AI Friends Chat Together: Google’s New Agent2Agent

In partnership with

Date: 19-Apr-2025

Hey AI enthusiast,

And just when you think you’ve seen it all…there’s always One More Thing in AI.

In this edition:

• Google’s New Agent2Agent Framework: Helping AI Tea …

• Amazon released a speech-to-speech AI called "Nova …

• UAE Business owners- Close More Deals than Ever w …

• 🦙 Llama 4: The Open-Source AI Arsenal Has Arrived …

• More AI news scoured for you from X.com

  • Samsung announced a partnership with Google.

  • Runway released Gen 4 Turbo

  • Midjourney released V7

  • ElevenLabs launched its MCP server

If you’re building or just trying to stay ahead with AI news—this one is for you. I do the heavy lift of curating the news so that you don’t have to!

Best,

Renjit

PS: Founders- In case you want to prototype and test AI automation /Agentic AI deployment in your business, schedule an appointment here:

Google’s New Agent2Agent Framework: Helping AI Teams Work Together

Imagine your AI helper finding job candidates, scheduling interviews, and running background checks—all without your help. Google’s new Agent2Agent (A2A) is making this possible.

Here’s how it works:

✅ AI Teamwork Made Easy

A2A lets different AI programs talk and work together, even if they’re built differently. They can share tasks and skills to finish complicated jobs like planning deliveries or hiring new workers.

📌 Built for Businesses

A2A helps with quick jobs or tasks that take months. It works securely and can handle messages through text, sound, and video. Plus, you see live updates as things get done.

🔑 5 Simple Ideas Behind A2A

• Made for teamwork—no shared memory or special tools needed.
• Uses common internet rules (HTTP, JSON-RPC, SSE).
• Secure right from the start.
• Good for short tasks and long projects.
• Works easily with text, audio, and video.

🤝 Better Together with MCP

A2A manages how AI agents talk. Another system that we wrote about earlier, called MCP, helps agents use different tools and internal databases (of companies). Together, they create smart AI teams that handle tasks better.

🌐 Open for Everyone

Google has made A2A open source for anyone to use. Big companies like Deloitte, Cognizant, BCG, and Wipro are already helping develop it- you can be sure that client deployments will follow soon.

So, What’s the Big Deal?

Google’s A2A means companies can easily have many AI programs working together. For businesses and startups, this means faster, easier workflows, and lots of new ways to use AI. Take a look at this framework early if you are building AI agents.

Now’s the perfect time to start thinking about how A2A could help your business.

Amazon released a speech-to-speech AI called "Nova Sonic"

Available on Amazon Bedrock, the model generates speech with a 1.09 sec latency, outperforming the best OpenAI models at 20% cost. This latency is important when you try to book a customer’s reservation in a restaurant (has to sound human and be delivered with reasonable delays or else it becomes a toy).

Nova Sonic delivers speech understanding and generation through a single, unified model. Makes it easier for builders to develop voice applications that maintain important context and nuance for customer service, AI agents, and other use cases across industries. It’s available in Amazon Bedrock now. The below link takes you to a Travel Agency use case where the AI Agent takes the booking.

So what?

AI powered Voice is going to be the next frontier for business discoverability (like Search is today), and this means that as folks who are thinking about building businesses, we need to be using voice apps to serve our customers.

Look forward to seeing what you build with Nova Sonic and take advantage of the installed based of millions of devices that have Alexa installed.

UAE Business owners- Close More Deals than Ever with AI

Build targeted lead lists, send custom outreach, get sales coaching for your sales team, and keep your CRM up to date with the our powerful AI automations.

We work with mid-sized businesses in the UAE to solve their sales related problems.

Are you a business owner? Interested in discussing how we can increase sales and reduce your operating expenses? We have a free launch package worth AED 25,000 to give to 10 selected businesses every quarter.

Speak to us (book time with us here) to see if you qualify.

🦙 Llama 4: The Open-Source AI Arsenal Has Arrived 

Meta just dropped a family of models that could reset the playing field — and it’s all open source.

🚀 1. Meet Llama 4

Scout- A compact powerhouse for developers.

• 17B active parameters

• 10M token context window

• Outperforms: Gemma 3, Gemini Flash-Lite, Mistral 3.1

• Runs on a single H100 GPU. Built for speed, code, and image understanding

Scout handles long documents and code easily — and does it fast.

🧠 2. Meet Llama 4

Maverick- The new king of multimodal AI.

• Uses 128 expert models for modular reasoning

• Best-in-class image grounding

• Outperforms GPT-4o & Gemini Flash. Matches DeepSeek v3 in reasoning & code

• Half the size, lower compute. Scores 1417 on LMArena. Also runs on a single GPU

This is not a toy model — it’s the multimodal engine for real-world use.

🧬 3. Coming Soon:

Llama 4 Behemoth- Still training — but already crushing:

• GPT-4.5

• Claude Sonnet 3.7

• Gemini 2.0 Pro

Especially strong on STEM benchmarks, this is the model behind Scout & Maverick.

💡So what?

• Fully open source -AI is not going to be dominated by closed players like Open AI. Meta is always going to be a contender to reckon with and that is good for soloprenuers.

• Multimodal-first architecture. No GPU farms needed

• Made for devs, researchers, and builders (that is you!)

• You can run it today — no paid API keys, no gatekeepers.

More AI news scoured for you from X.com

• Samsung announced a partnership with Google.
  Under the deal, the company will power its Ballie home robot with Google's Gemini and its own multimodal AI models. The football-like robot will roll out in South Korea and the U.S. this summer. Finally, a robot that looks “cute” enough to be welcomed to homes!

• Runway released Gen 4 Turbo 

  A faster version of its latest video model that can produce 10-second videos in just 30 seconds. Best part: it's available to all Runway users, including those on the free 'Basic' tier that provides 125 one-time credits

• Midjourney released V7, the latest version of its image generation model.
  
  The AI brings improved quality in generations, enhanced prompt adherence, and a voice-capable Draft Mode to help users iterate over their creations

• ElevenLabs launched its MCP server integration
  
  The move enables platforms like Claude and Cursor to access AI voice capabilities via simple text prompts. It can be used to create automated agents for tasks like performing outbound calls (founders alert!).

Help us to grow, dear reader and win!

If you want to interact with me on social media, follow me on LinkedIn and Twitter.