Welcome back. Just days after Sam Altman confidently declared that "none of our best people" would take Meta's $100 million signing bonuses, three OpenAI researchers did exactly that. Meta successfully poached Lucas Beyer, Alexander Kolesnikov, and Xiaohua Zhai—the trio who established OpenAI's Zurich office—for its "superintelligence" team. Turns out when you're throwing around nine-figure packages, loyalty has its limits.

In today’s newsletter:

• 🎒 AI for Good: Pearson and Google bring AI to the classroom

• 📱 Gemini replaces Assistant, new mobile model goes open-source

• 🚘 Uber considers funding Travis Kalanick's return to self-driving cars

🎒 AI for Good: Pearson and Google bring AI to the classroom

Pearson has signed a multi-year partnership with Google Cloud to deploy sophisticated AI-powered learning tools in K-12 classrooms. But this isn't just another education tech deal — it's part of Pearson's ambitious strategy to rebuild learning from the ground up using artificial intelligence.

What they're building: The partnership leverages Google's Vertex AI Platform, including Gemini AI models and LearnLM — Google's family of models specifically fine-tuned for learning. The system will deploy "agentic AI-powered study tools" that act as personalized tutors, adapting in real-time to each student's pace and learning style.

• Teachers get data-driven insights through BigQuery, providing comprehensive views of student progress and learning gaps for targeted instruction.
  

• The platform uses Google's Veo and Imagen technologies to automatically generate educational content, from interactive lessons to visual learning materials.
  

• Integration with Pearson's existing products, such as Connections Academy virtual schools and GED programs, extends the reach across various learning environments.
  

The bigger picture: This partnership is part of Pearson's broader AI strategy that includes similar deals with Microsoft and Amazon Web Services. CEO Omar Abbosh says the goal is to "move beyond a one-size-fits-all model to support each student on their unique learning journey."

Why this matters: Traditional classroom instruction still follows industrial-era models, designed for efficiency rather than individual learning. AI tutors could finally make personalized education scalable, giving every student, regardless of financial status, access to adaptive learning that was previously available only to those who could afford private tutoring. If successful, this represents a fundamental shift from teaching students to learning with them.

Building the Control Layer for Agentic AI with AI Gateway and MCP Servers

Learn how enterprises are deploying scalable, secure, and context-aware GenAI systems on-prem, in VPCs, or hybrid environments.

A practical and technical roadmap for AI officers, Data scientists, CDOs, CIOs, Platform heads, MLOps engineers and architects building production-grade AI infrastructure with AI Gateways and Model Context Protocol (MCP) based on our experience of enabling this for Fortune 500 Enterprises using our Platform.

Download the eBook to read the most important takeaways such as:

• Why AI Gateways are emerging as the central abstraction and orchestration layer in the GenAI stack

• How MCP allows agents to discover, authenticate, and invoke enterprise tools through machine-readable, context-aware schemas

• Deep dive into the architecture, protocols, and design principles shaping the next generation of enterprise AI infrastructure

Don’t fall behind - download now

Please don’t forget to also join us for a live session on “How leading enterprises are deploying MCP servers with centralized, security-driven architecture” on July 3. Register your presence to this webinar here.

📱 Gemini replaces Assistant, new mobile model goes open-source

Google is making two major AI moves this week that span both consumer devices and developer tools, signaling a continued aggressive push to embed AI across every level of the Android ecosystem.

Starting July 7, Gemini will gain the ability to control Android phones, similar to Google Assistant, allowing users to make calls, send messages, and manage apps, including WhatsApp and system utilities. The key change: these features now function even when users disable Gemini's data collection setting, removing a major privacy barrier that previously limited their functionality.

Users can now receive hands-free assistance without having to feed their conversations into Google's AI training systems. While the company still stores interactions for 72 hours for security purposes, this represents a significant step toward practical AI assistance with user-controlled privacy boundaries.

Meanwhile, Google released Gemma 3n after its preview last month at I/O, an open-source AI model designed to run entirely on personal hardware, even devices with just 2GB of memory. Unlike the closed Gemini system, Gemma 3n is downloadable and modifiable, processing text, images, audio, and video locally without cloud dependencies.

Technical capabilities: Gemma 3n introduces the MatFormer architecture and Per Layer Embeddings for improved memory handling, supports 140 languages, and includes a MobileNet-V5 encoder, enabling 60 FPS video processing on Pixel devices. The E4B variant is the first model with fewer than 10 billion parameters to surpass 1300 on the LMArena benchmark.

Missed Plaid Effects? Your competitors didn’t.

Plaid just launched a wave of new product updates—and the smartest teams are already moving.

From their most powerful fraud model yet to real-time cash flow insights and AI-powered tools, this year’s updates are built to give you an edge. 

Catch up now to see what’s possible. 

• Stop fraud at every step of the user journey

• Lend smarter with cash flow insights

• Speed up your integration with new AI tools

It’s all on demand: Zach Perret’s keynote, product deep dives with Plaid experts, and a fireside chat with Rocket’s CEO on the future of mortgage lending. 

You’ve got the link—now use it. 

👉 Watch now

• Is AI actually having an impact on our lives?

• Apple announces sweeping App Store changes in the EU

• Salesforce CEO says 30-50% of internal work is being handled by AI

• Meta just hired away three top researchers from OpenAI

• Meta is in advanced talks to buy AI voice startup, PlayAI

• CoreWeave is in fresh talks (again) to buy Core Scientific

• ‘Virality’ won’t last for some consumer AI startups

• DeepSeek’s momentum hits a wall thanks to U.S. export controls

• AI-generated ASMR is going viral, and it’s weirder than anyone expected

• Perplexity: Engineering Manager - AI Products

• LinkedIn: Editorial Prompt Engineer

• Pinecone: Hosts high-speed vector databases for search, chatbots and AI memory

• Salesforce: Try the world’s #1 CRM for 30 days and get 40% off when you buy*

• HeyGen Agent: The world's first creative operating system

🚘 Uber considers funding Travis Kalanick's return to self-driving cars

Eight years after Uber founder Travis Kalanick was forced out as CEO, the ride-hailing giant is considering helping fund his acquisition of a Chinese autonomous vehicle company. The potential deal would reunite Kalanick with his former company while addressing Uber's growing competitive pressure from Waymo.

Uber is in early talks to help finance Kalanick's acquisition of Pony.ai's U.S. subsidiary, according to sources familiar with the negotiations. Key details of the potential arrangement:

• Pony.ai was founded in Silicon Valley in 2016, but operates primarily in China

• The company has permits for robot taxis and trucks in both the U.S. and China

• Pony.ai went public last year with a $4.5 billion market cap

• Kalanick would run Pony.ai while keeping his CloudKitchens CEO role

• The deal would separate Pony.ai from its Chinese parent amid regulatory pressures

The discussions signal mounting pressure on Uber from autonomous vehicle services. CEO Dara Khosrowshahi sold Uber's self-driving unit to Aurora in 2020 after years of losses and a fatal 2018 crash involving a test vehicle.

Kalanick has criticized that decision. At the Abundance Summit in March, he said Uber was "really only behind Waymo but probably catching up" when he was ousted, adding: "Wish we had an autonomous ride-sharing product right now. That would be great."

Pony.ai has prepared its U.S. operations for sale since 2022, including creating separate source code. The company raised $260 million in its U.S. IPO and operates testing programs across multiple states.

The potential partnership marks a dramatic reversal. During Kalanick's tenure, Uber invested heavily in autonomous technology, poaching Carnegie Mellon robotics researchers and acquiring the startup Otto, which led to a bitter legal battle with Waymo that was settled in 2018.

Khosrowshahi has since pursued partnerships, integrating vehicles from companies like Waymo rather than developing in-house technology. But as Waymo expands and competitors emerge, Uber appears increasingly concerned about being left behind in the autonomous transition.

The talks remain preliminary, and financial details haven't been disclosed. It's unclear what role Uber would take as an investor or how the partnership would function operationally. What makes Pony.ai attractive is its dual-market presence and regulatory positioning — the company already holds operational permits in both the U.S. and China, providing any acquirer with immediate market access that would typically take years to develop independently.

This would be one of Silicon Valley's most fascinating comebacks. Kalanick, ousted amid scandals, returning to challenge his successor's autonomous vehicle strategy — with Uber's own money.

Khosrowshahi's 2020 decision to sell Uber's self-driving unit appeared prudent, as it cut losses and focused on profitability. But Waymo's expanding robotaxi service now poses an existential threat to Uber's driver-dependent model. Funding Kalanick's acquisition would essentially acknowledge that the partnership strategy is insufficient.

For Kalanick, this represents validation that autonomous vehicles were always critical to Uber's survival. Whether that vision translates to execution remains the trillion-dollar question.

🤔 Your thought process:

Selected Image 1 (Left):

• “Peripheral vision gut feel has me 3-0 this week!!!”

• “AI forgot the kitchen line. Obviously it doesn’t play.”

Selected Image 2 (Right):

• “Thought there was to much that looked accurate like the weeds around the fence.”

• “I thought the absence of shadows in [the other image] seemed unrealistic.”

💭 A poll before you go

The Deep View is written by Faris Kojok, Chris Bibey and The Deep View crew. Please reply with any feedback.

Thanks for reading today’s edition of The Deep View! We’ll see you in the next one.

P.S. Enjoyed reading? Take The Deep View with you on the go! We’ve got exclusive, in-depth interviews for you on The Deep View: Conversations podcast every Tuesday morning. Subscribe here!

P.P.S. If you want to get in front of an audience of 450,000+ developers, business leaders and tech enthusiasts, get in touch with us here.