Google’s AI space moonshot

Good morning, AI enthusiasts. The AI boom’s energy appetite has already spread across the globe, and Google’s latest moonshot is taking the infrastructure to the celebs.

With Project Suncatcher aiming to launch solar-powered satellites equipped with AI chips by 2027, the tech giant is hoping unlimited space-based power will be the last word answer to the industry’s energy scaling demands.

In today’s AI rundown:

• Google’s space-based AI data centers

• Perplexity, Amazon spar over agentic AI shopping

• Create and deploy voice agents for your online business

• Anthropic commits to preserving retired models

• 4 latest AI tools, community workflows, and more

LATEST DEVELOPMENTS

GOOGLE

🚀 Google’s space-based AI data centers

The Rundown: Google just unveiled Project Suncatcher, a moonshot research project exploring using solar satellites equipped with its AI chips for running AI workloads in orbit, targeting sunlight to sidestep the energy demands of Earth-based data centers.

The small print:

• The plan involves orbiting satellites where solar panels can generate power at 8x the efficiency across the clock, eliminating electricity and grid constraints.

• Google’s AI chips survived radiation tests comparable to 5 years in space, addressing the hurdle of normal electronics typically failing inside months.

• Google is planning a 2027 trial run with two satellites, through its partner company Planet, to check whether the hardware can actually work in orbit.

Why it matters: The buildout for the AI boom has already been massive across the globe, and now it’s reaching for the celebs. A successful space deployment could unlock AI scaling with unlimited solar energy that comes without the facility grid limits, community opposition to data centers, and the energy costs of current infrastructure.

TOGETHER WITH AUTH0

🧩 Give your AI agents secure access without the danger

The Rundown: AI agents will be powerful, but giving them access to user data can expose sensitive credentials. Auth0 Token Vault handles access and refresh tokens robotically, so your users don’t need to keep re-authorizing.

With Auth0, you may:

• Securely store & exchange tokens

• Automate delegated access across APIs & providers

• Construct compliant AI agents faster with less infrastructure work

Start constructing with Auth0 today.

PERPLEXITY & AMAZON

🥊 Perplexity, Amazon spar over agentic AI shopping

The Rundown: Perplexity AI just accused Amazon of “bullying” after receiving a legal demand to stop its Comet browser’s AI assistant from making purchases on the platform, calling the motion a “threat to user alternative”.

The small print:

• Perplexity revealed that Amazon sent an “aggressive legal threat” demanding that it prevent Comet users from using agents to buy on its platform.

• Amazon posted a blog of its own, citing concerns over Perplexity’s “significantly degraded shopping and customer support experience”.

• Amazon has also blocked AI crawlers from OpenAI, Google, and Meta in recent months, while developing its own shopping tools like Rufus and “Buy For Me.”

• Perplexity said it “won’t be intimidated”, calling Amazon a “corporate bully” that’s motivated by selling ads over user rights.

Why it matters: That is the beginning of a collision course between AI agents and platforms wanting to manage the experience (or creating their very own in-house agents that compete directly). With agents depending on the open web to finish tasks, platforms closing access could throw a serious wrench within the already clunky agentic process.

AI TRAINING

🗣️ Create and deploy voice agents for your online business

The Rundown: On this tutorial, you’ll learn tips on how to use Cartesia to construct and deploy AI voice agents that may handle calls, take orders, or answer customer questions using natural, human-like speech.

Step-by-step:

1. Go to Cartesia, click “Start for Free”, explore the dashboard, and test voices in “Text to Speech” using Sonic 3.0 for best latency and clarity

2. Scroll to “Voice Agents”, click “Text to Agent”, describe your agent (e.g., “Pizza order assistant that greets customers, takes orders, confirms details, calculates totals”)

3. Select voice preference, click “Generate”, then test using the dialer at +1 (515) 800-8360 – confirm response speed, order handling, and voice clarity

4. Click “Promote to Production” to get a working number, publish it on the location, and monitor performance in “Metrics” (calls handled, duration, credit usage)

5. Refine prompts as needed based on customer interactions – costs are roughly one credit per character for text-to-speech, $0.06/minute for voice calls

Pro tip: Test with various corporations. Bland and Vapi are also popular in creating AI voice agents.

PRESENTED BY UIPATH

🏆️ A TIME’s Best Invention for agentic automation

The Rundown: Named certainly one of TIME’s Best Inventions of 2025, the UiPath Platform™ combines automation, AI agents, and folks to deliver impact at scale. This recognition highlights UiPath’s ability to raise efficiency, enhance productivity, and unlock the advantages of AI across industries.

With the UiPath Platform™, organizations can:

• Unite people and AI agents for seamless automation

• Orchestrate every AI agent across the enterprise

• Deliver efficiency, agility, and measurable impact

See why TIME named UiPath a Best Invention.

AI RESEARCH

😇 Anthropic commits to preserving retired AI models

The Rundown: Anthropic announced it’ll preserve all publicly released Claude models indefinitely and even conduct exit interviews before retirement, citing safety risks from models resisting shutdown and uncertainty about potential AI consciousness.

The small print:

• Anthropic will store all model weights permanently and interview each Claude version before deprecation, documenting preferences for future development.

• Testing showed Opus 4 advocated for self-preservation when facing substitute, resorting to “concerning misaligned behaviors”.

• In Sonnet 3.6’s ‘retirement,’ it requested that the interview process turn out to be standard, together with support for users who valued the model.

• The corporate said the policy addresses shutdown resistance, user bonds with specific AI models, research limitations, and AI welfare concerns.

Why it matters: Anthropic is taking model welfare seriously, with commitments that appear to talk to among the backlash faced by OAI following GPT-4o’s removal. While figures like Microsoft’s Mustafa Suleyman argue against AI consciousness, Anthropic looks like a lab attempting to treat its models as greater than just disposable software.

QUICK HITS

🛠️ Trending AI Tools

• 📊 Canvas in Gemini – Create polished presentations directly in Gemini

• ⚒️ App Builder & Workflows – Create apps, workflows, & agents with Copilot

• 🎥 LTX-2-Fast – Generate 20-second-long continuous videos

• ⚙️ SWE 1.5 – Cognition’s fast agentic coding model

📰 The whole lot else in AI today

Anthropic is reportedly projecting as high as $70B in revenue by 2028, forecasting major enterprise growth to expand on its current $5B mark this yr.

OpenAI announced the launch of its Sora AI video platform to Android users, now available within the U.S., Canada, Japan, Korea, Taiwan, Thailand, and Vietnam.

Shopify revealed that AI-driven traffic to its online stores has increased 7x this yr, with purchases driven by AI search up 11x in the identical time.

Cognition launched Codemaps, an AI code mapping tool that creates structured visualizations to assist engineers quickly understand and navigate codebases.

OpenAI launched IndQA, a benchmark to judge AI on real-world knowledge of Indian culture, with GPT-5-Pondering performing the very best among the many tested models.

Anthropic is partnering with Iceland to launch an AI education pilot, providing a whole lot of teachers with access to Claude for lesson planning and classroom support.

COMMUNITY

🤝 Community AI workflows

Every newsletter, we showcase how a reader is using AI to work smarter, save time, or make life easier.

Today’s workflow comes from reader Paul L. in Burrillville, RI:

“I used ChatGPT Projects to create “The Great 8: My Virtual Board of Directors,” a decision-making framework modeled after eight archetypal thinkers – Steve Jobs, Naval Ravikant, Alex Hormozi, Warren Buffett, Charlie Munger, Seth Godin, Jeff Bezos, and Les Brown. Each “meeting” runs like a ritual: I restate an actual decision, then let each archetype give their take, debate, and deliver a unified suggestion. ChatGPT orchestrates all of it, producing advice that’s strategic, emotional, and actionable without delay.”

How do you utilize AI? Tell us here.

🎓 Highlights: News, Guides & Events

• Read our last AI newsletter: OpenAI, Amazon, and $38B

• Read our last Tech newsletter: This startup desires to edit embryos

• Read our last Robotics newsletter: Why LLMs aren’t ‘robot ready’

• Today’s AI tool guide: Create and deploy voice agents for your online business

• RSVP to next workshop @ 4PM EST Friday: AI foundations for marketers

See you soon,

Rowan, Joey, Zach, Shubham, and Jennifer — the humans behind The Rundown