發生了什麼事
TechJuice reports that PewDiePie unveiled Ajax on October 2, an uncensored AI model with 9 billion parameters designed to run on home PCs through his Odysseus workspace. The model is fine-tuned from Alibaba’s Qwen3.5-9B and functions as an autonomous agent handling tasks like web search and email management. OpenAI banned PewDiePie’s account twice during development for , a process where outputs from one model train another. To remove safety refusals, PewDiePie used an open-source tool called Heretic. While early testing showed roughly 90% task completion rates, the model weights remain private as of launch, with the download link directing users to a data contribution page rather than the model itself.
TechJuice reports that PewDiePie unveiled Ajax on October 2, an uncensored AI model with 9 billion parameters designed to run on home PCs through his Odysseus workspace. The model is fine-tuned from Alibaba’s Qwen3.5-9B and functions as an autonomous agent within the workspace, handling daily tasks including web search, email management, calendar scheduling, and to-do organization.
OpenAI banned PewDiePie’s account twice during the development process. The first ban explicitly cited , where outputs from one model are used to train another. After reinstatement, a second ban occurred when he ran the model again to create seed data. Distillation violates OpenAI’s terms of service because it uses proprietary model outputs as training data for competing products.
To remove safety refusals from Ajax, PewDiePie used an open-source tool called Heretic, which identifies and strips refused prompts from the model’s behavior. He acknowledged the trade-off, stating that fewer guardrails mean more room for misuse, but maintained that Ajax retains restrictions against harming people and is not designed to provide dangerous actionable instructions.
Early testing showed roughly 90% task completion rates, though PewDiePie described Ajax as a work in progress. His development setup includes 10 GPUs with eight modified RTX 4090 cards carrying 48 GB memory each. Despite this high-end infrastructure behind development, Ajax targets standard home PCs for end-user deployment.
The model weights remain private as of launch. The download link currently directs users to a data contribution page rather than the model itself. The Odysseus workspace operates under an AGPL-3.0-or-later open-source license and remains free to use.
為什麼這很重要
This launch represents a significant shift in the local AI landscape by targeting mainstream consumers rather than just enterprise developers. By leveraging his massive YouTube audience, PewDiePie introduces self-hosted, privacy-focused AI models to a demographic that typically relies on cloud-based services like ChatGPT. The incident also highlights the growing tension between large AI companies protecting their intellectual property through terms of service and independent developers who use to build competing products. The explicit removal of safety guardrails, while acknowledged as a trade-off for misuse risk, positions Ajax as a distinct alternative in the market for uncensored local models.
Ajax joins a growing category of locally deployable AI models that prioritize user privacy and reduced censorship over cloud-based convenience, alongside Meta’s Llama, Mistral, and Alibaba’s Qwen family.
PewDiePie’s entry differs primarily through his creator’s massive audience reach. With over 110 million YouTube subscribers, his promotion of local AI deployment introduces self-hosted models to a mainstream consumer audience that enterprise-focused open-source projects typically miss.
The repeated bans by OpenAI highlight the growing tension between large AI companies protecting their models and independent developers building alternatives using techniques.
互動機制：它實際上是如何運作的
以互動方式探索這項發展背後的基礎技術。
crm_get_transaction(id='4092').A route planner searches possible journeys using explicit rules. What does this illustrate about AI?
接下來看什麼
Monitor whether the model weights are eventually released publicly, as they are currently private. Watch for any further legal or terms-of-service actions by OpenAI regarding the practices used. Observe how the mainstream audience reacts to the privacy and censorship trade-offs of using a locally hosted, uncensored model compared to cloud-based alternatives.
The current status of the model weights is private, with the download link directing to a data contribution page; users should watch for an actual release of the weights.
The legal and terms-of-service implications of the process may lead to further actions by OpenAI or other major AI labs.
The practical usability of the model on standard home PCs, given the high-end hardware used for development, remains a key factor for mainstream adoption.