“The person that turns over the most rocks wins the game.”
We are a research firm that happens to manage capital. We built a system to do exactly that — turn over more rocks than any desk could by hand — pairing the wisdom of Buffett and Munger with cutting-edge machine learning, spending compute to read every filing and recalibrate as the market changes.
Our process leans on a non-trivial — and ever-growing — amount of compute, and we pay for that entire stack ourselves. Clients carry none of it: beyond the stated fees, the infrastructure is on us.
The large majority of both Wyatt’s and John’s net worth is invested in the fund, beside our clients. We take the same gains and the same drawdowns, on the same terms.
The research engine never idles. Around the clock, our compute fleet is generating and stress-testing new quantitative models, refining our in-house deep-research agents, re-reading filings, and recalibrating the live book as markets move. What you see below is that work in progress, live.
See exactly what the fleet is working on right now — a live look at every accelerator and the job it is running, just below.
A look at the pipeline behind every prediction: each public filing is read by our language model, then fused with fundamentals, macro, and pricing in the ensemble to resolve a forward return-probability for the name.
Cumulative time-weighted return over the selected window, from the strategy's daily return series, reconciled to the broker's reported time-weighted return and benchmarked against the S&P 500 (SPY). The solid line is net of fees, after a 2% annual management fee assessed monthly and a 20% performance fee on profits assessed quarterly over a high-water mark; the dashed line is gross of those fees. Dots note each model deployment and strategy change — hover any point or changelog entry to read it. The 2-and-20 is an illustrative fee structure applied to the gross record; certain figures may be unaudited. Past performance is not indicative of future results.
| Period | Gross | Net of Fees | S&P 500 |
|---|---|---|---|
| Live (Jan 2026 – Jun 2026) | +17.3% | +16.4% | +8.8% |
| Max Drawdown | -12.4% | -12.5% | -9.1% |
Net of fees sits beside gross on identical horizons, benchmarked to the S&P 500. The net figure carries an illustrative 2% management and 20% performance fee, assessed over a high-water mark and modeled on the structure the fund is actually billed under. Toggle the view above to compare the full record with the period since the fund opened.
A first-principles research compulsion. He built the engine on the back of a decades-long software engineering career and guards the quiet needed for portfolio construction and model calibration.
wyatt@terrapenecapital.comOperational control, LP relations, and regulatory rigor.
john@terrapenecapital.comWe publish sparingly — preferring to communicate openly with each of our partners individually. What follows is a small selection of older pieces, generated at various times throughout this journey. They still say a good deal about how we thought in the past.
A framework that marries Transformer architectures to Graham-and-Dodd value investing — the QFig and SWA models, how they are trained, and an honest account of where they break.
Read →Lessons from The Outsiders and 100 Baggers applied to modern software, with Constellation as the lens and Box and Zscaler graded against it.
Read →