Google’s massive AI day

Good morning, AI enthusiasts. Google just turned its Cloud Next 2025 event into an AI showcase of epic proportions — revealing the corporate’s strongest chip ever alongside upgrades across its entire ecosystem.

With a brand new agentic coding platform, Gemini 2.5 Flash, an agent-to-agent framework, and upgrades to image, video, voice, and music platforms, the tech giant is becoming a one-stop shop for all things AI.

In today’s AI rundown:

• Big AI day at Google Cloud Next 2025

• Google’s protocol for AI agent collaboration

• Construct your first AI voice assistant with Vapi

• Samsung’s Gemini-powered Ballie home robot

• 4 recent AI tools & 4 job opportunities

LATEST DEVELOPMENTS

GOOGLE

🤖 Big AI day at Google Cloud Next 2025

The Rundown: Google announced a flurry of AI news at its Google Cloud Next 2025 event, including a brand new agentic coding platform, next-gen AI chips, upgrades to its video, audio, and image models, a brand new Gemini 2.5 Flash model, and more.

The main points:

• Google’s Project IDX is merging with Firebase Studio, turning it into an agentic app development platform to compete with rivals like Cursor and Replit.

• The corporate also launched Ironwood, its strongest AI chip ever, offering massive improvements in performance and efficiency over previous designs.

• Model upgrades include editing and camera control in Veo 2, the discharge of Lyria for text-to-music, and improved image creation and editing in Imagen 3.

• Google also released Gemini 2.5 Flash, a faster and cheaper version of its top model that allows customizable reasoning levels for cost optimization.

Why it matters: This quick summary doesn’t do justice to the complete breadth of announcements, with Google’s AI advances reaching across every aspect of the ecosystem. Between top models, advanced chips, and now its own competing IDE, Google continues to pack a completely massive punch with each recent AI reveal.

TOGETHER WITH METRONOME

🧩 Solve the pricing puzzle for AI products

The Rundown: Pricing an AI product right can often feel like solving a puzzle without the box cover. Join 49 Palms (ex-Simon Kucher, Creator of Monetizing Innovation) and Metronome CEO Scott Woody to work out the way to construct sustainable and effective pricing for AI.

On this session, you may learn the way to:

• Apply a 9-step pricing framework

• Monetize AI–whether it’s your core offering or an add-on

• Treat pricing as a product and confidently run pricing experiments

Register now to avoid wasting your spot.

GOOGLE

🤝 Google’s protocol for AI agent collaboration

The Rundown: Google launched Agent2Agent, a brand new open protocol that allows AI agents from different developers and frameworks to speak and collaborate, with backing from 50+ tech and repair giants, including Salesforce, SAP, and PayPal.

The main points:

• A2A enables agents to find capabilities, manage tasks cooperatively, and exchange info across platforms—even without sharing memory or context.

• The protocol complements Anthropic’s popular MCP, specializing in higher-level agent interactions while MCP handles interactions with external tools.

• Launch partners include enterprise players like Atlassian, ServiceNow, and Workday, together with consulting firms like Accenture, Deloitte, and McKinsey.

• The system also supports complex workflows like hiring, where multiple agents can do candidate sourcing and background checks without humans within the loop.

Why it matters: AI agents are rapidly growing in capabilities, but unlocking their true capabilities would require an ordinary, streamlined method to interact with one another across platforms and frameworks. A2A could construct on the success of MCP, helping construct an interconnected system that permits multi-agent collaboration to tackle complex tasks.

AI TRAINING

🎙️ Construct your first AI voice assistant with Vapi

The Rundown: On this tutorial, you’ll learn the way to use Vapi to create AI-powered voice assistants that handle phone calls naturally, with customizable voices and integration capabilities.

Step-by-step:

1. Head over to Vapi and create an assistant by either scratch or choosing a starting template.

2. Select your selected AI model that can power your conversations and your required transcriber for accurate speech recognition.

3. Select a voice from Vapi’s library or create your personal voice clone.

4. Finally, add tools and integrations that allow your assistant take in-call actions, like checking calendars, scheduling appointments, or transferring to human agents when needed.

Pro tip: Give attention to optimizing response latency (to under 1.5 seconds) for a natural conversation flow. We just did an intensive workshop on constructing AI Voice agents at The Rundown University, led by Jordan Dearsley, the Founder & CEO at Vapi.

PRESENTED BY INVISIBLE TECHNOLOGIES

📈 Implement AI That Actually Delivers ROI

The Rundown: Invisible trains 80% of the world’s leading AI models — in order that they know the way to make AI work for businesses. If your organization is having trouble constructing AI into systems and workflows, this free ‘The right way to Implement AI that Actually Delivers ROI’ playbook is for you.

What you may learn:

• What’s stopping you from getting more value from AI

• What a well-implemented AI application looks like

• How businesses should use AI (hint: not only as a chatbot)

Download the free guide today.

SAMSUNG & GOOGLE

🤖 Samsung’s Gemini-powered Ballie home robot

The Rundown: Samsung and Google just announced a significant partnership to launch Ballie—a soccer ball-sized home robot teased for years at Samsung’s CES events—with Gemini AI models under the hood.

The main points:

• Ballie can roam homes autonomously on wheels, project videos on partitions, control smart devices, and handle tasks through voice commands.

• The robot will mix Gemini models with Samsung’s own AI, delivering multimodal capabilities for voice, audio, and visual inputs.

• It is going to launch within the U.S. and South Korea this summer, with plans for third-party app support also within the pipeline.

• Ballie, first revealed at Samsung’s CES event in 2020, has undergone several iterations over time, but is barely now getting an official release.

Why it matters: The AI consumer robotics race is shaping as much as be the subsequent big tech frontier, and few are as well-positioned as Samsung to guide it—leveraging their expansive SmartThings ecosystem and, now, the facility of Google’s Gemini models to deliver truly ‘smart’ home robots that would redefine the category.

QUICK HITS

🛠️ Trending AI Tools

• 🗣️ Nova Sonic – Amazon’s recent speech-to-speech AI voice model

• 🤖 Cogito v1 Preview – Recent open-source LLM model family trained using IDA

• 🔎 Gemini Deep Research – Now available with 2.5 Pro Experimental

• 🧠 Nemotron Ultra – NVIDIA’s 253B parameter open-source reasoning model

💼 AI Job Opportunities

• 🎨 Palantir Technologies – Product Designer

• 🧑‍⚖️ Horizon3 – Senior Attorney

• 🚀 Faculty – Senior Manager (Growth)

• 📋 OpenAI – Project Manager, Recruiting Operations

📰 All the pieces else in AI today

Anthropic announced a brand new premium Claude Max tier, featuring options for $100/mo and $200/mo, offering as much as 20x higher rate limits and priority access to recent features.

The U.S. government reportedly halted planned restrictions on NVIDIA’s H20 AI chips to China, following CEO Jensen Huang’s guarantees of latest U.S. investments.

Moonshot AI released Kimi-VL, a light-weight 3B-parameter vision-language model that matches the performance of models 10x larger on reasoning tasks.

UCL researchers introduced MindGlide, an AI system that analyzes MS brain scans in seconds and outperforms existing tools by as much as 60% in detecting disease progression.

The NO FAKES Act was reintroduced to Congress, with YouTube, OpenAI, IBM and others joining entertainment leaders in support of laws to combat AI deepfakes.

OpenAI launched the ‘Pioneers Program’, aiming to partner with startups on creating industry-specific model evaluations and AI systems for real-world applications.

The EU unveiled the “AI Continent Motion Plan,” committing €200B to construct 13 AI factories and aiming to triple data center capability across Europe inside seven years.

COMMUNITY

🎥 Join our next live workshop

Join our next workshop this Friday at 4 PM EST with Dr. Alvaro Cintas, The Rundown’s AI professor. By the top of the workshop, you’ll have your personal AI assistant that quickens your workflow and helps you get more done, with less manual effort.

RSVP here. Not a member? Join The Rundown University on a 14-day free trial.

🤝 Share The Rundown, get rewards

We’ll at all times keep this article 100% free. To support our work, consider sharing The Rundown with your mates, and we’ll send you more free goodies.

See you soon,

Rowan, Joey, Zach, Alvaro, and Jason—The Rundown’s editorial team