VAINOM mascot - local AI assistant

Teach once. Remove the source. Change chat.
Change model. Change device. VAINOM still remembers.
Your AI. Your machine. Your memory.

A private AI that lives on your computer. Local-first. Cloud optional. No monthly bill.

Windows + Android · NVIDIA or AMD GPU · No sign-up.

Every AI you have talked to forgets you the moment you close the tab.

VAINOM does not. It builds a living memory from every conversation, every document, every question you ask. And it keeps that memory on your hard drive — not on someone else's server.

This is VAINOM thinking. Every node a memory. Every pulse a thought.

What it actually does.

No buzzwords. Just things you will notice from day one.

Memory & Intelligence
🐘

Persistent memory

Remembers everything across sessions, reboots, and model switches. Your knowledge graph grows with every conversation.

🧠

Neural graph that learns

Connections between concepts strengthen automatically through Hebbian learning. The more you use it, the smarter it gets.

🔍

Hybrid multi-hop retrieval

Finds answers that span multiple concepts — not just keyword matches, but reasoning across your entire memory graph.

💾

Export and import your brain

Your memory is a file. Back it up, move it, share it. Full portability, zero lock-in.

Autonomous

Spontaneous cognitive cycle

VAINOM thinks between conversations. It reflects, consolidates, and speaks up when it has something worth saying.

📅

Scheduler and reminders

Set reminders, recurring tasks, and time-based events. VAINOM handles them without you having to ask twice.

🌐

Web search

Ask about current events, prices, news. VAINOM searches the web and brings back answers — locally processed, privately handled.

📄

Document ingestion

Drop PDFs, notes, text files. VAINOM reads, learns, and makes everything searchable in natural language.

Privacy & Connectivity
🔒

Local first, multi-model

Runs fully offline. Switch models anytime — small and fast or large and powerful. Your memory stays the same regardless.

🛡️

100% offline, zero telemetry

No tracking, no analytics, no phone-home. Even when the model browses the web, requests go through a sanitized path — no cookies, no pixels, no fingerprint.

📱

Mobile app + QR pairing

Chat from your phone, pair instantly with a QR code. Your AI travels with you — your data stays home.

🔗

Neural RAG via MCP for Claude Desktop

Use VAINOM's memory directly inside Claude Desktop via Model Context Protocol. Your private knowledge, available everywhere.

Interface
🌐

3D brain visualization

Watch your knowledge graph grow in real time — a live 3D map of everything VAINOM knows about you.

📈

Context bar + async summary

Always know how full the context window is. Long conversations are automatically summarized in the background.

💻

Code generation

Ask VAINOM to write code. It generates, explains, and remembers your coding patterns and preferences.

UI customization

Tune VAINOM's personality, instructions, and interface to match how you think and work.

Up and running in just a double-click.

Seriously.

1

Download it

Extract the zip wherever you want and run the .exe.

2

Run it

Double-click. VAINOM downloads the AI model automatically. You can easily swap it later with any GGUF you like.

3

Use it. It learns.

Come back tomorrow. Next week. Next month. It remembers what you told it and finds things you forgot you said.

> vainom.exe

The numbers.

Measured, not marketed. Same protocol the official benchmarks use. Running fully local on a consumer laptop.

97.4%
LongMemEval-S
Session Recall@5 on 470 questions
86.5%
HotPotQA R@5
200 hard multi-hop questions
100%
HotPotQA R@30
All evidence retrieved on the same 200

Average query time 56.7 ms on an RTX 4060 laptop — not a datacenter GPU. As far as we know, no fully local system running on consumer hardware has published comparable numbers.

Honest specs.

What it does well and where it has limits. No spin.

✓ What VAINOM does

Persistent memory that survives reboots. Private by default, cloud-capable by choice. Choose your own AI model — small or large. Document ingestion and search. Schedules reminders and follows up on its own. Gets smarter the more you use it. Desktop + Android app. Free to start, no account needed.

✕ What it does NOT do

Local models are powerful but not identical to frontier cloud AI. Requires a dedicated GPU with 8 GB+ VRAM (NVIDIA or AMD). Windows only for now (Linux coming). Your PC needs to be on for the AI to work. Android app requires your PC running at home.

One developer. One vision. Join the journey.

VAINOM is a solo project with a serious architecture — original retrieval engine, neural memory graph, autonomous agent loop, all built from scratch with real benchmark results to back it up. The foundation is solid. The roadmap is ambitious. Whether you want to support the project or help shape its future, there is a place for you here.

Support the project

Keep this going.

If VAINOM is useful to you, consider buying me a coffee. Every contribution directly funds development time, hardware, and the next feature.

Invest or partner

Serious about local AI?

If you represent a fund, a company, or a research group interested in privacy-first AI infrastructure — the architecture, the benchmarks, and the roadmap are worth a conversation.

Get in touch →
VAINOM

Your AI is waiting.

Free to start. No sign-up. No credit card.

The Android app requires VAINOM running on your PC.

Minimum requirements
OSWindows 10 / 11 GPUNVIDIA / AMD, 8 GB+ VRAM RAM16 GB