Skip to content

Method

From your holdings to a briefing you'll actually read.

Connect your holdings, build the research file, and keep the picture current. Here is how the charting company puts it together.

  1. Step 01

    You connect a portfolio

    Link a brokerage account read-only, upload a CSV from your broker, or type in tickers. the charting company reads positions and weights — it cannot place trades, move money, or change anything in your account. Within a few minutes you have a coverage set: the specific list of companies the charting company is now responsible for.

  2. Step 02

    We build the research file

    For each holding, the charting company pulls the full document history — annual and quarterly reports, current reports, proxies, insider filings, the last several earnings calls — and builds a Position File. Financial data comes from XBRL, so the numbers tie to the filing rather than to a scraped table. This backfill runs once and typically takes under an hour.

  3. Step 03

    We keep the picture current

    From then on it runs on a schedule: EDGAR polled continuously, transcripts processed when they are released, licensed news feeds monitored throughout the day. When a document lands it is diffed against the last one, the changes are classified, and anything material is queued for the next brief — or, if it cannot wait, sent as an alert immediately.

  4. Step 04

    You get one page, before the open

    At 5:30am ET the brief compiles: what changed overnight across your holdings, ranked by materiality, each item two or three sentences with a link to the document it came from. On a quiet morning it is four lines long. That is the point — the brief is a function of what happened, not of a content quota.

  5. Step 05

    You ask, and everything cites its source

    Ask a question across your whole portfolio and the charting company answers from the documents, with every claim linked to the paragraph it came from. Export any of it as a memo you can send to a client, a partner, or your future self. Then you decide what to do, which is the only part of this that was ever interesting.

Step 04 · The output

One page, before the open.

At 5:30am ET the brief compiles. Items are ordered by materiality rather than recency, each one is two or three sentences, and every one carries the document it came from. The strip at the bottom lists everything the system reviewed and cleared — proof that a quiet brief means a quiet night, not a lazy system.

Fig. 1Morning Briefing — 6 August 2026, three items flagged from 41 documents.Illustrative product interface. Sample portfolio; companies shown in analyses are fictional.

The pipeline

What happens between a filing landing and you reading about it.

EDGAR
XBRL
Transcripts
Newswire
Normalisecontinuous
Diffon filing
Retrieve + verifyper query
Brief
Alert

Ingest

EDGAR submissions, full-text search and the XBRL company-facts API are polled continuously for every issuer in any coverage set: 10-K, 10-Q, 8-K, DEF 14A, S-1/S-3, 13D/G and Forms 3/4/5. Financial statement data comes from XBRL facts rather than parsed HTML tables, which is why the numbers reconcile to the filing. Transcripts and newswires arrive from licensed providers.

Normalise

Documents are segmented structurally, not as flat text: Item 1A stays Item 1A, MD&A stays MD&A, the Q&A portion of a call is separated from prepared remarks. Entities are resolved across ticker, CIK, ISIN and LEI, so a position maps to the right filer even after a ticker change, a spin-off or a re-domiciliation.

Diff

Filing comparison is section-aligned and paragraph-matched, then each change is classified: added, removed, materially reworded, or cosmetic. Cosmetic changes — renumbering, date rolls, formatting — are suppressed by default. This is the difference between a diff you will read and a diff that returns four hundred highlighted lines.

Retrieve

Every document is chunked with its structural metadata intact and indexed for hybrid retrieval, lexical and semantic, followed by reranking. When the charting company answers a question it retrieves spans first and reads only those spans. It does not answer from model memory.

Ground

Every factual sentence is anchored to a specific span in a specific document at a specific accession number. A verification pass re-checks each claim against its cited span before anything is published, and claims that fail verification are dropped rather than hedged.

Schedule

EDGAR is polled on a tight interval during filing windows and after the close, when 8-Ks cluster. Transcripts are processed on availability. Briefs compile at 05:30 ET. Alerts fire on ingest, not on a batch.

Fig. 2Research assistant — an answer with every claim anchored to a source span.Illustrative product interface. Sample portfolio; companies shown in analyses are fictional.

Step 05 · Grounding

A citation is a span, not a link to a PDF.

Linking to a 140-page annual report and calling it a source is a gesture, not a citation — nobody goes and finds the paragraph, which means nobody catches you when the paragraph does not say what you claimed.

At the charting company, a citation is an accession number, a section identifier and a character range. Click it and the filing opens scrolled to that passage, highlighted in place.

First brief lands the next morning.

Connect a portfolio, or just paste in ten tickers.

See pricing

Free tier, no card. Read-only brokerage connection. No trades, no advice.

Get to know the charting company.

We're opening early access in small groups. Leave your email to join the waitlist, and we'll let you know when your workspace is ready.