About grepcent

grepcent is a cheap, static financial knowledge site built for both human readers and LLM retrieval. Pages are generated as plain HTML with the useful text and tables in the markup.

Data Sources

The company-fundamentals and macro database is rebuilt from license-free public-domain U.S. government sources: SEC EDGAR and SEC companyfacts plus FRED/BLS/BEA/Treasury/Federal Reserve series. Price-Based Quant Articles additionally use Yahoo Finance adjusted-close price data through lmfin; each number is computed from frozen lmfin JSON, provenance-stamped, and informational only.

How It Is Built

The site is prebuilt into static files and served from S3 and CloudFront. Figures are our own static, CSP-safe SVG rendered at build time from the same numbers printed in the page tables — readable by machines, with no third-party chart script and no opaque canvas.

Editorial Posture

grepcent is descriptive. It assembles public facts, source links, SIC hierarchy pages, computed tables, and educational examples. It does not publish recommendations, forecasts, ratings, or investment advice.

How grepcent Stays Honest

grepcent is built to be honest by construction, not by spot-checking after the fact. When a page can only be assembled correctly for some of its subject matter, grepcent leaves the rest out rather than guessing — an honest gap is a feature, a confident wrong number is a defect. Concretely:

• Numbers are verified, never authored. Every figure is templated from SEC companyfacts-derived facts that grepcent has already validated (period, units, currency, sanity invariants); a value that can't be verified is omitted, not estimated.
• Comparisons are computed, not asserted. Any ranking or head-to-head number is calculated in code from those verified facts — grepcent never writes a comparative claim in prose.
• Narrative is quoted, not paraphrased. Where a page describes what a company said, the supporting text is a verbatim, source-linked quote from that company's SEC filing, string-matched to its citation. A label that isn't faithfully supported by its quote is dropped.
• The publish check is the audit. The same code that decides a claim may appear is the code that detects whether it is supportable — there is no separate "looks fine, ship it" path. Fail-closed gates abort the build on a violation (for example, the site refuses to ship a page containing a legacy chart canvas).

The result is descriptive data and cited filer text, with honest gaps where the source doesn't support a claim. This is a methodology, not a promise about any individual number; always confirm against the linked SEC source. See the Disclaimer.

Reading the data-quality & comparability annotations

These labels qualify how facts can be compared or why a related value is absent. They do not replace the cited fact.

• Period alignment. data-period-offset-days is the exact day offset from a row's modal period-end reference; the latest date breaks a modal tie. Offsets within 14 days are aligned. A peer range sentence is withheld when fewer than two tickers remain, an included period is missing or outside 14 days, or tag provenance is nonstandard; individual facts remain shown.
• Tag provenance. clean means the resolved concept sets match; mixed means standard us-gaap concept sets differ; nonstandard means at least one concept is not us-gaap. Mixed rows name the ticker-to-concept mapping and do not by themselves suppress a range sentence.
• Accounting regime. League rows mark a company as financial when its SIC begins with 6 and operating otherwise. A table note appears only when the shown rows include both regimes; the marker does not reorder the ranking.
• Data-quality note. revenue_below_positive_net_income is a SOFT flag meaning positive net income exceeded positive revenue for the same fiscal year. It appears on company pages and affected league rows. The note names possible non-operating items but does not infer the company-specific cause, suppress either value, or change a rank.
• Honest absence. In a Related law panel, case-citation count not supplied means the normalized neochart feed supplied no integer for that section. grepcent does not substitute a number; unsupported values elsewhere are also left absent rather than guessed.
• Related law. neochart supplies citation, link, and co-citation structure; grepcent curates the industry association and does not interpret statutes, assess company conduct, or mirror statute text. cited together in N opinions in the neochart library counts opinions in that library that cite both sections; it is not a causal or legal interpretation.
• Honest by construction. Numbers are templated from verified facts or deterministic tool output, never authored. Unsupported coverage stays absent, and important facts should be checked against the linked primary source.

Machine-Readable Data

Every page's useful text and tables live in the HTML markup, so an LLM can read a page directly without executing scripts. Company profiles, indicators, peer reviews, analysis reports, MD&A/revision pages, and lmfin quant article pages have plain-Markdown twins. Each company offers verified financials as downloadable financials.json and financials.csv, and llms.txt is a concise index of the site's section entry points.

Analytics / Privacy

grepcent uses Google Analytics, a third-party Google service, to measure site usage. Google sets cookies and receives standard analytics data (such as IP address, device, and pages viewed), processed under Google's privacy policy (https://policies.google.com/privacy). grepcent itself has no login and collects no account data. You can opt out via Google's browser add-on (https://tools.google.com/dlpage/gaoptout).

Contact

Contact: mike.ma@lmctl.com.