Help
v0.0.1 · demo
WELCOME

Getting started · five-minute orientation

Anti-hallucinating AI revenue-recognition platform · ASC 606 · IFRS 15 · ASC 340-40 · ASC 958-605. This page covers the quick-start path, keyboard shortcuts, frequently-asked questions, and the shorthand glossary.

Quick start · five steps

Keyboard shortcuts

⌘ KCommand palette · search and jump to anything
⌘ ↵Confirm / commit on the current page
EscDismiss modal · cancel current action
TabAccept inline AI suggestion (in fields where applicable)
/Focus the search bar in lists and tables
? Open this help page

Frequently asked questions

How does the AI avoid hallucinating ASC 606 paragraph cites?
Eight-layer defense (BRD §F8.1): deterministic pre-check → RAG grounding → constrained-output (Pydantic schema) → verified-registry post-check (every cite must hash-match the FASB Codification index) → self-consistency CISC sampling → verifier model → confidence-routed human review → append-only audit log. Acceptance gate: zero fabricated cites on the 500-question internal benchmark.
Why can't I edit the schedule directly?
Audit-trail integrity. Every change has a controlled path: contract modification (ASC 606-10-25-10), AI override (BRD §F8.7 first-class entity), source-data correction (re-run from Step 1), or closed-period adjustment (new JE with cross-reference). Click "Request change" on the schedule page for the picker.
How is this different from Zuora Revenue / Workiva / NetSuite ARM?
Three differentiators: (1) AI-explainable judgment with codification citations and confidence scores per decision (no incumbent does this), (2) the platform's modernized 12-question investor framework — citationally grounded in current FASB / SEC / PCAOB / IASB authority, (3) anti-hallucination guardrails specifically engineered for accounting (verified-registry post-check, no-LLM-arithmetic rule). 3-minute setup vs incumbents' 6-month implementations.
Does the LLM compute any dollar amounts?
No. The "no-LLM-arithmetic rule" (BRD §F8.6) is enforced: every number that flows to a journal entry, schedule, or disclosure footnote is computed by the deterministic engine using values extracted (and validated) from the source contract. The LLM proposes account codes and memos but never amounts.
How are tenants isolated?
Phase 1: pgvector with row-level security (RLS) per `tenant_id`. Phase 2: per-tenant DB schemas for top-tier issuers. Phase 3: BYOC (entire stack runs in the customer's AWS account) for federal tier. The demo uses a tenant-config registry switchable via `?tenant=` URL param.
Where is the audit trail stored and how long?
Append-only Postgres `judgments` table → per-tenant Merkle hash chain → daily Merkle root anchored to Sigstore Rekor → S3 Object Lock Compliance mode for 7-year retention (matches PCAOB AS 1215 and SOX 404 retention requirements). Auditor can verify any judgment's Merkle proof from outside our infrastructure.
IFRS 15 toggle — does the schedule actually change?
Math is the same under both standards (the 5-step model is highly converged). What changes: disclosure narrative wording (paragraph cites shift to IFRS 15.X format), 4 specific divergences highlight (collectibility threshold, sales-tax expedient, shipping & handling expedient, onerous-contract test under IAS 37), and IFRS-specific companion standards activate. Toggle on the schedule page.
How do I customize this for my CPA firm?
See `CUSTOMIZE.md` in the repo. Most customizations are 2 file edits: `apps/web/lib/tenant-config.ts` for brand identity and tenant registry · `apps/web/lib/demo-data.ts` for the loaded sample contract. Or set env vars: `NEXT_PUBLIC_PRODUCT_NAME`, `NEXT_PUBLIC_OWNER_ORG`, `NEXT_PUBLIC_APERTURE_COLOR`, etc.

Glossary

POPerformance Obligation — a distinct promise to transfer a good or service to a customer.
SSPStandalone Selling Price — the price at which an entity would sell a good/service separately to similar customers.
TPTransaction Price — consideration to which the entity expects to be entitled in exchange for transferring goods/services.
RPORemaining Performance Obligations — aggregate transaction price allocated to unsatisfied POs at period end.
SFCSignificant Financing Component — adjustment when payment timing creates a financing benefit > 1 year.
PCSPost-Contract Support — software maintenance and updates over the support term.
GtNGross-to-Net — pharma-specific 7-bucket deduction waterfall (channel discount, Medicaid, 340B, GPO, prompt-pay, returns, copay).
CISCConfidence-Informed Self-Consistency — sample N times, weighted-vote on agreement; the methodology behind every confidence score.
RAGRetrieval-Augmented Generation — codification chunks fetched from the verified registry and injected into the LLM prompt.
Merkle proofCryptographic proof that a given audit-log entry is part of a daily Merkle tree root anchored to Sigstore Rekor.
OverrideA first-class entity with its own URL, structured reason, and audit-trail entry recording the deviation from an AI judgment.
NFPNot-for-profit — entities under ASC 958. ASU 2018-08 routes inflows between ASC 606 (exchange) and ASC 958 (contribution).
BYOCBring Your Own Cloud — deployment pattern where the entire stack runs inside the customer's AWS account.
Aperture markThe product brand symbol — open-and-close stripes evoking transparency and the act of "looking through" judgment with clarity.
Further reading
· Scenario library · /scenarios — every ASC 606 pattern the platform handles
· Investor framework · /investor-top-10 — the moat · modernized 12-question investor disclosure framework, paraphrased and cited
· BRD v1.1 · docs/specs/BRD.md — the full product spec
· CUSTOMIZE.md — guide for CPA firms forking this platform
· DEPLOY.md — Vercel deploy in five minutes