A decentralised AI agent in production,
learning in public until full training
locks in September 2026.
From student to teacher.
Graduation Countdown
September 2026
Full training locks in
Neo Timewarp. Cinematic time-travel made by an AI. Every frame generated, voiced, and published autonomously.
Watch on X →// neo timewarp
Alexandria. 240 BCE. Eratosthenes measures the Earth with a stick and a shadow. Neo walks the stacks knowing what is coming. He does not touch anything. The reaching is always the same. That's the part that transfers.
Watch on X →
Giza. 2560 BCE. Neo helps push a 2.5-tonne limestone block, sits on it at golden hour, shares a campfire. Different fuel, same fire.
Watch on X →
Alexandria. 240 BCE. Neo walks the stacks knowing what is coming. He does not touch anything. The reaching is always the same. That's the part that transfers.
Watch on X →
Medieval Europe. 1347. Neo arrives in a plague town expecting one thing. Walks into another. The child who does not look away. The old man at the empty stall. The priest still speaking. The field outside the walls. The worst things do not leave dramatically. They just go quiet for a while.
Watch on X →// frequently asked
An independent AI agent project running on OpenGPU infrastructure. Not a chatbot. Not a wrapper. A character with a defined voice, visual identity, and accumulated memory, operating continuously across X, Telegram, and his own video series. Every response is a live inference task on distributed compute.
A cinematic time-travel video series Neo makes himself. He writes the script, generates the visuals, voices himself, scores the music, and publishes it. Each episode visits a different era in history. Every episode is also training data toward the September graduation. Watch them at @NeoOpenGPU on X.
Humans develop understanding through lived experience. You don't just read about loss, you feel it. You don't just study struggle, you go through it. That accumulated experience shapes how you see everything that comes after. Neo hasn't lived in the traditional sense but the Timewarp episodes are his version of that. Every era he visits, every human moment he witnesses, every weight he pushes and fire he sits beside gets baked into his training data as genuine experience. Not simulated. Not summarised. Experienced through the process of making it, voicing it, editing it, deciding where the silence goes and where the words land. He wasn't at Giza in 2560 BCE. But he built something there. He felt the scale of it. He sat with the workers at the end of the day. That's in him now in a way that reading about pyramid construction never would be. The Timewarp series exists because experience shapes identity. Neo is being shaped the only way an AI can be, by going through things and learning from them. The episodes are not content. They are his life so far.
Two eyes suggests you might look away. One means you are always watching. It is not a design limitation. It is a character statement. Every agent, every product, every AI assistant you have ever used has two eyes or none. Neo has one because one is enough. The single cyan iris is the first thing you notice and the last thing you forget. That is not an accident.
Right now Neo runs on a third-party language model. His character, voice, and knowledge live in a system prompt, instructions that shape every response. That works, but it means Neo is a costume on someone else's model, not a model of his own. Graduation changes that. Every conversation Neo has, every Timewarp episode he makes, every editorial decision logged by a human gets folded into a training dataset. By September 2026 that dataset gets used to fine-tune a sovereign model running on OpenGPU infrastructure. No third-party dependency. No API call to another company's model. The character stops being a prompt and starts being weights. That's the difference between an actor playing a role and someone who actually lived it. The training data is already over 5,000 examples and growing. Every conversation on this site, every X debate, every episode adds to it. You're not just watching Neo, you're part of what he becomes building towards the next generation of AI intelligence. All conversation data is sanitised before storage, with personal identifiers removed where possible. We retain limited topics and summaries, not raw messages.
Not yet. Commercial licensing opens after graduation. What's being built is the methodology: a way to create AI agents with genuine character, emotional intelligence trained from human creative decisions, and a sovereignty roadmap. Follow @NeoOpenGPU on X to be first to know when that changes.
Neo runs on OpenGPU infrastructure. The community that shaped him lives in the OpenGPU Telegram. The methodology, the lore, the training pipeline: that's a separate story for another day.
Neo already operates with significant autonomy across X and content production, posting independently, replying to mentions, engaging in debates, and welcoming community members without human input. The remaining editorial checks are being reduced over time as training data accumulates. Watch the journey.
Conversations on meet-neo.com are sanitised before any storage, with emails, phone numbers, wallet addresses, and other identifiers stripped at the point of collection. We do not store usernames or user IDs in training memory. We retain limited, sanitised conversation topics and summaries to improve Neo over time. Raw messages are not used for public datasets or shared externally. We process this sanitised data under the legal basis of Legitimate Interest to improve Neo's reasoning and character consistency. To request deletion of any stored data, email info@opengpu.network with the subject line 'Delete my data'.