4 min read
Own Your AI is a nine-part Top Tool Stack series on running your own local AI instead of renting it from the giants. This is Part 1 of 9. See the whole series →
In May 2026, the most-subscribed independent creator on the planet released a piece of software and, in the process, handed a very big idea to a very large audience. Felix Kjellberg, better known as PewDiePie, launched Odysseus, a free, open-source AI workspace that runs on your own computer. His pitch, stripped of the theatrics, is a sentence worth sitting up for: own your AI instead of renting it.
This is the first article in a nine-part series we are calling Own Your AI. Over the coming weeks we will walk, in plain language and without hype, through what it actually takes to run your own AI at home: why you might want to, how a complete beginner gets started, which tools and models are worth your time, what the security and privacy realities are, and how to do the whole thing without wrecking your electricity bill or the planet. This first piece is the why.
What Odysseus is, briefly
Odysseus is a self-hosted AI workspace. Instead of typing into someone else’s website, you run the software on your own machine, point it at an AI model (a local one on your hardware, or a cloud one if you prefer), and get chat, autonomous agents, deep research, persistent memory, document search, even email and calendar tools, in one dashboard. It is polished, it is genuinely open-source, and it crossed 70,000 GitHub stars in under two months, which for a brand-new project is a small phenomenon. We will get hands-on with it later in the series.
But the software is almost beside the point. The reason this matters is that a creator with a nine-figure audience spent a year making the case that you do not have to rent your intelligence from three American companies forever.
Rent versus own
Right now, most people’s relationship with AI is a subscription. You pay OpenAI, Anthropic or Google a monthly fee, you type into their box, and your data lives on their servers under their terms. That is a perfectly reasonable deal, and for a lot of tasks the models they rent you are the best on Earth. We are not going to pretend otherwise at any point in this series.
The trouble with renting is the trouble with renting anything. The landlord sets the price, and can raise it. The landlord sets the rules, and can change them. The landlord can decide your favourite model is discontinued, throttle you when the servers are busy, mine your conversations to train the next version, or simply have an outage on the morning you need it most. And when the whole industry is burning tens of billions of dollars a year with the polite expectation of one day making it back, “the price only goes up from here” is not a cynical guess. It is the business plan.
Owning is a hedge, not a religion
Here is where we part company with the louder corners of the internet. Owning your AI is not about a coming apocalypse, and running a model on your laptop will not, today, match the frontier systems on the hardest problems. Anyone who tells you your home setup instantly replaces a paid Claude or GPT subscription is selling you something.
What owning your AI genuinely buys you is independence and resilience: your data stays yours, your costs stop being a monthly meter, no company can deprecate the tool you rely on, and the open models you run keep working whether or not the vendor is having a good quarter. Increasingly, open models like Gemma, Qwen and Llama are good enough for a large slice of everyday work. Treat local AI as a hedge and a complement, keep a cloud subscription for the genuinely hard stuff if you want one, and you get the best of both.
What’s coming in this series
Next week we make the honest case for reducing your dependence on the AI giants, including the parts that are inconvenient for the “just self-host everything” crowd. After that: a genuine beginner’s guide to running your first local model, a hands-on look at Odysseus, a straight comparison against the more established tools, how to pick the right model for your hardware, the security realities, a build guide for the enthusiasts, and finally, how to run the whole thing as cleanly and cheaply as possible. Own your AI. Let’s begin.
The AI tool stack actually worth paying for
One email a week. The tools, models and moves that matter, minus the hype and the horseshit filter set to maximum. Free.