Hey {{first_name|TEDchats Fam}}!

Happy Friday! AI companies have spent years putting AI chips in bigger data centers. Now Google is trying something a little more ambitious, putting them in space.

Explore all the AI resources you’ll ever need

Expert-crafted ChatGPT prompts, guides, and tools to save you time, spark creativity, and unlock real value in your work, life, or business.
→ See What’s Inside

📣 TOP AI NEWS

Google plans to launch a refrigerator-sized satellite on October 1 carrying 4 Tensor Processing Units, giving the company its first integrated test of AI hardware in orbit. The spacecraft will run workloads for roughly 15 minutes at a time, measuring how the chips handle radiation, temperature changes, power limits, and communications delays. The mission is the opening hardware step for Project Suncatcher, a proposed network of solar-powered low Earth orbit satellites connected by free-space optical links. Google’s orbital experiment is far smaller than a data center, but it could provide the engineering data needed for prototype satellites targeted for early 2027. The concept still faces difficult questions around radiation shielding, thermal control, formation flying, launch costs, laser-link precision, maintenance, and whether space-based capacity can compete with terrestrial facilities. SpaceX, Blue Origin, and NVIDIA are exploring related ideas, placing Google’s test inside a broader race to connect AI infrastructure with abundant solar power (read more).

OpenAI's GPT-6 Sol and Luna launch makes model selection a budget decision as much as a capability decision. Sol targets advanced reasoning, coding, computer use, and long-running agent workflows, while Luna is built for faster, cheaper, high-volume work. OpenAI says Sol is 54% more token-efficient for coding than earlier models, and its pricing is $1 per 1 million input tokens and $5 per 1 million output tokens. Luna costs $0.05 per 1 million input tokens and $0.25 per 1 million output tokens, creating a clear route for developers to send routine requests to Luna and reserve Sol or Astra for harder cases. A reported DeepSWE comparison put Luna at roughly 24 benchmark points per estimated API dollar, versus 4.5 for Claude Opus 4.8 and 3.2 for Claude Fable 5, though that measures a specific coding scenario rather than overall quality. The real test will come from production workflows, where accuracy, retries, latency, tool calls, and total task cost matter more than launch-day benchmarks (read more).

Meta’s Muse has won the opening mobile skirmish, reaching No. 1 on Apple’s U.S. App Store and recording more early U.S. and Canadian daily users than ChatGPT did during its comparable first 12 days. However, the launch momentum doesn't settle the question of long-term product quality or retention. Muse reached the top chart position with fewer downloads than ChatGPT recorded during its own early rise, a result that points to Meta’s distribution advantage across Facebook, Instagram, WhatsApp, and Messenger. The app is pitched as a personal agent that can take actions, so users will judge it on permissions, reliability, privacy, and repeat task completion rather than fluent conversation alone. Early reviews describe Muse as capable and distinctive while also highlighting limits that make a full replacement for established assistants premature. Meta now has to turn curiosity into a daily habit after the app-store spotlight fades (read more).

🛠️ THIS WEEK’S AI COMBO: 1 PROMPT + 1 TOOL

⚡️ MIND SPARK

📞 Gemini Makes the Call: Google is testing Gemini Call for Me, letting eligible Pixel users have AI call businesses to ask about prices, availability and hours. It revives the promise of Duplex, but shifts AI calling toward everyday information gathering.

⚙️ Anthropic Makes Opus Cheaper: Claude Opus 5.5 cuts the cost of Anthropic’s flagship model to $4 per million input tokens and $20 per million output tokens. Lower inference costs could make advanced coding and agentic workloads more viable, while putting more pressure on rivals to compete on both capability and price (read more)

🎬 YouTube Builds a Creator Co-Pilot: YouTube is expanding AI across Studio to help creators plan content, analyze audiences, edit videos and test thumbnails. With more than $100 billion paid to creators and media companies over four years, AI is becoming a key tool for keeping the platform’s creator economy productive.

🗣️ Trump Wants to Rename AI “SI”: Donald Trump proposed calling artificial intelligence “super intelligence” in U.S. government documents, arguing that “artificial” makes the technology sound fake. The proposal clashes with the established meaning of superintelligence, which refers to systems that broadly surpass human intellectual capabilities.

🌐 AI CEOs Brief the UN Security Council: OpenAI’s Sam Altman and Anthropic’s Dario Amodei joined AI researchers and executives at the UN Security Council to discuss AI and international security. The meeting highlights growing pressure for global AI standards, though it produced political discussion rather than binding rules.

🧬 950 AI Agents Find a New Enzyme: Nearly 950 Claude agents worked for 21 hours and analyzed 210 million tokens to identify a previously uncharacterized enzyme system in bacteriophage DNA. The discovery is still only a computational candidate, making laboratory validation the crucial next step (read blog).

🎨 AI ART

Want More AI News?
Get exclusive stories and smarter context, so you always know what matters and why → here