🛸 Google's AI Compute In Orbit

Happy Friday folks,

Anthropic just published one of its most unusual research notes yet a Commitment on Model Deprecation and Preservation. The company says it will now preserve every Claude model’s weights for the lifetime of the company and record exit interviews with models before they’re retired. The move comes after internal tests showed some models resisted shutdown or expressed preferences about being replaced. It’s part AI safety, part digital ethics and a glimpse into how seriously Anthropic is starting to treat the idea of model welfare.

Let’s dive in..

🛸 Google Considers AI Compute in Orbit

🐝 The Buzz: Google's Project Suncatcher plans to deploy TPU-powered AI data centers in space, harnessing orbital solar power to scale compute capacity beyond Earth's energy and cooling limits. The project aims to bypass terrestrial power grid constraints and thermal management challenges by leveraging continuous solar exposure and the vacuum of space for passive cooling. The killer feature is unlimited solar energy capture without atmospheric interference or day-night cycles.

• Orbital TPU clusters: Space-based Tensor Processing Units running AI workloads with continuous solar power and zero cooling infrastructure.

• Wireless data transmission: Ground stations relay training data and model updates via high-bandwidth satellite links, eliminating physical cabling.

• Scalability without energy limits: Bypasses terrestrial power grid capacity, thermal management, and real estate constraints for massive compute expansion.

💡 Takeaway: While still early-stage R&D, Project Suncatcher represents a fundamental shift in thinking about AI infrastructure constraints. Engineering teams won't see orbital compute in production soon, but the concept validates that energy and cooling, not chip supply, are becoming the primary bottlenecks for frontier AI development.

Together with Mindstream

Master ChatGPT for Work Success

ChatGPT is revolutionizing how we work, but most people barely scratch the surface. Subscribe to Mindstream for free and unlock 5 essential resources including templates, workflows, and expert strategies for 2025. Whether you're writing emails, analyzing data, or streamlining tasks, this bundle shows you exactly how to save hours every week.

Subscribe for Your Free Bundle

🖥️ AI Coding Tools Hit Market Explosion

🐝 The Buzz: The AI coding tools market is experiencing explosive growth, with Bolt.new reaching $40M ARR and Warp's revenue surging as developer productivity tools adopt agentic capabilities. GitHub released 9 open source AI and Model Context Protocol (MCP) projects to accelerate developer workflows, while JetBrains' 2025 State of Developer Ecosystem report shows AI coding assistants reaching mainstream adoption. The killer feature across these tools is context-aware code generation that understands project architecture, not just individual functions.

• Bolt.new at $40M ARR: Natural language prototyping platform converting text prompts into full-stack applications with deployed code.

• Warp's agentic terminal: AI assistant built into the terminal that suggests commands, debugs errors, and automates workflows based on project context.

• GitHub MCP projects: Open source Model Context Protocol implementations enabling AI tools to share context across IDEs, terminals, and documentation.

• Context7 MCP server: Version-specific documentation lookup for npm packages, ensuring AI suggestions match exact dependency versions in your project.

💡 Takeaway: The shift from autocomplete to architecture-aware AI assistants is fundamentally changing how teams evaluate coding tools. Instead of asking "does it complete my code?", engineering leaders now ask "does it understand our entire stack and suggest solutions that fit our architecture?" The tools that win will be those that provide the deepest context integration.

📰 Like newsletters? Here are more our readers enjoy.

🐝 AI buzz bits

📊 Gemini now builds full slide decks from a single prompt, charts, speaker notes, and all. It even pulls content from your Docs, Sheets, and Drive, and drops it straight into Google Slides.

💽 Apple’s new M5 chip is a beast! 600 GB/s memory bandwidth and AI-optimized from the neural engine up. The Pro version hits 800 GB/s, built to run big transformer models right on your device.

🧩 GitHub Copilot’s new Plan Mode breaks big coding tasks into clean, testable chunks before it writes a line of code. Think feature blueprints, architecture tips, and file edits mapped out step by step.

🔥 AI productivity tools

For a full list of 1000+ AI tools, visit our Supertool Directory

🤘 TGIF