On 31 May, Felix Kjellberg, better known as PewDiePie, released Odysseus. The name is more dramatic than the software itself, which is a local-first AI workspace: a place to chat with AI models, organise notes, and run tasks, all without the data leaving your device. It is free, open source under the MIT licence, and does not report anything back to anyone.
The GitHub repository hit 77,000 stars in three weeks. That is not a normal number for a personal project.
What Odysseus Actually Does
Odysseus connects to AI models running on your machine via Ollama or to any OpenAI-compatible API endpoint. You bring the model; it provides the workspace. You get a chat interface, a document organiser, and basic automation tooling, none of which talks to a cloud server. Your conversations and data stay on your hardware.
The launch message was blunt: "The war on big tech has just begun." Kjellberg built this because he does not trust that the major AI providers are handling user data in ways he is comfortable with. The market agreed loudly.
Why the Privacy Angle Matters Here
South Africa's POPIA applies to how you handle client data. If you run an AI assistant that processes personal information about your clients and that information is sent to a US-based cloud service, you are operating under those providers' data handling terms. That is fine in many contexts, but there are industries where it is not: legal, medical, financial, and anyone dealing with information their clients gave them in confidence.
A local-first setup like Odysseus sidesteps that entirely. The data never leaves the premises. No third-party privacy policy to worry about and no data breach risk from a cloud provider you have no control over.
The Practical Limits
Running AI locally means the quality of the output depends on the model you run locally. A current laptop with 16GB of RAM can handle models in the 7-13 billion parameter range reasonably well. Those are capable but not at the level of GPT-5 or Claude Opus. For drafting documents or organising notes, that is usually fine. For complex reasoning or code generation, the cloud models still have an edge.
Odysseus is not trying to replace the cloud. It is giving people a credible alternative for the work they want to keep private. For a specific set of users — and there are apparently 77,000 of them — that is exactly the right product.