They forgot you on Monday. This does not.

They are building bigger windows. We built a mind.

Write a fact once. Get the same fact back. On your hardware. That memory becomes every product we ship — the archive, the residents, the seats, the file door, the language.

Long-term vision

Land the mind. Prove it. Then scale it.

First customers own the archive. Three logos or $15 million booked opens the next raise. Until then the factory stays on the product.

Phased buildout

Three phases from first install to Series A

01
Seed · $20M

Factory + first proofs

Get the first enterprise customers live, prove the product, then launch the rest of the lineup on that same archive.

$7M R&D · $3M air-gap · $3M enterprise · 6–16 week proof

02
Gate · Phase 0–1

First logos

Wave 0 Memory in the building. Three signed logos or $15 million booked. Either prong. Later products inherit the archive.

CORE $2M · DIVISION $8M · ENTERPRISE $25M · proposed, not live

03
Series A · $250M

Scale after customers

Series A is a scale raise after signed customers. Not a seed story. Seed enables the trigger. Series A funds the scale.

$80M run-rate or 15 logos + $15M task order

The getWrite once. Get the same fact back.
The traveler

Process-aligned doors for the $20 million room

Hover a layer. The hologram fades in. Eleven book beats and twenty-six paper doors — the mind, the seats, the factory, the next physics.

Seed ask
$20M
Product / deploy / evidence / credits
85%
Factory, air-gap, enterprise, clocks, credits — not a street campaign
Use of funds

$7.0M factory · $3.0M air-gap · $3.0M enterprise · $2.0M clocks · $2.0M credits · $1.5M IP · $1.0M sales · $0.5M reserve

The factorySix weeks to proof on their floor.
Why the unit cost wins

Eighty-five percent of the raise is the factory

Per recall

450,000× vs rented RAG

We do not meter the past as tokens. Per recall is about 450,000× cheaper than GPT-4 class RAG — a constructed comparison you can re-run.

Both clocks

13.834 µs and 5,127 µs

Isolated recall on a quiet machine, and a loaded miss against a 200 µs budget. We show both. Neither is the whole story.

The ask

85% factory, proofs, credits

Sales is $1.0 million — two AEs plus counsel. Unused pilot credit is the first line we stop if the gate misses.

Substitution

The invoice the CIO already defends

A 10,000-employee stack is already $4–8 million a year. We take that bill. Not a share of a national appropriation.

Refusal

Fail-closed at ϑ = 0.70

Below threshold the archive returns not found. It does not mint a polite cousin of the stored fact.

Custody

Their hardware

When the contract ends, the memory stays. That is why incumbents have to start over.

Why this factory

We sell a runtime on their floor

Custody

The archive they own

Proofs install sovereign memory — not a chatbot demo. Customer GPU rooms and SCIFs stay the customer’s capital.

Repeat

A factory, not an API key

Six, ten, or sixteen weeks. Same product. Wider archive. Without the factory, the sales desk has nothing to put on the floor.

Gate

No public brand until logos

Three signed logos or $15 million booked. Press and waitlists do not count. Leftover seed does not become outdoor media.

Their floorThe archive lives on the buyer’s hardware.
Now · 24-month clock

Four windows. One gate. Then Series A or a bridge.

Months 0–6

The factory

CORE proofs on a six-week clock. Named-account NDAs. Independent validation starts on the $2 million evidence line — pass or fail, published as measured.

Months 6–12

First installs

CORE converts — or the analog gets a real number. DIVISION motion starts. The honest read is whether an install exists on customer hardware.

Months 12–18

ENTERPRISE path and the gate

Sixteen-week proofs. Air-gap default for the buyer who cannot send tokens off-premise. Gate: three logos or $15 million booked.

Months 18–24

Series A or a bridge

Hit the gate and prepare $250 million. Miss it and raise a bridge against a rewritten account list. Unused seed does not become a street campaign.

The room

Two models. $20 million ask. Inherit the archive.

Pitch Deck is the long visual. Room deck is nine slides on the board. The book is twenty pages. The product is the archive they own.

$20 million is an ask. No customer logos yet. Forecasts are a plan. Full papers are diligence.