Approach
How the thesis becomes positions: staged buying, sizing and trim rules, valuation discipline and hedges, tested against the S&P 500, QQQ, a 60/40 portfolio and Bitcoin.
How we invest
- Staged buying. Target weights are not filled on day one. We deploy in tranches over roughly a year, and undeployed capital sits in T-bill funds, so the live book can sit well below target weights while we learn.
- 3pp trim rule. If a long drifts more than three percentage points above its target weight, we trim it back (structural cash sleeves are excluded).
- Starter positions. New names usually begin as starters; a drawdown of about 15% triggers a review rather than automatic averaging down.
- Delayed publishing. The public portfolio, performance and changelog files use a three trading-day lag, with weights in per cent only.
- Changelog. Buys, sells, trims, tranches and thesis corrections are logged in the changelog so outsiders can see when we change our mind.
We also invest in private companies; those books are not detailed here.
Valuation discipline
New ideas pass a six-criterion rubric (0–5 each; pass at 20 of 30): worldview fit, lasting rent capture, how much is already priced in, country tilt, low overlap with the existing book, and liquidity or exit. Crowded winners can stay on a watch list. We prefer paying for scarce bottlenecks over chasing last year's best chart. New positions are funded from cash first, not by selling winners.
Multi-asset long/short, and why we hedge
The public book is multi-asset: equities, commodity ETFs, REITs, short TIPS, AAA loan funds and T-bills, plus a trimmed short sleeve. Gross exposure can exceed 100%; net stays long.
We hedge because the base case can fail in correlated ways:
- Duration: short long-dated Treasuries (TLT) while term premium stays elevated.
- AI eats services: short traditional IT outsourcing and consulting (INFY, WIT, ACN), where generative AI compresses seats and hours.
- Ballast: gold, T-bills and short inflation-linked bonds for bust, money-printing and shock scenarios. We avoid long bonds.
Shorts are sized well below longs, scale with the live book, and must earn their keep as hedges. In November 2026 we cut the short sleeve roughly in half, dropping office REIT, high-yield credit and fitness-hardware shorts that were dragging returns without hedging much.
Backtests
Hypothetical, with hindsight. Backtests use Yahoo adjusted closes with monthly rebalance, 5bps costs and 1% short borrow. Tickers were selected in Nov 2026 — severe hindsight/survivorship bias. Private books are proxied by T-bill ETFs (v2) or BIL/PSP (v1). Not investable NAV; not investment advice.
Whole fund v2
2007-05-30 to 2026-11-05 · as of 2026-11-05| Series | CAGR | Volatility | Sharpe | Max drawdown |
|---|---|---|---|---|
| Whole fund v2 | 9.8% | 11.0% | 0.73 | −26.0% |
| 60/40 fund (AOR) | 7.8% | 11.8% | 0.52 | −24.4% |
| Nasdaq-100 (QQQ) | 14.9% | 22.1% | 0.65 | −53.4% |
| S&P 500 (SPY) | 10.1% | 19.4% | 0.49 | −55.2% |
| 60/40 | 7.7% | 11.9% | 0.51 | −34.7% |
| Bitcoin | 29.6% | 51.1% | 0.73 | −83.0% |
| Window | Total return | Volatility | Sharpe | Max drawdown |
|---|---|---|---|---|
| Trailing 12 months | 8.7% | 12.1% | 0.57 | −7.8% |
| Trailing 24 months | 33.2% | 11.8% | 1.08 | −9.8% |
Whole fund v1
2007-05-30 to 2026-11-05 · as of 2026-11-05| Series | CAGR | Volatility | Sharpe | Max drawdown |
|---|---|---|---|---|
| Whole fund v1 | 8.4% | 9.2% | 0.70 | −21.1% |
| 60/40 fund (AOR) | 7.8% | 11.8% | 0.52 | −24.4% |
| Nasdaq-100 (QQQ) | 14.9% | 22.1% | 0.65 | −53.4% |
| S&P 500 (SPY) | 10.1% | 19.4% | 0.49 | −55.2% |
| 60/40 | 7.7% | 11.9% | 0.51 | −34.7% |
| Bitcoin | 29.6% | 51.1% | 0.73 | −83.0% |
| Window | Total return | Volatility | Sharpe | Max drawdown |
|---|---|---|---|---|
| Trailing 12 months | 7.9% | 11.9% | 0.51 | −8.3% |
| Trailing 24 months | 31.9% | 10.9% | 1.11 | −8.3% |