How it works

The idea

Two forces are reshaping what holds value. Governments carry more debt than they can repay honestly, which tends to erode paper money. And AI is starting to do work people are paid for, which erodes the value of labor. The letter looks for things that survive both: scarce real assets, the parts of the economy AI needs rather than replaces, and corners of the market few people are watching yet.

The second half matters as much as the first. A good story that everyone already owns is a crowded trade, not an opportunity. So every asset is read for whether it has already run.

The weekly table

Every weekend, each asset class on the table is read on its own prices, the same way each week. Two columns carry most of it.

• Stage asks whether the long run shows excess or a crash. Normal means no sign of either, which is most of the time. Stretched and Late mean a big run is already behind it. Post-peak means it ran up and then fell hard. Recovering means it has climbed back above its trend after that fall.
• Timing reads the 200-day average price, for the next one to six months. Add means the trend is up and leading. Hold means keep what you have, no new money. Wait means below the line. Reduce means below the line after a big run, so cut back.

Neither is a forecast. Tested on US industries back to 1927, they are much better at flagging trouble than at picking winners: below the 200-day line the odds of a sharp fall roughly double. Positioning and sentiment are shown as background, because they did not sort later returns in our tests. Each table explains itself in a fold-out underneath it.

The lists

Separate screeners rank individual names: small-cap stocks, large caps, coins, graded cards, NFTs and AI collectibles. Each scores names on a short list of fixed traits and marks the best as Strong fit or consider. These are orderings, not forecasts, and not recommendations to buy. They are where calls start, not the calls themselves.

Calls and the ledger

A call is what goes in the paid letter: a specific thing to own, with a size, an entry, the conditions that would end it, the strongest case against it, and whether the author holds it. At most 2 new calls go out in a week, and a week with none says why.

Every call is scored from a price a reader could actually have got, not the price when it was written, and against a benchmark fixed on the day it went out. Every one goes on the ledger 30 days after subscribers get it, or as soon as it closes, winners and losers alike. Nothing is edited after the fact; a mistake gets a dated correction.

The author's own trading

The author does not trade a named asset from 3 trading days before publication to 3 trading days after. Where readers could move the price (small caps, small-cap crypto, memes, cards, NFTs, AI collectibles), the window is 5 trading days either side. Every call says whether the author holds it. No company pays for coverage.

What it doesn't claim

Nothing here predicts prices. A rule earns a place on the site only if it held up in a test; what didn't hold up is shown as background, labelled as such, or dropped. Old numbers are never carried forward: a reading too stale to trust prints as "no read".

Rules change only going forward. A change applies from its date, and no published call is rescored under it.