Rust + Axum
A reusable edgar-core library — typed statements, diff, sentiment, verification badges — behind a server-rendered Axum site.
Six things no other tool does at any price near this
A verified substrate, and a workspace on top of it
The verification layer is why you can trust it. The workspace is why you'd open it every day.
01 · Verification layer
Click any number, land on the line it came from
Income statement, balance sheet, and cash flow render as clean, interactive tables — and every figure wears a verification badge. Click it and the embedded viewer scrolls to and highlights the exact line in the original 10-K. Not a reformatted copy. The audited legal document.
Why it matters — a Bloomberg user can't click a revenue figure and land on the highlighted line in the filing. That single click is the whole proposition.
1
01a-statement.pngStatement view (
/statement) with the Verified badges.
2
01b-audit-trail.pngFact audit trail (
/fact) — cross-check & checks.
3
01c-source.pngThe figure highlighted in the stored 10-K (
/source).02 · Verification layer
See exactly where XBRL disagrees with the filing
Provenance extracts financials from the HTML independently, then cross-checks them against the XBRL data every vendor treats as ground truth. Each disagreement is cataloged and filterable by company, concept, and magnitude — the class of error that lets figures like 500 trillion shares slip into a database unnoticed.
Why it matters — good analysts have caught vendor data errors by hand before. This surfaces them systematically, across the whole corpus.
▤
screenshots/02-discrepancy-catalog.png
Capture the /discrepancies catalog — a table showing both values, magnitude, and status. Pick a company with a striking disagreement.
03 · Research workspace
Read what management quietly changed
Pull any prose section — Risk Factors, MD&A, Business, Notes — beside last year's. Additions in green, deletions in red, paragraph by paragraph. See exactly what was added, removed, and rewritten. Changes in risk-factor language routinely precede the events they describe.
Why it matters — year-over-year prose diffing is otherwise only found inside a terminal that costs $10,000 a year.
≢
screenshots/03-prose-diff.png
Capture the /diff view on Risk Factors or MD&A — show a screen with visible green additions and red deletions.
04 · Research workspace
Every ratio is a tree of verified inputs
Not just "ROIC is 14.3%." ROIC = NOPAT / Invested Capital — expanded down to each underlying figure, and every leaf of the tree links back to the highlighted line in the filing it was drawn from. No black boxes, no vendor methodology you can't inspect.
Why it matters — "here's exactly how I calculated it, traced to the audited filing" is the fiduciary's version of showing your work.
⋔
screenshots/04-ratio-decomposition.png
Capture the /ratios view with a ratio expanded into its components, source links visible on the leaves.
05 · Research workspace
Screen on the numbers and the prose at once
Combine quantitative and language criteria freely — "accruals ratio above 0.10 for three of five years AND a newly added risk factor." Every hit traces back to source: quantitative values to the input facts and their filing, NLP signals to the triggering passage.
Why it matters — quant screens and text screens have never lived in the same query, let alone one where every result is traceable.
⌗
screenshots/05-screener.png
Capture the /screen builder with a combined quant + NLP screen and its ranked results — ideally with a "why" trace visible.
06 · Research workspace
How trustworthy is this company's data?
A per-company transparency dashboard: how many XBRL discrepancies across its filing history, whether it has restated, how often it reaches for non-standard tags, how much its risk language shifts year to year. Not a score or a rating — the evidence, laid out, so you form your own judgment.
Why it matters — data quality is itself a signal. A company that constantly restates and mistags is telling you something.
◈
screenshots/06-integrity-profile.png
Capture the /integrity dashboard for one company — discrepancy count, restatement flag, extension-tag usage, prose volatility.