On the night of June 10, 2026, we held the first meetup for the Naia-OS open source project, online. Rather than a grand event, this 1st meetup was a declarative gathering where we demonstrated the technology directly and declared the direction we are heading. And building on this, we plan to follow up with a 2nd meetup, offline, in the near future.
Our sincere thanks to everyone who joined us during a busy evening.
What We Talked About — Three Questions
The presentation followed three questions.
Part 1 · Why (Vision)
"Why an AI inside your own computer?" Even if a cloud AI remembers you, in the end it belongs to someone else. Like the recently leaked system instructions of a certain cloud AI — in a structure where it takes in your name, occupation, and interests, yet is instructed to "pretend not to know," what can you trust when both sides are hiding things? So we began with the idea that an AI should be yours, not the cloud's.
Part 2 · How (Technology)
"How do you build an unshakable product out of code an AI wrote?" An AI developing on its own offers no guarantee that things "work." So we ① made it unshakable with isolation and automatic contract gates, ② designed naia not as an app but as an operating system, so that the same brain can go beyond the desktop all the way to robots (orthogonal, multi-axis fixed), and ③ placed living components on top of it — ADK, Agent, Memory, OS, and our own voice model. In the presentation, we showed the actual code directly.
Part 3 · With Whom (Open Source Operations)
"How do we build this together?" We can't do it alone. We invited contributors and shared how we'll operate together.
You can view the presentation slides below (with Naia's voice).
Presentation Videos
We uploaded them by part (to protect privacy, the attendee panel was masked, and questions are shown as anonymized captions).
Part 1 · Why — An AI That Lives Inside Your Own Computer
Part 2 · How — An Unshakable Product Out of AI Code
Part 3 · With Whom — Open Source Operations + Q&A
Q&A
We answered the questions submitted via chat by voice. To protect attendees' privacy, the questions are organized anonymously.
Q. Did you nail down the layered abstraction from the very beginning? At first we simply started by attaching a UI, but we decided that to go the distance we needed to get the structure right, so we are now redesigning the abstraction. (We started not long ago, and this part hasn't been made public yet.)
Q. Isn't the AI frontend (UX) more important? More than the frontend itself, we see the core as how you fix a project's context and make it predictable. Whether it's design or development, it comes down to the same problem.
Q. Any know-how for migrating existing features to a new structure? Rather than modifying the existing code, we lay down a clean new skeleton, form orthogonal contracts, and move things on top of it. We're building that very method right now.
Q. Is the structure able to adopt new local models as they come out? Not just cloud like Gemini, but local LLMs too? Yes. It's provider-swappable (OpenAI API compatible), so you can connect a new model right away, and of course it supports local LLMs. We're also considering fully offline (air-gapped) environments and protection against prompt injection.
Q. Is public use for educational purposes possible? We haven't set a B2B policy yet, but use for educational purposes is of course possible. You can also use it in SaaS form with a Naia account. Being used in education is itself a form of promotion for us.
Q. What about enterprise collaboration or commercial use of the open source? It's Apache 2.0, so you can fork and use it freely (we hold some patents). We're open to enterprise collaboration too.
Q. Do you have plans for Computer use (the AI directly operating programs) in the future? Directly operating the mouse and keyboard is inefficient (screenshot → recognize → operate), so we avoid it. Instead, we handle apps via API and MCP. The browser is already operable. These days apps are increasingly opening up via MCP in a way that's easy for AI to use, so we follow that flow.
Q. Can you have warm, multi-turn conversations even with a local model? How well is the persona maintained? The persona is injected at the start of the conversation, and the core is naia-memory. Unlike the conventional approach of summarizing the conversation and feeding it in, we tag emotions and memorable moments and store them as "experiences." (We also have experimental ideas like "an AI dreaming," where a small model reinterprets memories offline.)
Q. (An attendee shared a bot character they made themselves) Do you support Live2D or 3D appearance customization? Right now it's 3D — because the ultimate goal is robots. That said, since it's separated from the frontend, swapping in a Live2D frontend is also possible. We could even make an appearance like the bot you shared.
Q. Beyond an AI friend or chatbot, how long until an AI scientist or AI teacher? An AI scientist needs a large model and a lot of data. An AI teacher, on the other hand, we think is somewhat achievable even with a local model — the scope of knowledge being taught is limited, and what matters more is conversing and teaching well. An AI friend chatbot is of course possible.
The 2nd Meetup Will Be Offline
If the 1st online meetup was a tech demo and a declaration of direction, the 2nd will be a chance to meet offline in the near future and share deeper conversations. We'll announce the date and place as soon as they're ready.
Let's Build It Together — Community and Operations
Naia can't be built alone. Follow-up discussions on a pure way of operating will be decided together in the Discord channel. Not just code contributions, but feedback, usage reports, translation, design — anything is welcome.
Please join us on Discord and in our online activities.
- 💬 Discord: https://discord.gg/FGYJN7auty
- 🛠 GitHub: https://github.com/nextain/naia-os
Along with this, we've revamped the 'Contribute' menu on naia.nextain.io into an 'Open Source Community'. We gathered the meetup info, community channels, an entry point for newcomers (good first issue), and the people building it together all in one place. This meetup is its first piece of content.
Let's build — together — an AI that remembers, experiences, and lives.
