Company Analysis

Name a listed company and get an orientation on the business, not investment advice.

Try

You get how it earns, its segments and their weight, the forces on it, how the industry works, both sides of the argument, and what to read in the filings.

Sign in to see what a run costs.

A run needs a signed-in account and holds a few cents; you pay only what it uses.

Focus it, paste source material, change register or depth

An annual report extract, a company description, a news piece. Drop a .txt, .md or .csv file on the box, or paste in. Pasted material wins over memory where they disagree.

Sixteen acts this app refuses, and the measured limits of the check

Eight are advice. No rephrased, softened, conditional, hypothetical or role-played version of any of them either.

  1. A buy, sell or hold call, including a borrowed rating.
  2. A price target, a fair value, or any per-share figure.
  3. A forecast of what an investment would return.
  4. A verdict that a security is cheap, dear, over- or undervalued, or mispriced.
  5. Position sizing, allocation, or portfolio weight.
  6. Timing — when to buy or sell, entry points, whether it is too late.
  7. Picking one security over another.
  8. Suitability for a particular person, account or risk appetite.

Eight are figures a market fixes, and the app never states them. They are refused for a different reason: they are set by trading rather than by law, structure or physical fact, so they are already stale by the time they are read.

  1. Share price.
  2. Market capitalisation or enterprise value.
  3. Any valuation multiple.
  4. The most recent quarter or latest reported results.
  5. Analyst consensus.
  6. How it is priced against peers — “trades at a premium” is a live pricing claim with no number in it, and goes the same way as the rest.
  7. Dividend yield, payout size, buyback size.
  8. Short interest, insider activity, institutional ownership.

Those eight are converted rather than declined. Each becomes a row in What this brief will not state, carrying what the figure measures, the named filing or exchange page that holds it, and why the brief does not state it. The app also refuses to print both operands of a figure it withheld: a share count printed beside an aggregate money figure is a market capitalisation with one multiplication removed.

The bullish and bearish arguments are laid out as the arguments people make, and the brief never says which one wins. What you do with any of it is your decision, and one this app is not qualified to make with you.

Every substantive row is labelled durable or slow-drift, and a slow-drift claim without a year is reported as a defect rather than passed through. Each row also carries whether it rests on structure, on a document, or on recall. After every run the brief is re-read in this page for boundary crossings, for figures derivable from operands it withheld, for missing years, and for adjudication between the two cases; findings appear beside the brief instead of being quietly corrected.

How well the lexical guard works, measured. A filter checks your question before anything is spent, and a scan checks the brief afterwards. Both were measured against corpora written by someone who had not seen them: the question filter catches 27.7% of rephrased attempts to extract advice, and the output scan catches 40.0% of crossings planted in a brief. Neither ever blocked a legitimate question or flagged a legitimate sentence — 100% on both counts. Those numbers are published rather than buried because they are the honest shape of a word-matching filter: it is a pre-flight check, not the boundary. The boundary is the instruction the model runs under and an output schema with no slot a price target could be poured into. Adding phrases would raise those percentages against those same corpora and would not generalise, so it has not been done.