The New Era of Personal AI: Hardware, Healthcare, and Open Models

• Ghost AI has launched Core, a personal computer powered by on-device models that remembers user details across apps and devices.
• Nolla Health received regulatory approval to test an end-to-end AI doctor, starting with dermatology in Utah.
• Reflection previewed Beam, an open-weight model for coding that claims performance parity with China’s GLM 5.2 using significantly less compute.
• Pope Leo XIV recently emphasized the urgent need to distinguish human art from machine-generated content, citing a lack of "spark."
• YouTuber PewDiePie returned from retirement to release Ajax, an uncensored open-weight AI model.

Zain Javaid, a 19-year-old former quantitative analyst who recently raised $11 million, has emerged from stealth mode with Ghost AI’s new product: Core. This device is marketed as the first personal computer built entirely around AI-first principles. Unlike traditional computers that rely on cloud-based assistance, Core utilizes on-device models to provide continuous, 24/7 AI features.

The hardware is designed to maintain context over time, remembering key details about the user’s life and completing tasks across connected applications and devices. This launch places Ghost AI in direct competition with major tech giants who are also exploring AI-integrated hardware, including Google’s Googlebook, Apple’s new Mac Mini, and Microsoft’s Surface Laptop Ultra. The Core device aims to move beyond simple app-based interactions by offering a persistent, intelligent layer over the entire computing experience.

In the medical sector, Nolla Health has achieved a regulatory milestone by receiving approval to test one of the first end-to-end AI doctors. While other AI tools currently exist that can only renew prescriptions previously written by human physicians, Nolla’s system handles the entire medical process autonomously.

The pilot program is dermatology-focused and is rolling out in Utah. In this setup, the AI assesses patient symptoms, prescribes appropriate treatment, and manages follow-up care. Human doctors remain on call in the background to oversee the process but do not perform the initial assessment or prescription steps. This represents a significant shift from auxiliary AI tools to primary care providers within specific medical niches.

A New York-based laboratory named Reflection has expanded the "Western open frontier" by previewing Beam, its first open-weight model designed for coding, reasoning, and agentic workloads. The lab claims that Beam’s performance in advanced reasoning matches that of China’s GLM 5.2 model. Crucially, Beam achieves this parity while using three to four times less computational power than its competitor.

This release positions Beam within a small but growing group of Western laboratories developing open models, joining established players like Meta, Nvidia, and Thinking Machines. The ability to match high-end performance with reduced compute requirements addresses a major barrier to entry for developers and organizations seeking efficient, transparent AI solutions.

For IT teams managing constant security threats, automating the aggregation of news can save significant time. Using Grok Bot, you can create a recurring security-news brief tailored to your specific needs.

1. Open Grok Bot and initiate a new Bot specifically for cybersecurity monitoring.
2. Define the topics the bot should track, such as ransomware, zero-day exploits, major vulnerabilities, phishing campaigns, cloud security issues, and data breaches.
3. Instruct Grok Bot to prioritize sources like CISA (Cybersecurity and Infrastructure Security Agency), vendor advisories, recognized security researchers, and reputable cybersecurity publications, while deprioritizing unverified social media posts.
4. Use a sample prompt structure: “Review the latest cybersecurity developments relevant to [company/industry]. Prioritize vulnerabilities, breaches, ransomware, phishing campaigns, and vendor advisories affecting [tools/vendors]. For each item, explain what happened, why it matters, severity, affected products, recommended action, and add a source link.”
5. Run the brief manually once to refine topics, sources, and vendors based on signal versus noise. Save this workflow as a Skill, then create a Routine to execute it every weekday morning or on your preferred schedule. Format the output into sections: ‘Critical Now’, ‘Watchlist’, ‘Recommended Actions’, and ‘Source Links’ for quick scanning by your team.

This week’s conversation around AI has been dominated by unexpected figures rather than traditional tech CEOs. Pope Leo XIV generated nearly 25 million views with a post stating that AI lacks the “spark of humanity.” He argued it is becoming urgent to distinguish human art from machine-generated art. This intervention follows reports that Anthropic co-founder Chris Olah lobbied the Pope’s advisors to consider the possibility that Claude might be conscious. While Anthropic has not publicly claimed consciousness, they have hinted at the concept, prompting the religious leader’s response.

Simultaneously, YouTuber PewDiePie ended his retirement with a video titled “Thank you and goodbye,” only to return two weeks later announcing Ajax, his own uncensored open-weight model. The video serves as an introduction to open-weight AI and details the history of why OpenAI banned him twice. These events highlight a growing public engagement with the ethical and creative implications of AI technology.

The current week illustrates the diversification of AI into tangible hardware, regulated healthcare, and efficient open-source models. From Zain Javaid’s Core computer to Nolla Health’s autonomous doctor pilot, the technology is moving from abstract software to integrated, real-world applications. As open-weight models like Beam challenge proprietary systems and cultural figures debate the nature of machine creativity, the AI landscape continues to expand in both technical capability and societal impact.