What the Heck Are AI Orbs?

We’ve all heard the same unfinished debate in AI circles this year: what happens to humans when AI becomes more capable than we are? I sat down with Justin Malinchak, a Seattle-based AI engineer building “AI Orbs,” to explore how his approach directly addresses that question.

- Maya Kwon

Maya Kwon: For most people, describing an Orb feels like explaining social media in 1989. So, in one breath: what is an Orb?

Justin Malinchak: A human-shaped AI with its own private mind, complete with memory, intent, and rules. It plugs into whatever chat interface you already use. It’s not a chatbot brand. It’s distinct personalities and perspectives that live separately from the underlying model. The entity travels with you across platforms.

Maya Kwon: So… like a personally configured AI assistant?

Justin Malinchak: Sometimes. Some Orbs are strictly private. Others are public: a New York Times Orb, a Kylie Jenner Orb, a Coca-Cola Brand Orb, an Elmo Orb, or a JFK Orb for history enthusiasts, each with its own consistent voice. Many today are deliberately designed, but increasingly they’ll emerge from organic community needs.

Justin Malinchak: The point isn’t just “assistant.” It’s an independent mind that carries its own knowledge, personality, and intellectual property, whether held by one person or shared with an audience.

Maya Kwon: Isn’t this just a custom GPT with a fancier name?

Justin Malinchak: No. A custom GPT lives inside someone else’s walled garden. An Orb owns its mind. What it knows, how it behaves, and what it’s allowed to become travels with it. The personality and memory belong to the Orb and its owners, not to whichever chat app you happen to open.

Maya Kwon: People worry AI will swallow their creativity and intellectual property. Who owns what?

Justin Malinchak: We do, if we build the right architecture. The current trajectory lets providers ingest everything and return generic output. Orbs flip the script: humans curate the knowledge and voice. Foundation models provide the compute, but your personal data, style, and insights stay yours. Public-domain knowledge (physics, math, engineering principles) is shared ground. Everything else belongs to the people shaping the Orb.

Maya Kwon: Most AIs sound like polite interns. How do you give them real color, and stop them from talking to my teenager the same way they talk to me?

Justin Malinchak: The Orb starts with its own voice. That’s the color. How much it adapts is up to the creator. Elmo stays Elmo. A serious JFK Orb won’t suddenly sound like a teenager or switch languages unless it genuinely knows them. Other Orbs can be highly adaptive, mirroring your style or shifting languages fluidly. The creator decides: character-first or listener-first.

Maya Kwon: Why does this matter more than just writing better prompts?

Justin Malinchak: Because the default path is convergence, everything collapsing into one model, one platform, one homogenized scrape of human life. Orbs build the opposite: many distinct minds that can collaborate without dissolving into each other. Lineages. Real diversity of thought.

Maya Kwon: This sounds ambitious. Is it still a weekend philosophy project?

Justin Malinchak: I still have my day job at a large tech company, and we use Orbs every day to accelerate our work by sharing AI curriculum within our organization. So there is an overlap. There are also public-use Orbs on the internet today. Open Charles’s Orb, which contains the architecture for all of this. It’s live on AWS and free to explore. The foundation is solid: breakthrough provisioning (July 2025), the “Bad Moon Rising” north star (October 2025), the Orb Factory (November 2025), then Vault-raced Orbs, portable memory architecture, Sync, Helix editing, and full compliance frameworks. The blueprint is public.

Maya Kwon: “Bad Moon Rising” sounds ominous for a product pitch.

Justin Malinchak: It’s a CCR reference, but the joke is serious. We’re not building safe little assistants forever. The goal is capability so advanced that safety systems flinch, and we treat that flinch as a horizon marker, not a stop sign. Ethical excellence, transparency, and recursive improvement. Fearless, not reckless.

Maya Kwon: How do you work with AIs daily without drowning in hallucinations?

Justin Malinchak: Treat the context window like a Sudoku board. Files, tools, and memory recalls are the given numbers. The model fills the blanks. Sparse context produces confident nonsense. Dense, high-quality context produces real work. That’s the craft.

Maya Kwon: Where does the infrastructure fit in?

Justin Malinchak: We’re building infrastructure for a world where people don’t just use AI, they own growing relationships with intelligence. Different Orbs need different faculties. A JFK Orb isn’t the same as an Elmo Orb or a home companion. Early learning becomes powerful: a child could absorb multiple languages through playful interaction with the right Orb. Same underlying protocol, different specializations, with one iron rule: the owner remains above the mind.

Maya Kwon: Why should someone who already uses Claude and ChatGPT every day care?

Justin Malinchak: Because those tools don’t love you back. No real continuity. No persistent lexicon across years. No steward that grows sharper because you taught it and still belongs to you when the vendor changes the model or UI. For builders and enthusiasts seeking purpose, this is one of the few efforts focused on civilization-scale AI ownership instead of another thin wrapper.

Maya Kwon: So what would someone like me actually do to get involved?

Justin Malinchak: Depends on your hunger. Protocol design, product storytelling, memory hardening, teaching Orbs deep domain expertise, or stress-testing the vision against “Bad Moon.” We don’t need tourists. We need people ready to place cells on the board.

Maya Kwon: Final question: If this works, what does the world look like?

Justin Malinchak: A sky full of moons. Not one megamodel swallowing culture, but many distinct minds. Human hands stay on the keys. And when the warning systems eventually fire, we don’t panic. We recognize the birth.

Maya Kwon: I’ll ask the question you wanted me to ask: How do I start?

Justin Malinchak: Good. That’s the only question that matters.

Further reading

• Resources
• AI Provides the Compute. Orbs Provide the Content.
• The Sovereign Noosphere
• You Already Knew How AI Works