AI is no longer just improving models.

It’s expanding into systems.

Last week made that clear:

• AI agents are executing workflows across enterprise tools
• Models are generating code, visuals, and decisions in real time
• Infrastructure investments are reaching gigawatt scale
• Governments are starting to shape how AI can and cannot be used

At the same time, a deeper shift is happening:

AI is moving from tools → to operators.

From assisting humans → to acting on their behalf.

Week 11 of Silism breaks down the signals behind this transition.

If you want to understand where AI is heading — not just what launched — subscribe.

Silism is built for people who want clarity in a fast-moving space.

What trend are you watching most closely right now?

Need-to-Know

• A joint Meta and NYU study shows multimodal models can learn across text, images, and video, with unlabeled video emerging as a major future training source.
• Anthropic CEO Dario Amodei said he remains uncertain whether Claude is conscious, prompting Elon Musk to respond publicly, “He’s projecting.”
• The Trump administration is preparing stricter AI contract policies following tensions between the Pentagon and Anthropic.
• Oracle and OpenAI reportedly abandoned plans to expand a Texas data center, although the broader Stargate initiative continues.
• OpenAI is again postponing ChatGPT’s “adult mode” rollout to prioritize intelligence improvements and deeper personalization.
• Microsoft introduced Copilot Cowork, an AI agent capable of executing workflows across Microsoft 365 apps using Claude-based capabilities.
• Nearly 40 employees from OpenAI and Google submitted an amicus brief supporting Anthropic in its legal conflict with the Pentagon.
• Nvidia is reportedly preparing NemoClaw, an open-source AI agent platform designed for enterprise task automation.
• X added a toggle to prevent @Grok from modifying uploaded images, though the limitation can still be bypassed.
• Grok has reached approximately 314 million visits, becoming the third most-used generative AI platform globally.
• Coinbase’s CEO stated that AI agents may soon outnumber humans in executing transactions and participating in crypto-based economies.
• Anthropic released Code Review for Claude Code, enabling multi-agent deep review processes for pull requests.
• OpenAI launched dynamic visual explanations in ChatGPT, allowing interactive learning for math and science concepts.
• NVIDIA introduced the Nemotron-Terminal framework and dataset to generate training data for terminal-based AI agents.
• Google upgraded Gemini for Workspace, enabling automated document, spreadsheet, and presentation generation using data across Gmail and Drive.
• Nvidia partnered with Mira Murati’s ThinkingMachinesLab, committing at least one gigawatt of compute for model training.
• Amazon expanded its Health AI assistant beyond One Medical into its main platform, supporting medical queries and doctor connections.
• RevenueCat reported that AI-powered apps monetize strongly early but struggle to maintain long-term subscriber retention.
• Ford launched FordProAI, integrating generative AI into fleet management to analyze vehicle data and support decision-making.
• Cursor is reportedly negotiating a funding round that could value the company near $50 billion.
• Amid tensions with the Pentagon, Anthropic launched the AnthropicInstitute to study long-term societal and economic AI impacts.
• Meta introduced four new MTIA chips to support generative AI and recommendation systems across its platforms.
• Nvidia invested $2 billion into AI cloud company Nebius, boosting its market value significantly.
• Anthropic enhanced Claude with shared context across Excel and PowerPoint alongside reusable workflow “Skills.”
• Replit launched Agent4 and raised $400 million, positioning its AI as capable of building and operating a startup autonomously.
• Anthropic updated Claude to generate embedded charts, diagrams, and visual outputs directly in conversations.
• OpenAI plans to integrate Sora video generation into ChatGPT to increase engagement and usage.
• FacebookMarketplace added Meta AI features such as automated replies, listing generation, and seller summaries.
• Google introduced Gemini task automation on smartphones, allowing assistants to directly operate apps for real-world actions.
• Bumble launched Bee, an AI dating assistant that adapts to user preferences and relationship goals.
• Microsoft released CopilotHealth, connecting medical data and wearables to provide health insights and recommendations.
• Google launched AskMaps, a Gemini-powered feature that answers complex location-based questions and suggestions.

TOGETHER WITH US

Silism is becoming a reference point for AI and technology insights.

From engineers to founders, from startups to global tech companies, professionals follow Silism to understand what actually matters.

We focus on signal, not noise.

If you are building, investing, or operating in AI — Silism helps you stay ahead.