Methodology

Source Curation

The process begins with people rather than screens. We interview traders, portfolio managers and industry engineers we respect and ask one question: who do you actually read? Names that recur across those conversations form the panel. The result is a small set of independent research desks, listed on the Sources page, each publishing primary work rather than commentary on other people's work. The panel is deliberately small, because adding a mediocre source dilutes the signal of a sharp one.

From Reading to Themes

We then read everything the panel publishes, not a sample. Each report is reduced to its claims, the claims are de-duplicated across shops, and the survivors are clustered. An idea that appears in one shop is an opinion; the same idea arrived at independently by several desks with different methods is a theme. Six clusters cleared that bar, and they are the six baskets of the index. Five come directly from the clustering; the sixth, Onchain Finance, is a Freeport house theme anchored on Hyperliquid, sourced from Capital Flows on the listed proxy and our own work on the rest.

Expression and Symbol Selection

For each theme we study how the authors themselves express the view: which instruments they name, on which side, and with what caveats. Where two credible desks take opposite sides of the same name, we lean toward the desk that discloses a real book and a real profit and loss, and we let price arbitrate the rest: a short thesis on a stock the market keeps rewarding is treated as unproven, and every short in the book is a genuine laggard rather than a brave call against a rising price. The output is a symbol list per theme, each name on the side the research supports.

Correlation and Optimization

With the candidates chosen, we run regressions on daily returns across the symbols to map how they move together, within each theme and across themes. Expected returns and upside are set from the published predictions of the credible sources themselves rather than from price targets of our own. The portfolio is then optimized against that correlation structure, so overlapping expressions of the same buildout do not stack silently into one oversized bet.

Sizing

Position sizes blend three weighting schemes. Market capitalization defers to market efficiency: the size of a company carries information. Equal weight defers to our own uncertainty: we do not pretend to know which expression of a theme wins. The thesis weight is where opinion enters: symbols that express a theme purely carry more than diluted expressions of it. The blend of the three sets each final weight, and the single strongest conviction in the book, Hyperliquid, is sized as a double-weight anchor.

The book runs approximately 100% long against 50% short, so it is structurally net long and the shorts fund a hedge rather than a bet on ruin. Where no short clears the bar, as in optics and foundry today, the theme runs long-only rather than shorting something for symmetry.

Factor Posture

The construction has consequences we accept rather than hide. Reading sharp practitioners makes the book long market beta, because the panel writes about businesses being built, not businesses failing. It makes the book long momentum, because people tend to write about stocks that are working. It makes the book long size, small over large: the panel covers far more than the mega-caps, and the equal-weight component pushes the same way. And it makes the book long growth, because the research today concentrates in technology rather than value. The realised betas published on the index page, roughly 1.9 to the market and 1.7 to momentum with positive size and growth loadings, are those choices showing up in the data.

Risk Targeting

The long and short legs are tuned so the book carries roughly two times market beta while the factor exposures above stay within reason. Volatility is targeted near 40% annualized, which is aggressive but not unreasonable for a concentrated thematic book. The arithmetic: dividing 40% by 16, the approximate square root of the trading days in a year, gives a daily standard deviation near 2.5%; multiplying by 0.8, the approximate correction from a standard deviation to an expected absolute move, gives an expected daily move of about 2%. A reader should expect the index to move about two percent on an ordinary day, and more when its themes are in the news.

Prices and Marking

Raw, as-traded closing prices in US dollars. The book is mostly US-listed equities and dollar-quoted crypto; the few foreign listings, the Korean and Japanese memory makers, convert at the day's exchange rate. Splits are applied as an append-only factor against the inception mark, taken from the vendor's corporate-actions feed rather than inferred from a price move. Returns are price-return: dividends, borrow cost, commissions and funding are all excluded, including any of them that would have helped.

The Chart

The page opens on a curve that runs back to the start of the year. Everything before the June 1 marker is a backtest, drawn dashed and labelled as such. The names were chosen with knowledge of which had already worked, so a line anchored to January inherits that hindsight. From June 1 the book is paper-traded and marked daily. Forward tracking removes hindsight from the returns; it does not remove it from the composition, and no methodology does. That is a real limitation, stated plainly.

What We Refuse to Price

A position we cannot price honestly is never swapped for a proxy nobody wrote about. The live examples are Lighter and Kalshi in onchain finance and ChangXin Memory in the memory basket: the thesis wants them, so they sit in the book at their intended weights, but they have no reliable public mark yet, so the performance line carries them flat until one exists. The chart never invents a number for them.

The Aggregate

The six baskets taken together are one book, approximately 100% long against 50% short. It is structurally net long and heavily semiconductors, with a sizeable onchain sleeve anchored on Hyperliquid and a set of price-confirmed shorts: the hyperscalers lagging the silicon they buy, the seat-based software that AI is repricing, the crypto majors with no cash flow, and the device makers paying up for scarce memory. Six baskets do not make it diversified: the themes rhyme, and most of the book leans on the same AI buildout from different angles.

We publish every position with its weight, the gross and net exposure, and the realised volatility and factor betas. A reader who concludes this is levered semiconductor beta with a crypto sleeve is reading it correctly, and should be able to reach that conclusion from the page rather than from a spreadsheet.