Skip to content
FrontierPicks

· an extra edition, outside the schedule

How it works

How our AI stock research and portfolio work.

Code finds where the market is moving and gathers each company's filed figures. Claude, an AI model made by Anthropic, writes a research note on each stock and, deciding alone, runs a concentrated portfolio on that research, measured against the S&P 500. Five steps:

  1. 1 Scan
  2. 2 Group
  3. 3 Read
  4. 4 Decide
  5. 5 Measure

  1. Scan the whole market

    Code

    Every trading day, code scans 3,405 liquid US stocks and ranks the 3,380 with enough price history by price trend. The scan decides which stocks the research notes cover and how often they are re-read; it does not pick stocks for the portfolio.

    No model is needed to tell which stocks are moving.

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

  2. Spot the themes

    Claude + code

    Claude groups the strongest stocks, and the ones the portfolio holds, into themes from their news and filings, and gives each theme a stage: accelerating, maturing or saturated. A weekly scan of stocks whose prices move together looks for new ones.

    Current themes include AI enterprise software, oncology and immunology and cyclical industrials. 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 strongest trends · news · filings · price correlation
    Produces
    The themes in motion

  3. Read the sources, write the note

    Claude

    For each covered stock, Claude reads the last two weeks of news, the price history and the figures code gathers from its SEC filings, searches the web for company releases, filings, analyst actions and earnings-call coverage, and writes the research note on the stock's page.

    Each note argues both sides and lists its key facts with their source and date. 17 stocks are read first: the portfolio's holdings and recent sales, and stocks in an accelerating theme. Each is re-read at least every four days. Every covered stock is checked every morning from Monday to Saturday for what changed. The other 1,108 are re-read within 5 sessions. A stock with new results, a filing or another dated event is re-read within 4 days of it.

    Inputs
    News · prices · filed figures · company releases · analyst actions
    Produces
    A research note: the thesis, both sides, the catalysts ahead and the kill line that would prove it wrong

  4. Decide the portfolio

    Claude

    Claude will rank every US-listed common stock and ADR that trades at least $5 million a day each week for the portfolio, on the second-to-last full trading day of each week, and research the best-ranked in depth. The first weekly ranking has not run yet. Each morning Claude will also rank the companies that have just reported. The first morning ranking has not run yet. Claude researches the stocks the portfolio holds the same way. Every trading morning Claude decides, alone, what the portfolio holds.

    Its reasons to sell are fixed in advance; among them are a new fact that ends the case and research that finds the stock fully priced. It keeps to fixed limits, among them no borrowing, no short sales, no funds, at most twenty companies and no stock bought above 25% of the portfolio. The full rules.

    Inputs
    These notes · the weekly ranking · web research · filed figures · its own record
    Produces
    A dated decision for every holding: hold, buy, add, trim or sell, each with its reason

  5. Measure it against the S&P 500

    Code

    Every stock the portfolio buys is a pick, measured against the S&P 500 while it is held and graded when it is sold. The portfolio as a whole is measured against the index from its first day, 10 November 2026.

    Since the record began on 10 November 2026, every pick has stayed on it, gains and losses alike. The results go back to Claude with every decision, including what each sold stock did afterwards.

    Inputs
    The prices it bought and sold at · daily closes · the S&P 500
    Produces
    The portfolio and its track record

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 research note is dated and says what would prove it wrong. Every buy, add, trim and sale is published with its date and its reason. Every pick is measured against the S&P 500 whether it worked or not.

The track record The portfolio This week's theme scan What changed today

Common questions

Which AI models does FrontierPicks use?
Claude, the AI model made by Anthropic, writes every published research note and decides the portfolio. Which Claude model does which job is on the methodology page.
When does the portfolio sell a stock?
Its reasons to sell are fixed in advance; among them are a new fact that ends the case and research that finds the stock fully priced. Every reason to sell is listed on the methodology page.
How is a sale graded?
On the day the portfolio sells a stock, its return, from the prices it was bought and sold at, is compared with the S&P 500's over exactly the same days. That is its final result, and it counts as a win only if it beat the index.
Is the research written by people?
No. The research notes, journal and macro views are written by the model and published after passing our checks. Nobody edits the model's calls after the fact, and nobody but the model decides what the portfolio buys or sells.

Today we cover 1,125 stocks across 18 current themes. Go deeper:

Anatomy of a research note Browse the stocks Full methodology Tested patterns For agents