Google’s Gemma Discovered New Scientific Knowledge. Is This AGI?

Plus: Anthropic’s Claude Skills, Spotify’s AI music deals, and Microsoft’s Copilot+ PCs debut at GITEX

Researchers from Google and Yale University used a 27-billion-parameter Gemma model to uncover a new cancer mechanism that had never been observed before. The system, called C2S-Scale 27B, analyzed single-cell data and predicted how certain drugs would change the way tumors interact with the immune system.

What the study found:

• Unexpected link: The model identified silmitasertib, a known drug, as able to help the immune system detect tumors only when small amounts of interferon were already present.
• Lab confirmation: Experiments in living cells confirmed the prediction, showing about 50% higher immune visibility with the drug combination.
• Real biological reasoning: The AI didn’t just repeat known facts but formed a new testable hypothesis that smaller models had missed.
• Research value: Teams at Yale are now studying how this discovery could guide future immunotherapy treatments.

The result gives a glimpse of what might come next in science. It shows how large models can act as research partners, finding connections that years of human study have overlooked. Still, it leaves a question hanging over the field: if machines can make correct biological predictions, how do we define discovery in the age of algorithms?

Anthropic has rolled out Claude Skills, a framework that lets users upload small folders of code, documents, and rules so the model can perform specialized tasks. The update blurs the line between chatbots and software agents, turning Claude into something closer to a customizable worker than a conversational tool.

Here’s what’s inside:
• Task-specific modules: Skills act like plug-ins that load only when needed, containing scripts and reference files.
• Cross-platform reach: The same Skills now run across Claude apps, Claude Code, and Anthropic’s API.
• Developer access: A new /v1/skills endpoint allows teams to version, deploy, and audit Skills directly.
• Executable code: Certain Skills can run small programs for complex workflows like financial modeling or branded content generation.

Claude Skills is efficient, yes, but also unsettling. Once companies start uploading internal logic, brand rules, and workflows, Claude stops being a general tool and starts becoming infrastructure. That’s a new kind of trust, and it comes with questions: who audits the skills, who sets the boundaries, and how do you know when the machine starts thinking for itself?

Spotify just signed deals with Sony, Universal, Warner, and Merlin to create AI tools that keep artists and songwriters in control of how their work is used. The company says the goal is simple: let musicians decide if AI can touch their sound  and make sure they’re paid when it does.

What’s happening

• New partnerships: Spotify and the major record labels will co-develop “responsible AI” tools that respect copyright and give artists a choice to opt in.
• Product direction: The company didn’t reveal specific features yet, but early ideas include tracking when an artist’s voice or music appears in AI-generated tracks.
• Policy shift: Spotify recently tightened its AI rules to stop spam uploads and duplicate songs, and began labeling when AI is used in a track’s creation.
• Next step: Spotify has formed a new research lab focused on developing creative AI tools that align with its “artist-first” principles.

This is Spotify’s line in the sand is that AI can coexist with musicians without erasing them. For years, streaming disrupted how music was sold. Now, AI threatens to rewrite how it’s made. If Spotify can build systems that keep credit and cash flowing back to the people behind the music, it might pull off something rare in tech: progress that still feels human.

Microsoft introduced new Surface Copilot+ PCs at GITEX Global 2025 in Dubai, marking a major push to make every Windows 11 computer an AI-powered device. The lineup includes the first Surface Laptop with built-in 5G, designed to run AI locally through the company’s Copilot+ system.

What’s new:

• Always-connected AI: The Surface Laptop and Surface Pro 12 feature next-generation chips from Intel and Qualcomm, combining AI acceleration with all-day battery life and 5G connectivity.
• Copilot built-in: Users can say “Hey Copilot” to activate voice assistance or use Copilot Vision to analyze on-screen content and perform tasks like photo edits or document reviews.
• Local intelligence: Upcoming tools like Copilot Actions will manage files, while a new Manus AI agent in File Explorer can build websites directly from local documents.
• Enterprise-ready: The devices integrate with Microsoft Intune and the Surface Management Portal, giving IT teams control over data and lifecycle management.

The launch comes as the UAE doubles down on its national AI strategy, aiming to become the “AI Capital of the World.” Microsoft’s showcase at GITEX highlights what it calls the next phase of computing: agentic AI, where systems act autonomously to plan, reason, and collaborate with users. Microsoft is making a statement. AI belongs not in the cloud, but on your desk.

⚠️ Final days. The bonus-share offer ends October 17, 2025.

Act now to get up to 25% extra shares before the window closes.

Why GenAI.works:

• Infrastructure for agentic AI systems

• Trusted by 200+ companies / 13M+ global community

• Bootstrapped to $3M ARR

👉 Put your capital where tomorrow’s economy starts.

Secure My Bonus Now

Reg CF offering via DealMaker Securities. Investing is risky.