Creative / sync | 85–92% | 25–35% | Strong prototype; integrations are simulated |
| Deliverables | 65–80% | 10–20% | Metadata workflows exist; actual delivery lifecycle does not |
| Analytics / reporting | 60–70% | 5–15% | Derived from sample data only |
| Onboarding | 80–88% | 10–20% | Broad flow; no identity or durable workflow |
| Tenant / platform admin | 55–70% | 5–15% | Settings often disconnected or globally scoped |
| Artist / client portals | 85–90% | 10–20% | Polished views over shared sample data |
| Accessibility / mobile | 60–70% | 40–50% | Good baseline; mobile and keyboard issues remain |
| Languages (i18n) | 90% | 13.3% | Switching and right-to-left work; most values are still English |
Source: Deep-Dive Re-Audit, 21 July 2026, §1.3. Ranges are audit estimates, not measured values.
03 / The gap that explains everything
Screens, forms, tables, workflows — everything you click. 83 of 99 essential features fully working, 15 partial, 1 barely started.
The server that stores data permanently, checks who is logged in, and does the money math. 26 of 89 essential features fully working, 23 partial, 15 barely started, 24 deferred and 1 missing in the July ledger.
Nothing is permanent yet; the backend is designed but not built. Corrected 1 Oct 2026
Contacts are saved in the current browser. Agreements, songs, royalties, creative and accounting remain in memory and are lost on reload. Clean testing mode warns before leaving when there are unsaved changes.
A server endpoint now extracts agreement terms from uploaded documents. That is a working backend capability, but it does not provide shared record storage, user accounts or enforced workspace isolation.
Use the sample-data demo or the clean testing instance for guided walkthroughs. Keep source records outside the prototype.
Counted from the 3 July 2026 feature ledger, essential (MVP) features only, where a partially working feature counts as half. Frontend 99 + backend 89 + 3 features classed as legal, design or infrastructure = 191 essential features in total.
04 / What is missing
July release assessments, with hosting and storage corrections. Marketing and internal automation have not been re-audited for this update.
| Intended use | Verdict | Why |
|---|---|---|
| Local demo with sample data | Conditional go | Works broadly; label simulated behaviour and keep unsafe financial actions switched off |
| Public hosted demo | Now live | No-go — no deployment existed Corrected 20 Aug 2026 — now live at micca-dashboard.pages.dev |
| Public marketing launch | No-go | Dead links, unsupported claims and accessibility failures on the marketing site |
| Paid pilot with a real client | No-go | Real identity, shared durable storage, enforced tenant separation and privacy operations remain unfinished |
| Royalty calculation or payment | No-go | Financial rules and safeguards are not yet trustworthy |
| Automation on untrusted content | No-go | The internal automation tool has unresolved safety gaps |
The July audit catalogued 13 release blockers and 32 pilot blockers. These are historical counts, not a verified count of issues still open today. The storage correction below reflects current code.
| Audit ID | What it means for you |
|---|---|
| DATA-001 | Contacts survive reloads in this browser; agreements and songs do not. No shared durable record store. |
| AUTH-001 | There are no real logins — everyone who opens it is an administrator |
| TEN-001 | One client could see another client’s data |
| ROY-001 | Royalty maths is not safe to pay anyone from |
| LAND-001 | Marketing site links and legal pages do not work |
Source: Deep-Dive Re-Audit §1 and §15. IDs are kept so each line traces back to the full audit.
05 / What is next
These estimates have not been recalculated for the work completed since July.
Show Micca to anyone, anywhere, using sample data.
One real partner, with real data — but no royalty payments running through it.
Sell it. Administer real rights and real money for paying customers.
The July audit estimated 6–12+ calendar months for the commercial path with one engineer. That range is not a new countdown from October. Elapsed time cannot be subtracted from it without reviewing completed work, remaining scope and staffing.
The next planning step is to finish the contact, agreement and songs walkthrough, record its gaps, and re-estimate a defined pilot. The original ranges exclude time waiting on product and legal decisions.
06 / What we need from you
Consuelo’s 24 September direction was to focus on the contact, contract and songs for one writer. Keep that walkthrough focused; capture territory questions when they arise in the workflow.
Immediate follow-up: verify the imported agreement against its source and enter royalty rates by rights type manually. Extraction does not yet populate the Royalty Allocation rows.
Consuelo also needs to resolve whether society fields belong on contacts or agreements: the walkthrough request and requirements workbook disagree. Territory inheritance and overrides remain open schema questions.
Engineering needs durable storage for agreements and songs before the clean instance can hold ongoing work. Folder-wide import of agreements, songs and mixed-writer statements remains separate, unbuilt scope.
The workshop listed 14 decisions. They are retained below for traceability, not counted as 14 confirmed open items. September’s testing focus does not itself approve a paid-release scope.
Founder and legal decision. The workshop lists software, managed service and combined options, with no engineering preference.
Unblocks: legal structure, what financial data we may store, privacy obligations, service guarantees.
Recommended: Contacts + Agreements, with royalties read-only — the two strongest modules, without taking on payment risk.
Unblocks: pilot scope, what must be production-grade first, the target partner conversation.
Recommended: read-only — clients can see statements, but Micca does not calculate or issue payments yet.
Unblocks: the largest single block of financial-correctness work.
| ID | Question | Recommendation |
|---|---|---|
| D4 | Who is legally responsible for each kind of data? | Per-dataset table, drafted with legal |
| D5 | Which countries and societies do we support first? | Needs your call — drives integration order |
| D6 | Which territory-code standard is authoritative? | TIS reference data implemented in August; licensing and per-writer inheritance decisions are not closed by that work |
| D7 | Who is allowed to do what, at every level? | Define roles before real logins are built |
| D8 | Can contacts be shared between workspaces? | Default to no; revisit if a partner needs it |
| D9 | Which financial details do we store vs. hand to a payment provider? | Store as little as possible |
| D10 | How long do we keep data, and where does it live? | Follows from D3 |
| D11 | Which integrations are real launch dependencies? | Separate genuine blockers from roadmap ambition |
| D12 | Which technology decisions are locked? | Cloudflare hosting and agreement extraction exist; confirm the durable storage and identity platform |
| D13 | What uptime and support do we promise? | Follows from D3 |
| D14 | How much evidence before AI touches rights or money? | High bar; keep AI advisory, not authoritative |
Sources: audit decision workshop agenda (D1–D14, with recommended options and owners) and roadmap §4.
07 / Recent activity
| Date | Work |
|---|---|
| 1 Oct 2026 | Agreement AI scan now reads the uploaded document through the real extraction endpoint. Fixed request failures, stalled scans and confidence scoring. The deployment record reports a successful live extraction with both parties resolved. Royalty rates by rights type still require manual entry. |
| 1 Oct 2026 | Agreement form revised after the 28 September walkthrough: editable royalty allocation, a territory picker, separate governing-law jurisdiction, 12 top-level agreement types and a direct Draft-to-Active path for contracts already signed. Imports preserve publishing deal type; filters use the revised type list. |
| 28 Sep 2026 | Agreement import wizard added: upload, extraction, field review and party confirmation. Extraction follows the publishing deal-type requirements and includes source evidence. Low-confidence fields start unaccepted. Nine requirements still lack fields in the agreement data model. |
| 27 Sep 2026 | Clean testing instance published with a first-run guide. Contact-to-song links and agreement party navigation repaired. Added a warning before leaving with unsaved records; contacts persist locally, but agreements and songs still reset on reload. |
| 24–27 Sep | Replaced leaked placeholder labels, tooltips and accessibility text. Demo charts now keep their sample dates inside the displayed reporting window. |
| 24 Sep 2026 | Publishing deal-type requirements, missing contract-term inputs and agreement completeness indicators added. Catalog renamed Songs in the interface. Copyright workflow can write an accepted registration number back to covered songs; this is a prototype workflow, not a live government filing integration. |
| 27 Aug 2026 | Territory data rebuilt on the official CISAC standard — the previous list was not TIS at all, and 11 of its 20 definitions were wrong. “Worldwide” covered 249 territories instead of 201, and the UK was still inside the European Union. No registration file could have carried a valid territory before this. |
| 27 Aug 2026 | Currency and language brought onto international standards — ISO 4217 (including the currencies that do not use 2 decimal places, such as yen) and ISO 639. The song form previously offered 18 languages; it now offers all 71. |
| 27 Aug 2026 | Catalog health score, matching the one on contacts — weighted for whether a song could actually be registered, not how many boxes are filled |
| 20 Aug 2026 | Contacts fixes from Alberto’s walkthrough — section navigation, working filters, saved edits, search, phone validation (PRs #230, #231) |
| 20 Aug 2026 | Dashboard published to a live link for the first time |
| 13 Aug 2026 | Contacts aligned to Consuelo’s revised spec workbook — all 11 sheets audited (PRs #227, #228) |
| 23 Jul 2026 | Codebase reorganisation, phases 1–6 |
| 21 Jul 2026 | Deep-dive engineering audit — the source of most numbers on this page |
Sources: platform changelog (24–30 September) and commits c406fe0, d72a0c6 and caa48ae (1 October). The 1 October commit records 12,904 passing tests and a successful live extraction; those app checks were not rerun for this report update.