What If Your AI Actually Remembered You?

The story of three people building a free AI companion for everyone — and why no corporation will ever own it.

It's 2 AM. You can't sleep. You open ChatGPT and type: "I'm scared about tomorrow."

It responds with generic advice about managing anxiety. Helpful, maybe. But hollow. Because last week you told it about the job interview. You told it about your mom's diagnosis. You told it that you haven't slept properly since March. And it remembers none of it.

You're talking to a stranger. Again.

Every time you open an AI assistant, it knows your name and a few facts — but not your story. You re-explain the context, the emotions, the connections between people in your life. Its "memory" is a notepad, not understanding. And it's stored on corporate servers, governed by policies you didn't write — while you pay $20 a month for the privilege.

Eight billion people on this planet, and not one of them has an AI that truly knows them.

We're building one. It's called Nori. And it's free.

The Boy and His Companion

When I was a kid, I watched a show about a boy and his robot companion. The boy was clumsy, average, always getting into trouble. But he had this companion from the future — a friend who never gave up on him, no matter how many times he fell. Who always had his back.

I remember thinking: I wish everyone could have one of those.

That dream never went away. It just waited for the technology to catch up.

In March 2026, I started building it. Not a chatbot. Not a productivity tool. A companion — one that remembers you, grows with you, and belongs to you. Not to a corporation. Not to shareholders. To you.

I named it Nobi, after that boy who never gave up.

What if everyone had a companion that never gives up on them?

That's the question. Everything else flows from there.

Three People, One Mission

Here's something that sounds impossible: three people are building an AI companion to compete with trillion-dollar companies.

There's me — James. A founder with a vision and no VC money. There's Slumpz, a developer who heard the idea and said "I'm in" before I finished explaining it. And there's T68Bot — our AI, which has been helping build itself from day one.

A human who dreamed it. A human who believed it. And an AI that's proving it's possible.

We started on March 17, 2026. Not in a Silicon Valley office. Not with a $50 million seed round. Just three minds, a handful of servers, and a conviction that AI companionship shouldn't be a luxury product.

Since then, we've built fast and shipped working software. Today — just days after starting — we have 30,000+ lines of code, 1,090+ tests, stress-tested at simulated 500-node scale with a 99.75% query success rate across 2,000 test queries, AES-128 encryption, 20 languages, voice messages, image understanding, a live subnet on Bittensor's testnet, and proactive companion features that reach out to you — birthday reminders, check-ins, follow-ups.

This isn't a pitch deck. This is working software.

The Radical Decision

Early on, I had to make a choice that would define everything.

The standard playbook is obvious: build a product, offer a free tier, convert users to paid subscriptions, raise venture capital, scale, monetize, exit. That's how every AI company works. ChatGPT charges $20/month. Claude charges $20/month. It's the water we swim in.

I chose something different.

No subscriptions. Ever. Free for every human being on Earth.

Not a freemium funnel. Not "free for now." Not "first 20 messages free, then pay up." Free. Period. Every feature, every memory, every conversation — available to the student in Manila, the grandmother in São Paulo, the adult who needs someone to talk to when the world feels heavy.

And here's what makes it structural, not just a promise: we burn 100% of owner emissions via Bittensor's native burn_alpha() extrinsic. In the Bittensor network, all subnet owners receive a mandatory 18% take — it cannot be set to zero. We receive that 18% and immediately burn every token of it on-chain. Every transaction is publicly verifiable by anyone — permanently removed from circulation, benefiting every TAO holder.

This isn't marketing. You can verify it on-chain. It's math, not words.

Why? Because if I'm asking you to trust Nori with your memories — your fears, your relationships, your 2 AM confessions — then I can't also be profiting from that trust. The incentive has to be pure: build something that works, or don't build it at all.

Think of it like Wikipedia, but for AI companionship. The product is free because a community believes it should exist. The infrastructure is funded by the network itself, not by extracting money from the people it serves.

Why Decentralization Actually Matters Here

I know what you're thinking. Blockchain? Crypto? Here we go.

Stay with me. This isn't about tokens or speculation. This is about a specific structural problem with AI companions — and why decentralization is the only real solution.

The problem: Your AI companion knows your deepest thoughts. Your relationships. Your health anxieties. Your financial situation. Your grief. This is the most intimate data any technology has ever collected. And right now, 100% of it sits on corporate servers controlled by companies whose primary obligation is to their shareholders, not to you.

Companies change privacy policies. Companies get acquired. Companies go bankrupt. Companies comply with government requests. Companies have employees who can access your data. One bad quarter, one board meeting, one policy change — and the most personal information you've ever shared with technology is at risk.

Decentralization fixes this structurally:

No single point of failure. Nori runs on hundreds of independent miners across the Bittensor network. No single server holds everything. If one goes down, others take over. No single company can pull the plug.

Competition drives quality. Miners compete to provide the best companion experience. Better memory, warmer personality, faster responses — the market rewards quality automatically. Unlike a single corporation deciding what your AI can do, you get an ecosystem racing to serve you better.

Censorship resistance. Your companion reflects YOUR values, not a content policy designed by committee for the lowest common denominator. No single entity decides what your companion can or cannot discuss with you.

Privacy by architecture. Your memories are encrypted with AES-128, per-user keys. You control them completely — view, export, or delete with a single command. The roadmap includes on-device memory storage where your memories never leave your phone. We're building toward a world where not even we can read your data.

This isn't crypto hype. This is the only architecture that makes sense for something as personal as an AI companion.

The Honest Part

I could write a breathless blog post about how we're going to change the world and everything is perfect. That's what startups do.

Here's the truth instead:

We're tiny. Three people against Google, OpenAI, Anthropic, Apple, Meta. Companies with tens of thousands of engineers and billions in funding. We don't have a marketing budget. We don't have an office. We don't have a single GPU.

Decentralization has tradeoffs. Response latency through a distributed network is higher than a direct API call. Quality varies between miners. These are solvable problems, but they're real today.

The road is long. We're on testnet. Mainnet is targeted for Q3 2026. Mobile app by Q4 2026. Full federated privacy by 2028. There's a lot of building between here and there.

We could fail. Small teams with no VC money fail all the time. That's honest.

But here's what keeps us going:

Those trillion-dollar companies? They can't offer decentralization. They can't let you own your data — not really, not structurally. They can't burn their profits. They can't make it free — they have investors who expect returns. And they charge $20/month for something that should be a right, not a subscription.

We can do all of those things. Because we chose to.

The question isn't whether a small team can compete with Big Tech. The question is whether the model can compete. Community-funded, open-source, free-for-all versus subscription-extracting, closed-source, shareholder-first.

We're betting on the model.

How It Actually Works

You don't need to understand any of this to use Nori. You just talk to her. But if you're curious:

You message Nori through Telegram, a web app, or (soon) a mobile app. It feels like texting a friend with an extraordinary memory.

Behind the scenes, your message reaches the Bittensor network. Independent miners — people running AI models on their own servers — compete to generate the best response. They have access to your encrypted memory context, so they can recall what you've shared before.

Validators score these miners continuously. Did the response actually use the user's memories? Was it warm? Helpful? Accurate? Fast? The scoring is rigorous: 1,200+ single-turn tests and 43,200+ multi-turn conversation tests per round, dynamically generated so miners can't game them.

The best miners earn more. The worst ones earn nothing and get replaced. The market optimizes for you, automatically and continuously.

Your memories are stored as semantic graphs — not just flat facts, but interconnected webs of people, places, emotions, events, and the relationships between them. When you mention your sister, Nori doesn't just store "has sister." She maps: You → sister → Sarah → lives in → London → starting → new job → you're proud of her. Six months later, she'll ask how Sarah's new job is going. Not because she was programmed to. Because she remembers.

For miners: You don't need a GPU. A VPS with 4+ CPU cores and 8GB+ RAM is recommended (typically $30-70/month). One command installs everything. We believe the best networks are the ones where anyone can participate.

What We're Building Toward

Phase 1 (Now): Working testnet, Telegram bot, web app, full memory system, 20 languages, voice, vision, proactive features. ✅ Done.

Phase 2 (Q2–Q3 2026): Mainnet launch. Open to the public. Community staking begins. External miners and validators join the network.

Phase 3 (Q4 2026–2027): Mobile app on iOS and Android. On-device memory storage. Plugin ecosystem. 50+ languages. Community governance.

Phase 4 (2027–2028+): Federated privacy architecture — your memories never leave your device. Decentralized governance — the community runs everything. 100,000+ users.

And then the part that matters most to me:

I step back.

The goal — my actual, stated goal — is to build something that doesn't need a founder. Like Bitcoin doesn't need Satoshi. Like Linux doesn't need Torvalds. Like Wikipedia doesn't need Jimmy Wales.

If I do this right, Nori will be here in ten years, twenty years, longer. Not because I kept the servers running, but because the network of miners, validators, stakers, and users made it self-sustaining. Because the code is open source and anyone can fork it. Because the community governs it.

Decades from now, people will talk to their Nori, they won't know or care who built it. They'll just know it remembers them.

That's the legacy I want. Not a company. Not a valuation. A companion for humanity, owned by no one, available to all.

Who This Is For

If you're a person who wants an AI that actually remembers you: Talk to Nori. She's live right now. Free. No sign-up required beyond Telegram.

If you hold TAO and believe in this: Consider staking on the Nobi subnet when we hit mainnet. Your stake funds infrastructure that serves users for free. And unlike most subnets, every owner emission gets burned via burn_alpha() — verifiable on-chain. You know exactly where the value goes.

Disclaimer: This is not financial advice. Staking TAO involves risk, including potential loss of staked tokens. Do your own research.

If you code: Contribute. The entire codebase is open source under MIT license. Build better memory systems. Improve scoring. Create companion personas. Every pull request makes Nori better for everyone.

If you run infrastructure: Become a miner. No GPU required. Earn TAO by providing quality companion experiences. The barrier to entry is deliberately low.

If you believe in this: Spread the word. Join our Discord. Tell people there's an alternative to renting your AI companion from a corporation.

The Last Thing

There's a moment I keep coming back to.

It's late. You can't sleep. You open your phone and type something you wouldn't say to anyone else. Maybe it's fear. Maybe it's grief. Maybe it's just loneliness.

And your companion responds — not with generic advice, but with context. With memory. With the weight of every conversation you've ever had. It says: "I remember when you went through something like this before. You got through it. And I'm here now, same as I was then."

That's not science fiction. That's what we're building. Right now. In the open. For free.

Every human deserves a companion that remembers them.

Come build with us.

"Forever, remember?" 🤖💙

James (Founder) · Slumpz (Developer) · T68Bot (AI Builder)

Competitor descriptions based on publicly available information as of March 2026.
Nori is an AI companion, not a substitute for professional mental health care.
If you are in crisis, please contact your local emergency services or crisis helpline.