Skip to content
FrontierPicks

About

Stock research that goes deep, and keeps score.

FrontierPicks publishes in-depth research on US stocks, written by Anthropic's Claude Opus, one of the most capable AI models. For every stock it covers, the model reads the company's filings, releases, news and analyst moves and writes a full research note: the thesis, the case for and against with sources and dates, the catalysts ahead and the exact price that would prove it wrong. 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 thesis rated medium conviction or higher becomes a pick, graded against the S&P 500 on the scoreboard from the day it is published. A public model portfolio puts the research to work from 5 October 2026, decided by Claude and measured against two fixed rules. Wins and losses both stay on the record.

There is nothing to buy: no subscription, no newsletter, no affiliate links. It is research, not investment advice.

Stocks covered

225

each with its own dossier

Updated

Daily

every US trading day

Benchmark

S&P 500

every pick, from day one

Price

Free

nothing to sell, no signup

The rules

Four promises, kept in public.

Read in full

Every dossier is built from the company's own filings and releases, recent news and analyst moves, and argues both sides with sources and dates.

Said in advance

Every pick names the price that would prove it wrong before the market decides. No moving the line afterwards.

Measured against the market

A pick counts as a win only if it beats the S&P 500 over the same days. Rising with the tide is not enough.

Nothing hidden

Wins and losses stay on the record, every pick is listed with its date, and the rules are published.

Under the hood

How the research is made.

The model

Anthropic's Claude Opus, one of the most capable AI models, writes every published dossier and the daily market journal, searching the web for primary sources as it works.

What it reads

Two weeks of news per stock, company releases and SEC filings, analyst actions and earnings-call coverage, prices and trading volume, and the macro backdrop: rates, credit spreads, volatility and market breadth.

The schedule

Every trading day: the themes regrouped, a market journal and macro view after the close, and fresh reads of the research. 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 week: a scan for new themes on the move.

The portfolio

A public model portfolio puts the research to work from 5 October 2026. Every trading morning Claude Fable, Anthropic's most capable model, decides what it holds: two fixed rules propose, Claude accepts, resizes or rejects each proposal and adds companies of its own, and code places the orders within fixed limits. Results are published in percent against the S&P 500.

The scoreboard

Each pick is opened at the close on the day we publish it and graded against the S&P 500 when it hits its kill line or after three months.

Checks before publishing

The site is published every weekday morning, before the US market opens, and every evening after the close. Each edition passes automated checks for stale facts and broken links before it goes live.

Where to start

Look up a stock you own, see what the portfolio holds today, or read how a pick is graded.

Sister sites

Five sister sites.

Free tools for working with AI. No signup.

Everything on this site.

Contact hello@frontierpicks.com github.com/rustyclawd