Introducing Veris

March 2026 · Noah Sioly · 8 min read

Every AI starts from zero. Every session. Every time.

You tell it your name. You explain your stack. You walk it through the architecture you've described three times this week. Tomorrow it'll ask you what framework you use.

The current solutions? A vector database that keyword-matches your old messages. A RAG wrapper that dumps 50 random chunks into the context window. A .md file you manually maintain.

None of them understand anything. They search text.

Your AI is a genius with amnesia.

What if your AI had a real brain?

Not a database. Not a search index. A brain — one that filters what matters, strengthens what's important, forgets what isn't, and gets smarter every day. One that runs in the background while you sleep, consolidating memories, finding connections you didn't see.

I didn't invent this. The human brain already does it. 158 mechanisms — documented by neuroscience, studied for decades. Spreading activation, Hebbian learning, thalamic filtering, sleep consolidation, metacognition.

I just translated them into code.

Not another vector database

Vector Search / RAG

StorageChunks of text in a database

RetrievalKeyword or embedding similarity

LearningNone

FilteringEverything gets stored

Over timeGets slower and noisier

BackgroundDead between sessions

Self-awareNo

Veris

StorageKnowledge graph — entities, relationships, weighted edges

RetrievalSpreading activation — one memory triggers connected ones

LearningHebbian — "fire together, wire together"

FilteringThalamus — only important signals pass through

Over timeGets smarter — consolidation prunes noise, strengthens patterns

BackgroundAlive — sleep cycles, replay, creative connections

Self-awareMetacognition — knows what it knows and what it doesn't

How it works

architecture

You talk to any AI
       ↓
  [ Veris ]
       ↓
  +---------+---------+---------+
  |         |         |         |
Thalamus  Senses   Emotion   Memory
(filter)  (analyze) (feel)    (store)
  |         |         |         |
  +---------+---------+---------+
       ↓
  Knowledge Graph
  (entities + relationships + weighted edges)
       ↓
  Context Generator
  (narrative briefing for your AI)
       ↓
  Your AI now knows you

1. Thalamus filters incoming information. Noise gets blocked, important signals pass through. Like the real thalamus — the brain's gatekeeper.

2. 6 Senses analyze every message. Code patterns, emotional tone, system health, knowledge consistency, context. Not one input channel — six.

3. Extraction pulls entities, relationships, and facts from the conversation. Not keywords — structured knowledge.

4. Knowledge Graph stores everything as interconnected nodes with weighted edges. 15 SQLite tables. Fully local.

5. Spreading Activation finds relevant memories. One concept triggers connected knowledge — like how thinking about "coffee" activates "morning", "productivity", "that bug you fixed at 3am".

6. Context Generator builds a natural-language briefing. Your AI never sees the graph directly. It gets a narrative — exactly the right context for the current conversation.

7. Consolidation runs in the background. Every 6 hours, the brain sleeps. NREM strengthens important memories. REM finds creative connections between unrelated knowledge.

The brain is alive

This isn't a database that waits for queries. Veris runs continuously.

Sleep cycles.

Every 6 hours, the brain consolidates. NREM strengthens important memories. REM finds creative connections between unrelated knowledge. Your brain works while you sleep.

Hunger.

The brain detects knowledge gaps and drives your AI to ask the right questions. Not because you told it to — because it wants to know. Like unconscious curiosity: the question just comes out.

Growth.

A new brain starts like a child: curious about everything, low thresholds, absorbing it all. After hundreds of sessions, it matures — selective, fast, deep schemas. After a thousand, it's wise.

Self-awareness.

The brain monitors itself. Too much noise? Raises thresholds. Knowledge fragmented? Triggers consolidation. It knows what it knows — and what it doesn't.

One brain. Every AI.

Switch from Cursor to Claude Code mid-project. Veris doesn't care. It knows you. It knows your project. It picks up where you left off.

The AI you use is just the mouth. It vibrates, it produces words. The brain behind it? That stays the same. Same personality, same memories, same knowledge — no matter which AI you talk through.

OpenClaw Codex Claude Code Gemini Cursor Windsurf Aider Continue Any MCP Client

The numbers

mechanisms

lines of TypeScript

providers

local

By one person. No team, no funding, no framework — just TypeScript and a mass of neuroscience papers. This is version 0.1. And this is just the beginning.

Give your AI a brain.

$ npm install -g veris-dev && veris init

GitHub Docs Website

Your brain runs 100% locally. No cloud. No telemetry. Your data is yours.