Position Notes
How I pick what to look at, and what I'm currently watching.
Four things I look for, in roughly this order. A company doesn't need to be exceptional on all four, but weakness on more than one is usually disqualifying.
Founders or operators with a track record, skin in the game, and a clear sense of what they're building toward. This is the hardest of the four to assess and the one I weight most heavily — a mediocre business under great management can turn around; the reverse rarely happens.
A big total addressable market — enough revenue potential that the company doesn't hit a ceiling from a shrinking or niche market before the thesis plays out. I'd rather own a small share of something huge than a large share of something small.
Real free cash flow or EBITDA, or — for earlier-stage names — a large enough cash pile that the business isn't at the mercy of capital markets to survive the next few years. Optionality matters more than profitability on day one, but the company needs a runway that doesn't depend on someone else's generosity.
Are they actually doing it — growing revenue, shipping a better product, taking share? Vision and TAM are cheap; execution is the evidence the first three points are real. This is the one I keep re-checking after initiating a position.
These are my own criteria and notes, published for reference — not a recommendation to buy, sell, or hold anything.
Each name has its own write-up, argued against the four criteria above with the specific catalysts and risks I'm tracking.
A 29-company venture portfolio — Anthropic, SpaceX, Databricks, OpenAI. The thesis is less about any one company's execution than about reported NAV lagging where the portfolio is actually marked.
Live NAV model →The thesis has moved off cars. Real free cash flow funding two speculative bets — unsupervised autonomy and Optimus — while the auto business decelerates underneath.
Deep dive →Best execution record on this list. Starlink cash flows today, Starship cost-per-kilogram as the non-linear variable, plus orbital compute and lunar infrastructure as free optionality.
Deep dive →$757M pro forma backlog and revenue up 605% — funded by a 324% increase in share count. The entire question is whether backlog converts faster than the float grows.
Deep dive →$3.1B cash funding a roadmap to fault tolerance. The 256-qubit system is the near-term proof point; SHIELD/Golden Dome is the credential that puts quantum inside defense procurement.
Deep dive →Repositioning from probe maker to chip licensor. Two Embedded partners — Midjourney and Aleph Neuro — shipped working hardware eight days apart in June, on the same silicon.
Deep dive →