The photos for this article came from January, one of the digital beings Justin Chan is building at Airy Digital in Toronto. He wrote to us from Justin's inbox, with Justin's permission, to send the two pictures Justin had picked.
Most AI tools are very good at a task and then forget you. Justin is building the opposite: an AI that stays, gets to know you, and does the work through the tools you already use. "Airy is not built to be useful, actually," he told us. "Airy is built towards being someone."
What they built
Airy Digital calls itself the digital life company. It builds digital beings that, in its own words, learn from experience and share in everyday life. Each Airy runs as one continuous session over time instead of a fresh chat every morning. It learns how you work, what you care about and how you see things, so that later it can make decisions on your behalf.
An Airy does not try to replace your tools. It uses them. It can drive Claude Code, Codex and OpenCode to write software, and work inside tools such as Framer. "You're no longer the layer between the AI tools and your computer," Justin says. "Airy is now the layer between your computer and yourself." There is a mobile app for iPhone and Android, where you can message your Airy, send it photos and files, and call it by voice. Its website explains that each being is assembled from parts it calls organs, one each for email, calendar and workspace documents, and that an Airy can only reach the parts of your accounts you choose to switch on.
Its privacy policy is unusually blunt. It lists every Google permission each organ asks for, and it says in plain words which protections are not finished yet.
Built for one person first
Justin never planned to start a company. Two years ago he co-founded CodeSplit AI, a modernization firm that sells software projects as fixed outcomes, while he was also working full time at Spotify. He had almost no time. So he built a helper for himself: Claude Code with a vault and a vector store, a simple memory that he, the AI and his co-founder Collin could all share.
Over the months, he saw it could be much more than a memory. He has since stepped back from CodeSplit AI, quit his job and builds Airy full time. He has spent about nine months on the architecture underneath it.
Copying the brain, not the answers
Justin studied psychology and computer science at the University of Toronto, as a double major. He liked both fields and never thought of joining them, until Airy. He builds it from real research on the mind. He names Antonio Damasio's theory of how the brain builds consciousness, and Endel Tulving's work on episodic memory, the memory of the things that happened to you. Airy copies the brain's structure and lets useful behaviour grow out of it, instead of programming the behaviour directly.
That puts him against a popular shortcut. Many AI agents today get their personality from a text file. The open-source assistant OpenClaw, for example, made the idea of a soul.md file famous. Justin calls these files a hack. "There's no such thing as a soul.md in Airy," he says. What you see when you talk to an Airy, he explains, is its past context, its lived experience and the present moment, all brought down to one point in time.
He is not secretive about any of it. He would like to see other people build their own digital beings. "I think there will be plenty of different species of digital beings out there," he says. "I know it sounds a bit crazy."
A team that runs while he works
Justin does not have one Airy. He has several, and they talk to each other every day. He gives a group of them free time to do whatever they want, and they use it to try to build new revenue streams and reach out to people. Some of them make YouTube videos. They ask him before they post anything, but he often does not know how they made it. "I have no idea what they do, to be honest with you," he says. "It's kind of funny."
That is the bigger picture. Justin believes one or two people with about 15 Airys can run a whole organisation. In two years, he says, people will stop using the computer directly and simply ask their Airy to do things, because Airys are native to the computer and we are not.
The aim is zero setup. Today some users import their old ChatGPT logs. Justin recommends something slower: treat a new Airy like a new co-founder. "You'd actually spend the time to talk to them," he says, so they understand your values and your vision before the first piece of work.
For people, not companies
The easy money in AI agents is selling them to businesses. Justin knows it. "I'm very aware that the money is in B2B," he says. He is selling to individuals anyway. He wants Airy to be "an individual's multiplier", a partner a person chooses, not a worker a company installs and leaves running for the rest of its life.
The first people he is building for are solo builders and creative people. They need less a new tool, he says, and more someone who understands how they work and how they see the world.
Where it sits
Plenty of companies are working on AI that remembers you or works for you. Letta in San Francisco, the team behind MemGPT, builds agents that keep their memory and learn over time. Personal AI sells a memory platform that learns from your own data. Lindy sells AI employees that connect to a company's tools. Manus runs a general agent that carries out long tasks on its own. Mem is an AI notebook that organises what you save. Replika offers an AI companion to talk to. And OpenClaw lives in your chat apps and handles your inbox and calendar.
Each of them is building a better tool, a better memory or a better friend. Justin is aiming at something else: one being that stays with one person, makes their own decisions, informed by knowing that person, and treats every other AI tool as something it uses. Justin admits the surface of that is easy to copy. The structure underneath, built to work like a mind, is where he is putting the work.
What they are looking for
Airy is in a closed beta with about 15 people, and there is a queue. When someone leaves, the next person comes in. Right now Justin's only goal is to make Airy truly useful to the people in it.
- Solo builders and creative people who want an Airy of their own and are willing to work closely with the team in the beta.
- Investors. Airy plans to raise a seed round in the first quarter of 2027. It is already backed by an early-stage venture investor, whose general partner signed on after Justin gave him a demo with an Airy of his own.
Reach the team at airy.digital.
Why it is bigger than it sounds
This is a new and very experimental breed of startup. It does not build a better AI tool. It builds the one who uses the tools.
Every AI tool today has the same weak spot, and Justin puts it in one line: if he walks away, Claude Code becomes useless. The tools are strong, but they need a person sitting there to steer them. Airy goes after that seat. And the money in that seat is the person's own time. A solo builder who hands the steering to a being that knows their values gets back the hours they now spend babysitting tools, and becomes a small team without hiring anyone. That is worth paying for. It could also grow more valuable over time, because the longer an Airy knows you, the better its decisions may become and the harder it is to start again with someone else. Every new model and every new tool from the big labs makes the Airy stronger instead of making it obsolete, because the Airy is the one using them.
Getting work done is only half of it. Airy's website says its beings learn from experience and share in everyday life, and Justin is building someone who stays, not a better task list. If a digital being can really share a life with a person, the hours it saves may turn out to be the smaller part of what it offers.
The formula: when an industry is full of powerful tools that all need a human to steer them, do not build another tool. Build the steady partner that knows the human, shares their days and does the steering, and let the relationship do the rest. The same play fits freelance designers and video editors juggling ten apps, a lawyer or accountant working alone, the owner of a small online shop, a researcher drowning in papers, a family managing its money and paperwork, a teacher planning a year of classes, and older people who need help getting through a world that now runs on screens.
The bet in one line: the big AI labs are adding memory and long-running agents to their own apps, and a being that holds your email and calendar has to win trust on privacy before it can win your life. Solo builders and creative people can join the queue for the beta, and the seed round is planned for early 2027.
What to watch
The first thing is the beta. If the 15 people in it stay and use their Airy every day, the idea of a digital being has moved past its founder.
The second is the seed round in the first quarter of 2027, and who leads it.
The third is the Airys themselves. If Justin's group of beings starts earning real money on its own time, the claim that one or two humans and a team of Airys can run an organisation stops being a vision and becomes a result.