Six positions. Every thesis on the record.

Concentrated by design: "safe" compounders paired with volatile high-upside names. Weights and cost bases as of December 31, 2025. Dollar position sizes are shared with paid subscribers only.

META Meta Platforms

Weight42.42%

Cost basis$93.67

2025 return+12.74%

The largest position and the highest conviction: distribution wins AI. Meta is the world's largest distribution platform, and it monetizes AI through ads and engagement already — no waiting on AGI timelines. Bought at $93 in the 2022 crash (down from $336). The acknowledged risk is open-ended capex — $600B in data-center commitments — but unlike cloud hyperscalers, Meta can pull back anytime, as the 2023 "year of efficiency" showed.

What would change my mindIf engagement or revenue-per-user starts declining the way it did in 2022, or if AI spending becomes structurally unproductive with no path to pull back — the distribution-moat thesis breaks.

HALO Halozyme Therapeutics

Weight19.46%

Cost basis$53.26

2025 return+40.77%

The favorite thesis in the portfolio. Halozyme leads injectable-drug delivery and is acquiring competitors to extend the franchise. It was called out as a stock that "should be at least $100" — at the end of 2025 it traded at a P/E of just 14 while delivering a "double beat on earnings every quarter." The patent-cliff fear is considered way overblown. This is also the bubble-resilient pick: medicine profits whether AGI arrives in two years or ten.

What would change my mindIf the acquisition spree destroys rather than extends value, the every-quarter double-beat pattern breaks, or the patent-cliff risk turns out to be real after all.

NET Cloudflare

Weight11.77%

Cost basis$40.08

2025 return+83.09%

The post-search internet bet: as agents and LLMs reshape how people find things online, Cloudflare is best positioned for the AI internet. Up 83% in 2025 — "I don't own enough of this stock." Held through the February 2026 software selloff on the view that AI disruption accrues to its platform rather than destroying it. A long-term hold despite the premium valuation.

What would change my mindIf AI agents commoditize the network layer instead of accruing value to it — or if growth ever stops justifying the premium multiple.

NTRA Natera

Weight10.94%

Cost basis$147.35

2025 return+44.72%

Called "the largest beneficiary of AI in the market right now" — scale plus distribution plus data, growing revenue 30%+. Bought in November 2024 on the view that biotech, not chatbots, is where AI makes real progress. One of the two biotech picks that beat the ARK Genomics ETF by a wide margin in 2025.

What would change my mindIf the 30%-plus revenue growth stalls or the scale-and-data advantage stops converting into durable market share.

U Unity Software

Weight7.39%

Cost basis$28.18

2025 return+96.57%

The roller coaster: bought at $28 in late 2022 after the fall from $197 to $15, now a turnaround bet under CEO Matthew Bromberg. Unity's engine reaches some 4 billion monthly users, and the monetization upside is in-game ads — "your margin is my opportunity" against Applovin's 65% net margins. AI making content creation easier means more games, and more games need engines. The stated sell range is $50–$55 (later raised toward $55–$60).

What would change my mindThe plan already includes the exit: $50–$55 is where this gets sold. Before that, a stalled turnaround or a broken monetization story would end it early.

CELH Celsius Holdings

Weight4.34%

Cost basis$46.07

2025 return+68%

Bought at $46 on a −22% earnings crater in November 2025 — "another roller coaster, much like $U" — then added to in January 2026 with the proceeds from selling Microsoft. The thesis: high margins, huge growth, and the secular shift from coffee to energy drinks. "Invest in what one consumes." The top "cult stock" buy-list pick for 2025.

What would change my mindIf the energy-drink secular trend reverses or the high-margin, high-growth profile compresses structurally — the crater-buy only works if the business was fine.

Recently exited. Microsoft (~3.6% weight) was sold in January 2026 and folded into Celsius — impatience with a "mundane" compounder. Occidental was sold at $42 in late 2025 on a broken oil thesis; it then ran 54% — a published mistake. AMD and Intel were sold before both ran. The exits get the same disclosure as the entries.

Six positions. Every thesis on the record.

See every buy and sell the moment it happens.