One of the biggest content creators on the planet stopped using cloud artificial intelligence (AI) and built his own, running at home on his own hardware. PewDiePie named the project Odysseus, released the code for free, and summed up the idea in one sentence: "your own AI desktop, running on your hardware." This is not a tech enthusiast's quirk. It is a clear signal of where serious AI is headed: private by default.
This post explains, in business language, what he built, why it matters for your company even if you have never heard of "self-hosted," and what the realistic path looks like for having private AI without setting up a data center in the office. At the end, a checklist to help you assess whether your current AI respects your data.
What PewDiePie built with Odysseus?
Odysseus is what is called an "AI desktop": a single application where you chat with the AI, research, write, edit images, and set the AI to execute tasks on its own. The difference that changes everything is where it runs. Instead of sending every question to a company's server somewhere else, Odysseus runs on the owner's own machine. The project's motto is direct: local first, privacy first, and no telemetry. Telemetry is the technical name for "the program keeps reporting what you do to the manufacturer." Here, it reports nothing.
In practice, the project brings together in one place what is normally spread across several paid services: conversation with memory (the AI remembers what you have discussed), agents that work autonomously, email assistant, deep research with cited sources, side-by-side model comparison, and a catalog of over 270 AI models for the user to choose from. The philosophy is "bring whatever models you want," without being locked into a single provider.
The tone of the project is lighthearted, it even features the cyclops Polyphemus from the Odyssey giving a one-star review by just screaming. But the thesis behind it is serious and written without beating around the bush on the page: it is open source and free, with no sales team, no demo request, no Trojan horse. His motivation was simple: running local AI was fun and powerful, but the existing tools felt like a step backward. So he built his own.
What "private AI" and "self-hosted" mean for your business
These terms sound like things for the IT department, but the concept is simple and matters directly to the business owner. Cloud AI is like storing your documents in a room rented from another company: convenient, but your information passes through someone else's house, where they set the rules and can read, store, or use what comes in. Private AI (self-hosted) is having the room inside your own company: the data goes in and out without passing through anyone else's hands.
Translated into what matters: with private AI, the revenue spreadsheet, the customer history, the confidential contract, and the pricing strategy do not leave your control to feed another company's server. You decide who sees it, where it is stored, and for how long. It is the difference between discussing your financials with your accountant in a closed room and shouting the numbers in the middle of a public square.
Why should a business care about this?
When one of the biggest names on the internet makes a point of taking AI off the cloud, the market signal is that privacy stopped being a detail. For your business, this translates into four concrete gains:
- Protection of sensitive data: customer, financial, and strategic information does not become raw material for an outside system. This connects directly to the LGPD, the Brazilian law that holds your company responsible for what happens to the data it collects.
- End of single-vendor dependency: if you are locked into one AI and it changes its price or rules, you have nowhere to go. The freedom to switch models is bargaining power.
- More predictable cost: instead of paying per use to a third party, with the bill rising without warning, you control spending within your own structure.
- Trust from the people who hire you: being able to tell your client that their data never leaves your house is a sales argument, especially in healthcare, legal, and finance.
"Running at home" is not for everyone (and that is fine)
Now the honest part, because selling an illusion helps no one. PewDiePie has the time, technical knowledge, and equipment to set up and maintain his own AI. He is a special case. For most businesses, copying this to the letter would be a mistake: setting up and maintaining AI servers requires expensive hardware, specialized staff, and constant maintenance. It is the equivalent of deciding to generate your own electricity instead of plugging into the grid. It makes sense for the odd case, rarely for the average business.
The important point is to separate the principle from the execution. The principle Odysseus defends, "your data, your control, no outside party watching," is absolutely right and applies to any business. The execution, building everything yourself from scratch, does not fit the reality of most companies. The good news is that there is a middle ground.
The middle ground that makes sense for businesses: a private agent built for your operation
You do not have to choose between "handing your data over to someone else's cloud" and "becoming a technology company that maintains AI servers." The path that makes sense for most is to have a private AI agent, built to order for your operation, with privacy embedded in the design: the data stays in your environment, under your rules, and whoever handles the technical side is someone who understands it, not your internal team stumbling in the dark.
It is the same logic as having a vault room. You do not need to manufacture the vault, install the alarm system, and train the guards yourself. You hire someone who does that well, define the access rules, and keep control of what is inside. Real privacy, without the cost and headache of building everything on your own.
How Steply delivers private AI for businesses
Steply builds customized AI agents for each company's operation, with privacy treated as part of the project and not as fine print in the contract. In practice, this means assembling the agent while deciding upfront where the data lives, who accesses it, and what can never leave your environment, connecting the AI to your internal systems without handing your information over to feed a third party's product.
The effect for your business is the best of both worlds: you keep the control that PewDiePie sought by running AI at home, without needing to build and maintain the infrastructure yourself. The AI works with your data, inside your rules, and you answer confidently when the client, legal, or auditors ask "where does our information go?"
If your company uses AI but has never stopped to ask where the data goes, or if you want to take advantage of AI without giving up your business confidentiality, that is exactly the problem Steply solves.
Frequently asked questions about private AI and self-hosted
What is private or self-hosted AI?
It is artificial intelligence that runs within the company's or person's own environment, instead of sending data to a third-party server in the cloud. In practice, the information does not leave your control: you decide where it is stored, who accesses it, and how long it is kept. The Odysseus project by PewDiePie is an example of this model taken to the extreme, with everything running on the user's own hardware and with no telemetry.
What is the difference between cloud AI and private AI?
With cloud AI, every question and every document passes through an external company's server, which sets the rules and can store or use that data. With private AI, the processing happens in your own environment, with no intermediary. The cloud is simpler to start with. Private AI gives you more control over sensitive data and more cost predictability.
Does private AI relate to the LGPD?
Yes, quite directly. The LGPD holds your company responsible for where the data it collects ends up. When customer information does not leave your environment to feed a third-party system, it becomes much easier to prove where the data is, who accessed it, and that it was not used without authorization. Private AI reduces the legal risk of leaks and misuse.
Does every company need to run AI on its own server like PewDiePie?
No. Setting up and maintaining AI servers requires expensive equipment, a specialized team, and constant maintenance, which rarely makes sense for a company that is not in the technology business. The realistic path for most is to have a custom-built private AI agent, with privacy embedded in the project, without having to build and care for the infrastructure on your own.
What is PewDiePie's Odysseus?
Odysseus is a free, open-source AI desktop created by PewDiePie that runs locally on the user's hardware. It brings together in one application conversation with memory, autonomous agents, an email assistant, research with cited sources, and a catalog of over 270 models. The focus is total privacy: local first, no telemetry, and no dependence on any cloud service.
Is it possible to have private AI without becoming a technology company?
Yes. The solution is to hire someone who builds the private AI agent for you, defines where the data lives, and handles the technical side, while your company keeps control of the access rules and what can or cannot leave the environment. It is like having a vault room without needing to manufacture the vault.