The scorecard, losers first.

Most research publications show you their winners. Our method says credibility compounds from admitted misses faster than from claimed hits, so this page leads with what went against us. Portfolio-level figures are from our brokerage's statements and analytics for the publisher's own account, year to date 2026, with the as-of date noted on each figure; itemized position figures are verified against our position records before they appear. All figures are shown as percentages.

The year so far

Two measures, both published. Time-weighted return (+17.1%) is the industry standard for judging management: it strips out the timing of deposits. Money-weighted return (+14.7%) includes that timing, and ours is lower; the roughly two-point gap is largely attributable to the timing of our own contributions, though the two figures carry different as-of dates (August 28 and August 30) and the gap is an approximation, not an exact decomposition. We publish both because picking whichever number flatters is exactly what this page exists not to do. The money-weighted series peaked near +24.7% in the week ending June 19 and has given back meaningful ground since; that drawdown is part of the record too.

Against the indexes

The comparison a track-record page owes you, printed rather than implied. Two honest caveats: this book is roughly 97% equities, concentrated in AI-adjacent themes, so the Nasdaq-100 is the harder and fairer comparison, and a two-month-old margin over it is weather, not climate. A third: the rows above carry slightly different as-of dates (the book and the S&P 500 through August 28, the Nasdaq-100 through August 26), which makes the comparison approximate; at the next update all series will be marked to a single date rather than restated here without their sources. We will keep printing the rows either way, because omitting the benchmark is how newsletters flatter themselves.

What went against us

Correction, August 30, 2026: earlier versions of this page quoted per-position returns from our broker's movers report, including OKLO at −90.8% and RKLB at −78.3%. Those figures are per-symbol money-weighted returns, which weight the timing of every buy and sell and can differ sharply from a position's simple total return: verified against our position records, RKLB is −42.3% and OKLO is −23.5%. Itemized rows on this page now use simple total return verified against primary records; money-weighted math is reserved for the portfolio headline, where it is labeled. This note stays up because corrections belong on the record too.

A position's return is not the same thing as a thesis's quality: entry timing, instrument selection, and sizing all show up in these numbers alongside the idea itself. We print them anyway. If we only showed you the research and never the book, you would be right not to trust either.

What went for us, not yet itemized

We had individual winners this year. We hold itemized winners to the same verification bar as the losers above, and they have not cleared it yet: closed positions need the realized-performance report, not a screenshot. Until they clear, we claim no number for any of them. The losers cleared first, on purpose.

The rest of the book

Across the forty rows of our broker's movers report for the period (cash included), 24 show gains and 16 show losses. We quote only the direction of those rows here: the report's percentages are money-weighted per symbol, which folds trade timing into each figure, so magnitudes get restated as simple total returns and verified against position records before they are published. If this page's figures ever stop reconciling to primary records, the correction gets published in the same place, like the one above.

The conditions behind these positions are tracked at The Conditions: 29 published, all scored.

Where the book leans

Portfolio theme exposure as of August 30, 2026, per our brokerage's classification. Themes overlap, so weights do not sum to 100.

The account is roughly 97% equities and 3% cash as of August 30, 2026.