The Coordination Book.
P4 governs the decision, transaction and physical action. P3 supplies the watts.
Where we are on the S&P 500.
AMBER+ — Stage A active / Stage B unconfirmed. Phase 2 remains SPX 7,197 or lower.
Macro determines sizing speed. It never rewrites company valuation zones.
The five stages of the AI cycle.
P1 and P2 produced the early returns. They are cycle indicators now, not the destination for new capital. P4 is the receiving phase. P3 remains because all coordination consumes power.
Memory
Scarce compute input. Contracted memory remains a cycle sentinel. The permanent book no longer needs cyclical concentration here.
Shells / interconnect
Physical infrastructure and optical buildout. Useful historically. Crowded and cyclical. Not a strategic receiver.
Power
The operating physical floor. Licensed nuclear, dispatchable generation, onsite torque. Permanent because P4 still requires real electricity.
Coordination
Operational control, financial coordination, physical action. The businesses that govern intelligence, authorize economic action, and convert intent into movement.
P4-BMA · ICE
P4-CUBER
P4-DORCL · GOOGL
Physical AI
Robotics, autonomy, real-world compounding. Uber is the P4 gateway into this phase — not a separate sixth AI book.
What we want to own.
The fund is no longer primarily trying to own every layer of the AI buildout. The permanent book is migrating toward the businesses that govern intelligence, authorize economic action, coordinate physical action, and own the durable power floor underneath the system.
Crisis cash floor is 15%. Unfilled target weight remains cash in addition to that floor.
Intelligence becomes investable when it creates governed economic action.
Twelve permanent names, grouped by phase. Target is destination. Funded is what the ladder has authorized. A shortfall is not permission to chase.
Grey gap = destination exists, price not open. Amber = company gate. Green = a zone is live. Gap is not an instruction to buy.
Why each name is in the book.
Role, moat, price, what evidence unlocks the next tranche, and what kills the seat.
Next legal actions.
Stage A is active: R0 only, 0.25%–0.75% initial entries, DAY orders. Macro sets speed. It does not rewrite company zones.
No new capital.
P1 and P2 names stay on the book until harvested. Stops still apply. Proceeds go only into an approved permanent-core entry — otherwise they remain cash.
Is P4 actually monetizing?
The payment is the receipt proving the agent has created economic value. This belongs in P4 — not a separate sixth AI phase.
MA = permanent core trust seat · Visa = reserve · COIN / NET / FIGR = research watches, not permanent core yet.
Green only when the metric is economic value, not narrative activity. Missing fields render as em dash.
Validators influence size, never price.
Independent conviction · specialist conviction · crowding · strategic conflict. 13F is a confidence modifier, not an entry signal.
Two eras. One account.
Buildout era · P1–P3
Memory · shells/interconnect · power.
Coordination era · P4
Operational control · financial coordination · physical coordination.
P1–P3 created the fund’s early returns. P4 is where the fund expects durable organic compounding to migrate as AI moves from capex to economic action. P3 remains because all coordination ultimately consumes power.
Contribution by P4 / P3 / transition / cash is not in this export — rendered as em dash, not computed here.
What we actually did.
Trades and rule changes are separate ledgers. No after-the-fact rule makes a prior deviation look compliant.