Skip to content
Contact us

SteadyShares

Stock research for private investors, where every figure carries its source and an AI assistant answers from the site’s own data.

Our own product, not client work steadyshares.com(opens in a new tab) Read 19 October 2026

The job

Research tools for private investors tend to be priced for institutions, or padded with scores nobody explains. SteadyShares is the plain version: research built from filed numbers, every figure with the source it came from, and every model with its limits printed beside it. It is educational information, not financial advice, and it says so.

What we built

  • A stock screener, and a research page for each company it covers, with a fair value from one published model whose method and blind spots are written down
  • The holdings large funds file each quarter with the US regulator (their 13F filings), fund by fund
  • US politicians’ disclosed trades, read from the public filings
  • Research studies recomputed from the database every morning, each saying what it cannot tell you
  • A data API and an MCP connector over the same data, so another AI agent can use it, and live figures in Google Sheets and Excel

Where the AI is, and how it is kept honest

Answers fetched, never recalled

The assistant answers from the same data as the rest of the site. Every figure it gives is fetched and carries its source, so when it cannot find a number it leaves a blank rather than a wrong one. 2, 3

Picks made by rules, not by a language model

The picks on each market come from a fixed set of published rules over filed numbers, and the reason under each one is assembled from that company’s own figures. No language model writes the reasons. 4

Its weaknesses, scored in public

A scorecard published on 27 July 2026 rates grounding 9 out of 10, and saying “I don’t know” 3 out of 10: the assistant hedges where it should refuse, and the post calls that its biggest risk. The scores are its own assessment, and it says so. 3

An audit of its own data

When the political trades feed was first profiled, only 1,190 of its 5,000 rows were stock trades, and 3,972 rows came from a single filer. The study says so on the site, and its live figures are recomputed every morning. 5

What it does not do yet

  • There is no evaluation set you can run yet, so the assistant’s scores are a self-report.3
  • Fund holdings are as filed: US-listed positions only, and up to 45 days old on the day they are filed.4, 7
  • Nothing is for sale yet: signed-in accounts get the full limits at no cost until billing is switched on.6

What it shows about the work we would do for you

Answers that show their working

The assistant’s rule, fetch the figure and show where it came from, is the rule behind our knowledge offer: an answer you can check in two clicks, or a blank.

Knowledge that cites itself

Limits printed beside the claim

Every model on it says what it cannot see, and the weakest score is published with the strongest. Our record does the same for an agent: what it refused, beside what it did.

The record

Sources

  1. 01
    SteadyShares: About, and its editorial rules steadyshares.com(opens in a new tab)

    Weight: primary; the product’s own published page (our product) Re-checked: every 30 days Last read: 19 October 2026

  2. 02
    SteadyShares: home page, the AI advisor (“grounded, cited answers”) steadyshares.com(opens in a new tab)

    Weight: primary; the product’s own published page (our product) Re-checked: every 30 days Last read: 19 October 2026

  3. 03
    SteadyShares: “SteadyShares’ AI agents: an honest scorecard”, 27 July 2026 steadyshares.com(opens in a new tab)

    Weight: primary; the product’s own self-assessment, published as such Re-checked: once a year Last read: 19 October 2026

  4. 04
    SteadyShares: how the AI picks are made steadyshares.com(opens in a new tab)

    Weight: primary; the product’s own published page (our product) Re-checked: every 30 days Last read: 19 October 2026

  5. 05
    SteadyShares: “We audited our own data, and three quarters of it was not what it claimed” steadyshares.com(opens in a new tab)

    Weight: primary; the product’s own study, with its method Re-checked: once a year Last read: 19 October 2026

  6. 06
    SteadyShares: pricing (“Nothing is for sale right now”) steadyshares.com(opens in a new tab)

    Weight: primary; the product’s own published page (our product) Re-checked: every 30 days Last read: 19 October 2026

  7. 07
    SteadyShares: methodology, with its limits steadyshares.com(opens in a new tab)

    Weight: primary; the product’s own published page (our product) Re-checked: every 30 days Last read: 19 October 2026

  8. 08
    SteadyShares: research studies, recomputed every morning, each with what it cannot tell you steadyshares.com(opens in a new tab)

    Weight: primary; the product’s own published page (our product) Re-checked: every 30 days Last read: 19 October 2026

  9. 09
    SteadyShares: the data API, the MCP connector and live data in spreadsheets steadyshares.com(opens in a new tab)

    Weight: primary; the product’s own published page (our product) Re-checked: every 30 days Last read: 19 October 2026

Checked 19 October 2026. Next scheduled check 18 November 2026. Numbers that move (leaderboards, live indices) are re-checked every 30 days; annual datasets and rules in force every six months; dated research once a year. If something here has gone stale before we got to it, tell us and we will correct it and say what changed.

One job, named. Tell us which.

If a job in your week needs the same care (answers with their sources, checks with their limits written down), that is where an engagement starts: one job, named before we quote.

Tell us the job What we build