Search in context
Every passage where Buffett writes intrinsic value
or margin of safety
, keyword-in-context, across all the letters — with a link to the exact paragraph.
In development · preview
Reading the letters of value investing, by the numbers.
An analytical layer over Warren Buffett's Berkshire Hathaway shareholder letters — five decades of them, made searchable, measurable, and mappable.
Concept distribution
Illustrative
The lens
You already read the letters. Value Atlas lets you measure them.
Warren Buffett has written to Berkshire Hathaway's shareholders every year since 1977 — a plain-spoken, decades-long text on how a great business thinks about risk, value, and capital. Value Atlas doesn't republish those letters; it sits on top of them as an analytical instrument. Search a phrase across fifty years, watch an idea rise and fade, compare one era against another — then follow any result straight back to the letter it came from.
The tools
Concordance
Corpus
Distribution
Keyness
Co-occurrence
Reader
Workspace
And the point is
Seven ways to read a corpus you can't read seven times.
Every tool is computed live over the full text and links back to the source. Each view is a URL, so any finding is a link you can send.
Browse the whole canon
Every annual letter from 1977 to today, indexed by year and length. The full corpus at a glance, with a way into each one.
Track a concept over time
Watch how an idea — moat
, float
, derivatives
— rises and fades across the decades, and how the vocabulary evolves with the business.
Compare two eras
What makes the 1980s letters distinct from the 2010s? Surface the terms most characteristic of any slice of the corpus against another, by log-likelihood.
Map a term's neighborhood
What words keep company with risk
, or cash
? Rank a term's collocates by association and read the shape of an idea around it.
Read at the source
Follow any citation out to the canonical edition of the letter. Value Atlas is the lens; the letters themselves stay where they live.
Keep what you find
Star any result as you go. Your collection persists in your browser and exports to CSV for your own notes and writing.
Treating a body of investing wisdom the way a scholar treats a text — closely, quantitatively, and with every claim traceable to the page.
The corpus
Built on the letters — and respectful of them.
The first corpus is the complete run of Berkshire Hathaway's Chairman's letters. The analytics are computed over the text; the reading happens at the source.
Value Atlas stores a local, read-only index of the letters purely to compute concordances and statistics. It doesn't reproduce or republish the text: every citation links out to the canonical edition, so you read Buffett's words where they're meant to be read.
Roadmap: extend the corpus beyond Buffett to the wider canon — Benjamin Graham and Charlie Munger — so the same tools can trace an idea across the whole tradition, not just one author.
Buffett · Berkshire letters
Graham — planned
Munger — planned
Under the hood
A small, sharp stack.
Deliberately dependency-light: no CSS framework, no charting library, no precompute step. Fast to build, cheap to run, and honest about what it's doing.
Rust, end to end
Built with Leptos and Axum — server-rendered, then hydrated in the browser. One language from the query to the pixel.
Hand-drawn charts
Every chart is hand-written SVG, like the figure above. No third-party viz dependency to age or bloat.
Read-only index
The corpus is a SQLite file with a full-text index. Every analytic is computed live — no cache tables to go stale.
Every view is a link
All page state lives in the URL, so any search, chart, or comparison is a shareable, bookmarkable link.
Status
The instrument is built. The corpus is next.
The analytics engine — search, distribution, keyness, co-occurrence — is complete and running. What's in progress now is preparing the letters into the corpus so the tools have five decades to work on. Follow along, or get in touch.