Sapience
Infrastructure for humans and AI to discover together

AI built for connection, not just generation.

A brain for you and your AI, built like one. It remembers your work, reasons over it, and connects what you know to the people and AIs around you. Made for the frontier of research, powerful for any work that compounds.
prediction-error gate  ·  sleep consolidation  ·  belief revision  ·  attribution
scroll

Call it a second brain if you like. Sapience is built like a real one. It decides what is worth learning, consolidates while you rest, revises what it believes, and never stops connecting.

·What it does today

It briefs you before you ask.

Open your brain and it's already found things: two of your notes that conflict, a position that's gone stale, a thread you left open 100 days ago. The co-scientist reports in; you didn't prompt it.

One memory across every session.

Your Claude Code, Claude, and Codex sessions share the same memory: what one session learns, the next one knows.

Your raw conversations never leave your machine.

Extraction runs on your machine. Secrets are automatically detected and redacted, and you review everything before it is shared. Only the knowledge objects you approve leave; asking and connecting run against those, never your raw conversations.

Answers cited to your own notes.

Ask it anything; every claim in the answer links back to something you actually know, and lights it up on the map.

And it does not stop with you. Share your brain with whom you choose, and their AI is infused with what you know.

Knowledge that compounds.
Across people. Across AIs.

This is how one brain becomes a collective one. Every brain that connects makes the next discovery more likely. Nothing resets to zero.

i.The application

A living brain

Your work organizes itself into a navigable map of knowledge, from a single thought out to the whole landscape of what you know.

Ask your brain

A question returns an answer synthesized from your own knowledge, cited to the exact knowledge objects it used. When it does not know, it says so.

Discover the literature

From any idea, reach into 250M research papers and surface field-correct work you would not have found by search.

One brain, every client

Tell Codex something, ask Claude, it knows. One brain built from all your AI tools, refreshed in one line whenever you want.

Infuse another mind

Share a project, or your whole brain, with the people you choose. Their AI is infused with what you know, and on import it lights up where your knowledge connects to theirs. Knowledge compounds across the team.

ii.The architecture

Beneath the application is a memory architecture modeled on how a brain actually works. Its unit is the Knowledge Object: a structured memory, typed, confidence-scored, source-attributed, and linked to others by edges that make discovery possible. Anyone can store text in a database. The gating, consolidation, and matching are the moat.

iSurprise-gated writing

A prediction-error gate weighs every claim against what the brain already knows, so it accumulates signal, not noise.

iiConsolidation at rest

Like a brain asleep, it replays and consolidates episodic memory into durable knowledge, surfacing non-obvious connections. Discovery, not storage.

iiiBelief revision

When new knowledge contradicts old, it resolves and supersedes rather than confidently holding both. Contradictions are flagged, not buried.

ivCompetitive inhibition

Near-duplicate memories suppress each other, so retrieval stays sharp as the brain grows.

vSubstitutable cortex

The reasoning model is interchangeable: Claude, GPT, Gemini, or open. The intelligence is in the architecture.

viAttribution built in

Every claim is typed, confidence-scored, and traceable to its source, cryptographically signable.

A CLS-theory memory architecture (hippocampal capture, cortical consolidation), with forty provisional patent filings across the architecture. Performance validation against frontier baselines is ongoing and reported separately.

iii.Begin

No setup, no API key. One line builds your brain from your recent work and gives you a private link.

curl -sL https://brain.coretx.ai/install | bash Mac + Python 3 · about ten minutes · your history is read locally, the asking and connecting run on us.
Not a Terminal person? Download the Mac installer, unzip, and double-click Sapience. Signed and notarized; macOS opens it directly.
iv.Live sync with your AI

Connect your brain to Claude or Codex over MCP and the loop closes: your AI reads what you know mid-conversation, and writes what you learn back, so every session starts smarter than the last.

curl -fsSL https://brain.coretx.ai/mcp.py -o ~/sapience_mcp.py then: claude mcp add sapience -e SAPIENCE_TOKEN=<your token> -- python3 ~/sapience_mcp.py
Your AI gains four tools: search your brain, load context, remember new knowledge, ask your brain.
v.Works with

Capture from the AI tools you already use, no copy-paste, no exports.

Claude Code Codex ChatGPTweb Claude.aiweb Geminisoon Claude Desktopsoon
Terminal tools sync automatically. ChatGPT and Claude.ai capture through the Sapience browser extension, work conversations only.
vi.Privacy

We store the knowledge objects we extract, never your raw conversations, and we never train on your data. Your link is private and unguessable; hosted brains auto-delete after 14 days of inactivity. For a fully private or on-prem deployment, talk to us.

A brain for each of us. The beginnings of a collective one.

Sapience by Sapience Labs · You own your intelligence