The investment thesis

Models became commodities. Memory did not.

Frontier model capability is available to everyone and improves on someone else's balance sheet. The durable, differentiated layer is the one that knows the person, and no model vendor can own it without owning the person's trust. That layer is Cortex Engine.

The problem

One buyer situation, repeated across the economy.

When people leave, what they know leaves with them, and they never got to keep it either. Every tool sold to fix it watches the individual for the institution, and people starve the watchers of the very signal they need. The category fails on trust.

Clinical

The nurse

Who holds the unit together. Replacing one runs $52k–61k [verify].

Wealth

The advisor

Retiring with twenty years of client judgment the firm cannot keep.

Public

The caseworker

Whose departure orphans dozens of family histories.

Trades

The lineman

Whose head is the manual, retiring in a silver tsunami.

The insight

Issue the memory to the person, not the platform.

The individual gets a companion they own; the institution buys the weather, plus the handoff a departing person chooses to leave. The buying motion already exists: institutions pay per seat for assistance programs they cannot see into, applied to crisis. This is the same motion applied to capability.

The person owns

The memory

A private companion holding their whole arc, sealed and encrypted, portable at the end. Switching cost compounds every week a seat is live.

~$1T / yr
lost to voluntary turnover in the US; replacing one person runs 0.5 to 2 times their annual salary (Gallup, 2019).
Detectable early
burnout is now a WHO-recognized occupational condition, readable in aggregate signal weeks before people leave (WHO, 2019).
$3.7B → $9.9B
the workforce-analytics market from 2024 to 2031, growing in the low teens (The Insight Partners, 2024).

Proof and moat

Three deployments. Each a configuration, not a rebuild.

LIVE

Cortex

Individuals

Sealed per-user memory proven end to end, external users onboarded.

LIVE

Grant Engine

Nonprofits

Human-approval trust layer shipped, three design partners engaged.

PILOT

Cortex for Campus

Higher ed

Pilot proposal in front of a state university now.

Why they cannot copy it

Model providers ship memory captive to their model and their business needs to see the data. Retention incumbents are built on watching individuals for the institution. To copy the estate split, either would have to break the business model that funds them. The trust position is structural, not a feature.

The model

Build once. Configure per vertical. Operate end to end.

Seats

Recurring

Sponsor pays per person per year. Scales with headcount, compounds with retention.

Configuration

One-time

Standing up the vertical: connectors, scoping, integration. A paid gate, written once, re-run.

Operation

Recurring

We deploy, monitor, and maintain. In regulated institutions this is why procurement says yes.

The raise

A seed round to industrialize the machine.

One founder and an AI-native build system stood up three vertical deployments, the security architecture, and this design language. The round retires key-person risk and turns replication speed into paying scale.

Team
Senior engineering to harden the engine and build the structural-privacy layer; deployment engineering to run pilots; institutional sales for the warm channels.
Compliance
SOC 2, penetration testing, and per-vertical counsel: the spine that institutional checks require.
Milestones
Cortex for Campus landed as the first beachhead, co-developed with a design-partner university, a named pipeline at coverage, two full cycles of outcome data, and the weather station shipped as product.

The vision

Every working person issued their own memory.

The institution funds it. The person owns it. The judgment that used to walk out the door compounds instead, for an entire career, from first semester to last shift.