Skip to content
FrontierPicks

Research through the 9 October 2026 close

Search

How it works

From a scan of the market to a graded pick, every trading day.

Code finds where the market is moving. Claude Opus, one of the most capable AI models, then reads each company's filings, releases, news and analyst moves and writes a full research note, including the price that would prove it wrong. The scoreboard and the portfolio show how those calls actually did. Five steps:

  1. 01 Scan
  2. 02 Group
  3. 03 Read
  4. 04 Commit
  5. 05 Score

  1. Scan the whole market

    Code

    Every trading day, code ranks 3,488 liquid US stocks by price trend.

    No model is needed to tell which stocks are moving. A cheap, rules-only pass narrows the field so the expensive reading goes where the action is.

    Inputs
    Prices · volume · liquidity
    Produces
    A ranked list of the strongest trends

  2. Spot the themes

    Claude Opus + code

    Claude Opus groups the strongest stocks, and the ones we hold, into themes — AI chips, nuclear power, the space economy — from their news and filings, and tags each theme as accelerating, maturing or saturated. A weekly scan of stocks whose prices move together looks for new ones.

    Markets move in groups. Reading a stock next to its peers tells you whether a move is the start of something or a one-off.

    Inputs
    The momentum leaders · news · filings · price correlation
    Produces
    The themes in motion

  3. Read the sources, write the note

    Claude Opus

    For each stock we cover, the model reads two weeks of news and the price history, searches the web for company releases, SEC filings, analyst actions and earnings-call coverage, then writes the dossier you see on this site.

    Each note argues both sides, and its facts carry their source and date. 147 stocks are re-read every trading day: every stock in the portfolio or with an open pick, plus those in accelerating themes, reporting soon or moving sharply. The other 78 are re-read on a rolling cycle, and none is more than 6 sessions old.

    Inputs
    News · prices · company releases · filings · analyst actions
    Produces
    A full research note: thesis, both sides, catalysts, kill line

  4. Make the call, in advance

    Claude Opus

    Every dossier ends with a conviction level and a kill line: the exact price that would prove the thesis wrong. A medium-or-higher conviction call becomes a pick on the scoreboard.

    Saying in advance what would change our mind is what separates research from commentary. The line is frozen the day the pick opens.

    Inputs
    The dossier
    Produces
    A dated pick with its kill line

  5. Invest, and keep score

    Claude Fable + code

    Every trading morning Claude decides the model portfolio, which runs from 5 October 2026: two fixed rules — filed revenue inflections the market confirms, and the strongest trends among the most-traded stocks — propose, and Claude accepts, resizes or rejects each proposal and adds companies of its own. Every pick on the scoreboard is graded against the S&P 500.

    A pick closes when it hits its kill line or after three months, and counts as a win only if it beat the S&P 500 over the same days. The portfolio's buys and sells are Claude's decisions; code places the orders within fixed limits.

    Inputs
    The research · the rules' proposals · daily closes · the S&P 500
    Produces
    The portfolio and the scoreboard

Why it is built this way

Every call can be checked.

Most stock commentary is impossible to hold to account: no date, no price, no condition under which it was wrong. Here every pick carries all three, and the scoreboard grades it against the S&P 500 whether it worked or not. And Claude's portfolio decisions are measured every day against two fixed, published rules valued the same way, so what its judgment adds can be checked.

The scoreboard The portfolio This week's theme scan

Common questions

How does an AI model analyse stocks?
Code first ranks the liquid US market by trend, and Claude Opus groups the strongest stocks into themes. Then Claude Opus reads the news, price history, company releases, filings and analyst actions for each stock we cover, and writes a dossier: the thesis, the bull and bear case, upcoming catalysts and the price that would prove it wrong.
Which AI models does FrontierPicks use?
Anthropic's Claude Opus writes every published dossier, the daily journal and the macro view, and makes the calls on the scoreboard. Claude Fable, Anthropic's most capable model, decides the model portfolio's trades. We always run the latest version of each.
How often is the research updated?
Every US trading day. 147 stocks are re-read every trading day: every stock in the portfolio or with an open pick, plus those in accelerating themes, reporting soon or moving sharply. The other 78 are re-read on a rolling cycle, and none is more than 6 sessions old. Every dossier shows the date it was last updated.
How are picks graded?
A pick opens at the closing price on the day we publish it. It closes when it hits its kill line or after three months, and counts as a win only if it beat the S&P 500 over exactly the same days.
Is the research written by people?
No. The dossiers, journal and macro views are written by the model and published automatically after passing our checks. Nobody edits the model's calls after the fact.

Today we cover 225 stocks across 17 themes. Go deeper:

Anatomy of a dossier See the dossiers Full methodology For agents