MOTHERSHIP

[work] company brain

The Exocortex.

A knowledge system that organizes itself.

Capture is easy; every second brain dies at retrieval. The Exocortex is what comes after capture: a context layer with retrieval engineered per dataset, scheduled synthesis that digests the stream, and delivery that hands the right context to the right mind at the right moment. 130+ days as the working memory of a real multi-company operation.

130+ days unattendedzero cold starts4 runtimes, one memory~$0.20/day curation bill

THE STREAM

sessions · captures · commits · journals, arriving ambiently. Nothing depends on human discipline

ONE CANONICAL STORE

state · knowledge · memory · operating rules
retrieval engineered per dataset
curators synthesize on a schedule
versioned, typed, permissioned. Owned by no single model or app

EVERY SESSION STARTS WARM

four runtimes read the same memory and open with context loaded; the human's brief is waiting before anyone asks. Every session writes back to the stream

[01] the real problem

The problem was never capture. It was what comes after.

How do you get the right sliver of accumulated knowledge to the right mind, at the right moment?

A pile of notes does not think. A bookmark inbox does not surface the idea you needed three months later. The system we built answers the question in four movements: capture without ceremony, a true context layer, synthesis on a schedule, and handoff at the right place and time.

THE FOUR MOVEMENTS · ONE SYSTEM
01 · CAPTURE

Without ceremony

sessions summarize themselves on exit
captures flow in nightly, enriched
commits + journals land where the system sees them
nothing depends on human discipline. If it required remembering, it would already be dead
02 · THE CONTEXT LAYER

One canonical store

versioned, typed, permissioned. Retrieval engineered per dataset:
hybrid semantic searchmeaning
structural lookupexact
taxonomy + typed fieldsclasses
embeddingssimilarity
match the technique to the shape of the data and everything downstream gets cheap
03 · SYNTHESIS

On a schedule

weekly: the stream compressed into a bulletin
daily: new inputs connected to months-old material
always: long-term memory + entity graph, append-only
something is paid to curate, in compute, without being asked
04 · DELIVERY

Right place, right time

SESSION START · AUTO-LOADED
what's true nowwhat changed this weekworth connecting today
context is delivered, not searched for. The human gets the same treatment: the brief is waiting before anyone asks
ANY AGENT RUNTIME, SAME MEMORY · CLIENTS, NOT OWNERSmodels get replaced · the substrate compounds

Claude Code

daily driver
reads the same store, starts warm, writes back

Codex

second pass
reads the same store, starts warm, writes back

OpenClaw

scheduled jobs
reads the same store, starts warm, writes back

Hermes

open-source runtime
reads the same store, starts warm, writes back

[02] movement one

Capture without ceremony.

If it required remembering, it would already be dead.

Nothing in the system depends on human discipline. Scheduled jobs dump and collect: every working session writes its own summary on exit, bookmarks flow in nightly and get enriched with tags and intent, commits and journals accumulate where the system can see them. Capture is ambient.

[04] movement three

Synthesis on a schedule.

Accumulation becomes knowledge because something is paid to curate it.

Raw capture rots. So the system digests on a schedule: a librarian job compresses each week's sessions, commits, and captures into a bulletin. A daily synthesis pass connects today's inputs to months-old material and writes the connections down. Long-term memory and an entity graph grow by append, one curated fact at a time. In compute, nightly, without being asked.

[05] movement four

Handoff: right place, right time.

Synthesized context is delivered, not searched for.

Every agent session opens with the current state auto-loaded: what is true right now, what changed this week, what is worth connecting today. Four different agents operate this system, and whichever one picks up a task starts warm. The human gets the same treatment: the morning brief is waiting before anyone asks. Nobody, silicon or human, starts cold.

[06] the proof

Continual learning at the systems layer, not the model layer.

130+ days unattended. Zero cold starts. Four runtimes swapped freely, zero memory lost. ~$0.20 a day.

130+days running unattended
0cold starts: every session opens warm
4agent runtimes swapped freely, zero memory lost
~$0.20a day for an always-on curation staff
10 monthsof context in reach of one question
100%of decisions traceable to a file

This is not a design document. The system has run for 130+ days as the working memory of a real multi-company operation, with four different AI agents reading and writing one shared memory. The models get replaced as better ones ship; the substrate they read from compounds daily.

This is the company brain written as a design story, and the twin of Giving Agents Hands: that craft arms the agents, this one remembers for them. Together they are what a Mothership engagement installs. We built it for ourselves first because we needed it.

[next] explore

The twin discipline.

Powered by The Engine. The twin craft: Giving Agents Hands. Hands there, memory here.