Daily AI Newsletter: Google Unveils Gemini 3, With Improved Coding and Search Abilities

Loading page…

In today’s email:

👀 Jeff Bezos to co-lead AI startup in first operational role since Amazon
🛫 Google rolls out its AI ‘Flight Deals’ tool globally, adds new travel features in Search
☠️ Peter Thiel sells entire Nvidia stake amid AI bubble fears
🧰 8 new AI-powered tools and resources. Make sure to check the online version for the full list of tools.

Google Unveils Gemini 3, With Improved Coding and Search Abilities

Key Takeaway: Google’s new Gemini 3 model is a big step up in coding, search, and document handling, positioning Google to seriously threaten OpenAI and Anthropic while still raising tough questions about whether the A.I. boom can ever pay for its massive costs.

More Insights:

Gemini 3 powers Google’s Gemini app and AI Mode in Search, blending conversational answers with live web retrieval.
Google says Gemini 3 hits 72% accuracy on a key benchmark and beats rival models on several standard tests.
The app has grown to 650 million monthly users, closing in on OpenAI’s reported 700 million weekly users.
Google AI Pro is now free for U.S. college students, a ~$240/year perk aimed at capturing the next generation of users.
Early business uses include multilingual meeting transcription, email organization, and large-scale document and data analysis.

Why it matters: This isn’t just another model upgrade—it’s Google trying to turn its search monopoly into an A.I. monopoly. If Gemini becomes “good enough” for coding, research, and daily productivity, power (and data) will centralize even more around a handful of tech giants. The open question is whether these ultra-expensive systems can generate real economic value beyond impressive demos—or whether we’re inflating an A.I. bubble that only a few companies can afford to keep alive.

Drexel’s MS in Artificial Intelligence & Machine Learning

Ready to turn curiosity into career momentum? Drexel University’s Master’s in Artificial Intelligence & Machine Learning is an interdisciplinary program led by visionary faculty across specialties. You’ll get hands-on with the newest methods and tools, while keeping ethics front and center. From deep learning to responsible AI, this is practical, industry-ready training that moves fast and goes deep.

Learn more

Jeff Bezos to co-lead AI startup in first operational role since Amazon

Key Takeaway: Jeff Bezos is stepping back into an operational role to co-lead Project Prometheus, a $6.2 billion A.I. start-up that aims to use advanced A.I. to transform engineering and manufacturing for industries like aerospace, autos, and computing.

More Insights:

Project Prometheus launches with a massive $6.2 billion war chest, making it one of the best-funded early-stage start-ups ever.
Bezos will serve as co-chief executive alongside Vik Bajaj, a former Google X and Verily leader with deep experience in ambitious, science-heavy projects.
The company targets “physical A.I.” — systems that learn not just from text, but from experiments, engineering data, and real-world manufacturing processes.
It has already hired nearly 100 employees, including top talent from OpenAI, DeepMind, and Meta, signaling an aggressive push into the frontier of A.I. research.
Project Prometheus joins a wave of labs (like Periodic Labs and Physical Intelligence) trying to use A.I. to accelerate scientific discovery, robotics, and complex physical design.

Why it matters: This isn’t just another chatbot play — it’s a bet that the next technological revolution won’t be about generating text, but about giving machines the ability to design, build, and experiment in the real world. If Bezos and his rivals succeed, the boundary between software and physical innovation could collapse, reshaping who controls manufacturing, scientific discovery, and even the future of space exploration.

An AI Podcasting Machine Is Churning Out 3,000 Episodes a Week — and People Are Listening

Key Takeaway: A tiny startup called Inception Point AI is mass-producing thousands of AI-generated podcast episodes every week, proving that automation can now scale podcasting like a factory — and people are actually listening.

More Insights:

Inception Point’s Quiet Please network cranks out ~3,000 episodes a week, has 12 million downloads and 400,000 subscribers, and aims for 150,000 episodes by the end of 2025.
Episodes cost about $1 to produce, only need ~20 listeners to be profitable, and focus on hyper-niche “edutainment” topics like biographies, localized weather, and trending figures.
The company leans on 120+ AI “host personalities” with backstories and quirks, but voices are still monotone and often lack the emotional nuance and spontaneity of human hosts.
Human review is reserved for sensitive or political topics, while multiple LLMs are used to cross-check facts — yet the sheer scale makes full editorial oversight impossible.
Competitors and industry voices see AI as useful for formulaic segments and scale, but remain skeptical it can replace the authenticity, improvisation and charisma of human-led shows.

Why it matters: This isn’t just about podcasts — it’s a glimpse of a future where content becomes infinitely cheap, infinitely targeted, and mostly machine-made, forcing us to rethink what “creativity,” “authenticity,” and even “quality” mean in a world drowning in perfectly optimized, soulless media.