Deterministic DCF
from real SEC data.
- Every number traces to a 10-K filing.
- Every formula traces to a peer-model lineage.
- Every refusal is documented.
LLMs guess. CIPHER proves.
~3-7 seconds on EDGAR-cached tickers
A DCF asks one question: what is this business worth if these assumptions are true?
A discounted cash flow model projects a company's future free cash flow, discounts those dollars back to today, and turns the result into a model-implied enterprise value, equity value, and value per share.
What it is for
Bankers use a DCF to pressure-test intrinsic value. It shows which assumptions drive the answer: revenue growth, margins, capex, working capital, WACC, and terminal growth.
Why it usually takes hours
A junior banker can spend 4-12 hours on a first-pass public-company DCF: pulling 10-K data, mapping GAAP tags, building formulas, checking circularity, formatting the workbook, and documenting sources.
Why a normal LLM is risky
A chatbot can sound confident while using stale numbers, inconsistent averaging windows, broken formulas, or fabricated line items. CIPHER uses deterministic EDGAR fetches, formula lineage, refusal logs, and a SHA-chained audit trail.
REAL SEC DATA
Direct from EDGAR. No APIs. No scraping. Just facts.
DETERMINISTIC
Same input. Same output. Every time.
ADVERSARIAL MINDSET
AURORA refuses every unsafe write per-cell, with logged reason. CIPHER doesn't guess.
AUDIT CHAIN
Every cell traced. Every source linked. SHA-chained.
REFUSAL AS OUTPUT
CIPHER refuses rather than fabricate. That's the point.
Five verified pipeline stages. Exactly what runs.
EDGAR
Fetch SEC Company Facts & GAAP line items (cached 24h)
POPULATE
Derive 5y mean ratios from real EDGAR data
ALCHEMIST
Peer-DCF schema synthesis (content-addressed SHA)
ALCHEMIST
Dialect-aware fill with per-cell write refusal by AURORA
RECALC
COMPUTE
Closed-form DCF defaults: model-implied EV → equity → $/share
on 22 real EDGAR tickers
Derived ratios (mean) — 5 sample tickers
RUN 2026-05-02| System | EDGAR-direct | Derived ratio | Audit chain | Deterministic |
|---|---|---|---|---|
| CIPHER | 100% | 100% | yes | yes |
| GPT-5.5 | 80% | 17% | no (by architecture) | no (by architecture) |
| Claude Opus 4.7 | 80% | 23% | no (by architecture) | no (by architecture) |
| Grok 4.3 Preview | 80% | 14% | no (by architecture) | no (by architecture) |
| Gemini 3.1 Pro | 80% | 23% | no (by architecture) | no (by architecture) |
AUDIT & ASSURANCE
Trace every number for PCAOB, SOX 404, and internal audit.
MODEL RISK
Deterministic outputs. Reproducible. Defensible.
BUY-SIDE & RESEARCH
Faster models. Cleaner inputs. Stronger theses.
DEVELOPERS
CLI, API, and SDKs. Integrate CIPHER where you build.
TEAMS & ENTERPRISE
One source of truth. Standardize across your organization.
Trust the Output. Ship with Confidence.
Because if it didn't cross CIPHER, you can't trust it.