Holdings

Data verified: September 21, 2026 close

"Positions" counts every open ticker in the live book, including index funds and the leveraged sleeve — the same 32 used throughout this page and site-wide. "All-time" is the account's real, dollar-weighted return since the $500 start in 2022, verified via Blossom and disclosed in Letters — the same +72.14% quoted on the homepage and Tear Sheet, not a separate cost-basis calculation. The Performance Tracker's own "All-Time (tracked)" figure will read differently from this one; that page explains why in its own methodology note, rather than these two figures quietly disagreeing.

Download as PDF Download Full Datapack (.csv) Download Portfolio Monitoring & Returns (.xlsx) →

Download the full Portfolio & Watchlist Research Report (.docx) → Download the Sector Landscape Overview (.docx) →

View the Performance Tracker: real daily return vs. SPY / QQQ / IWM →

The comps workbook groups all 41 operating companies across the book and watchlist into 9 sector peer groups (EV/EBITDA, P/E, PEG, margins) sourced from stockanalysis.com / S&P Global Market Intelligence as of July 13–14, 2026 — see the workbook's own Methodology tab for sourcing and limitations. Every position and watchlist page below also links its own real 3-statement model (Income Statement, Balance Sheet, Cash Flow, 5-year projection) where one exists; index funds (SPY, VOO, QQQ, SCHD), the leveraged NBIL ETF, and privately-held SPCX have no comparable fundamentals to model.

Exposure Dashboard

Aggregate sector risk, at a glance — the donut below and the exposure bars further down show the same 32 positions two ways. The point is an integrated view of the book: diversification by ticker is not necessarily diversification by economic driver.

In plain terms, a beta of 1.79 means that over the trailing regression window, this book has historically moved roughly 1.79x as much as SPY in both directions — up more on a good day, down more on a bad one. It's a measure of historical volatility relative to the market, not a prediction of future returns; full methodology below.

Factor tilt

What kind of book this actually is, stated in numbers rather than adjectives — see the Investment Policy Statement for the reasoning behind leaning this way.

"GAAP-profitable" and "pre-profit" are read directly off each position's own Fundamentals table (a blank or "N/A — unprofitable" P/E line counts as pre-profit; a real trailing P/E counts as profitable). The 121.7x weighted P/E is computed only across the profitable 57.2% of the book — blending in the pre-profit names would require either excluding them (which is what's done here) or assigning them an arbitrary placeholder multiple, and a made-up number is worse than an honest gap. Read together with the 0.24% weighted dividend yield above, the picture is a book built for capital appreciation from companies still reinvesting rather than paying out, not an income portfolio.

Sector groupings are mine, not an official GICS classification — see the By sector section below for the full methodology note. Portfolio Beta methodology: weekly closing prices for all 32 positions, pulled via Massive Market Data from January 2025 through July 2026 (up to 80 weekly observations per name, fewer for names with shorter trading histories — SPCX, NBIL, WYFI, and CRWV all IPO'd or launched after that window opened). Each position's individual beta is a population-covariance regression of its weekly returns against SPY's weekly returns over the same overlapping period; the 1.79 figure above is those 32 individual betas weighted by current allocation %. A beta above 1.0 means the book has historically moved further than SPY in both directions — consistent with a portfolio concentrated in semiconductors and leveraged/early-stage growth names. This is a trailing, not predictive, figure, and a single regression window is a limited sample; it will be recomputed periodically rather than treated as a fixed constant.

Macro Stress-Test

How the book is actually positioned under four scenarios — including the ones it's genuinely exposed to. This isn't a claim that every scenario is defended; the point is naming the real vulnerability alongside the real mitigant, not just the parts that sound good.

Portfolio composition

The Interactive Holdings Matrix below is the primary record: every position, its sector, live allocation weight, average cost per share, return, my own conviction rating, and compliance status. Click any column header to sort.

Allocation and return use the published share counts and prices through September 18, 2026 close; average cost is the per-share basis recorded in the portfolio file. Conviction is my own subjective rating (High / Medium / Low), not a statistical output. Status flags active thesis or process review; SNDK is above the 10% limit and remains Under Review on the Compliance Ledger.

By sector

The same 32 positions, grouped by sector rather than bucket — the lens the Exposure Dashboard's donut chart above already uses, restated here as bars next to the holdings table for a direct read.

Sector groupings are mine, built from each position's own fundamentals page — not an official GICS classification, since I don't have a licensed data feed for that. "Software & cybersecurity" and "AI / cloud infrastructure" are drawn as separate slices from "Semiconductors" here even though all three lean on the same AI-infrastructure theme discussed in the July letter — see the correlation matrix below for how correlated they actually are day to day.

Return contribution

Not every position that's up a lot has actually moved the book — a small position with a huge return contributes less than a large position with a modest one. This ranks each holding by today's weight × its total return, the top 10 contributors to where the book stands overall.

Contribution = current allocation % × total return since cost, in percentage points. This is a simplified attribution using today's weights, not a precise time-weighted decomposition — weights shift as positions compound, so this ranking will not sum exactly to the account's own +72.14% all-time dollar-weighted return; treat it as "which positions are doing the heavy lifting right now," not a reconciled performance bridge. SNDK alone accounts for roughly a fifth of the book's gross unrealized gain despite being just over a tenth of current weight — a concrete illustration of why its position size gets so much scrutiny elsewhere on this page.

Dividend & income

This book is built for growth, not income, but it's worth reporting honestly: the weighted-average dividend yield across the 87.57% of the book with a verified yield figure is 0.24%. SPY and VOO are excluded from that figure — Blossom shows a fundamentals table for neither on this site yet, and I'd rather leave them out than guess. The real income here is concentrated in a handful of positions: SCHD (3.25%), CAKE (1.51%), EL (1.70%), and AVGO (0.69%) do almost all of the work; the AI-infrastructure and semiconductor growth names pay nothing.

Correlation matrix

The July letter argues that much of this book leans on one macro bet — AI infrastructure, semiconductors, and power — and that if the theme cools, many positions fall together. Here's that argument made checkable: real pairwise correlation of daily returns (year-to-date) across every position except SPCX, which has too little trading history to compute a meaningful number yet. Darker cells mean two positions have moved together more closely; lighter cells mean they've moved more independently.

Computed from each position's own YTD price history (resampled to a common daily grid), Pearson correlation of daily returns. Grouped by theme, not alphabetically, so you can see the clustering rather than have to hunt for it — notice how the semiconductor block (SNDK–ARM), the cybersecurity block (PANW–CRWD), and the AI-infrastructure block (NBIS–GOOGL) each run noticeably darker against each other than against the non-theme names (META–TEM) or the index funds (SPY–SCHD).

AEA Compute & Memory Basket

An internal tracking basket, not a holding you can buy — a way of watching the compute/memory theme underneath several positions in this book move as a single unit, the way a desk might track an unofficial internal index. Equal-weighted across four names; two are current positions, two aren't held but sit squarely in the same theme.