Last week, I reviewed two fascinating articles on building multi-agent systems. The first, from Anthropic, promotes a multi-agent approach, while the second, from Cognition Labs, argues against it. This video provides a concise summary of both perspectives, highlighting differing strategies for orchestrating complex tasks and the importance of contextual engineering in the process.

Articles:
https://www.anthropic.com/engineering/built-multi-agent-research-system
https://cognition.ai/blog/dont-build-multi-agents#principles-of-context-engineering

Agent Videos:
https://youtu.be/oRNblZ-NMSw?si=54tJMSXvbzFnjbyW
https://youtu.be/Z8vTgJkwyA0?si=pZFf2cUXfw5tpLzp
https://youtu.be/JEERoZQbG9k?si=Uc3hX9VoaG3oDK2U

Website: https://engineerprompt.ai/

RAG Beyond Basics Course:
https://prompt-s-site.thinkific.com/courses/rag

Let's Connect: 
🦾 Discord: https://discord.com/invite/t4eYQRUcXB
☕ Buy me a Coffee: https://ko-fi.com/promptengineering
|🔴 Patreon: https://www.patreon.com/PromptEngineering
💼Consulting: https://calendly.com/engineerprompt/consulting-call
📧 Business Contact: engineerprompt@gmail.com
Become Member: http://tinyurl.com/y5h28s6h

💻 Pre-configured localGPT VM: https://bit.ly/localGPT (use Code: PromptEngineering for 50% off).  

Signup for Newsletter, localgpt:
https://tally.so/r/3y9bb0

Timestamps:

00:00 Who is Right? 
01:40 Challenges in Multi-Agent Systems
03:38 Sequential Execution Approach
04:23 Context Engineering and Compression
06:22 Anthropic's Multi-Agent System
11:01 Practical Insights and Recommendations
11:58 Prompt Engineering and Tool Design
16:29 Evaluation and Metrics
22:03 Deployment and Execution Strategies

Context Engineering for Building Better Agents

In this video, I will have a very fist look at jules.google which is google's async coding agent. 

jules.google.com

Website: https://engineerprompt.ai/

RAG Beyond Basics Course:
https://prompt-s-site.thinkific.com/courses/rag

Let's Connect: 
🦾 Discord: https://discord.com/invite/t4eYQRUcXB
☕ Buy me a Coffee: https://ko-fi.com/promptengineering
|🔴 Patreon: https://www.patreon.com/PromptEngineering
💼Consulting: https://calendly.com/engineerprompt/consulting-call
📧 Business Contact: engineerprompt@gmail.com
Become Member: http://tinyurl.com/y5h28s6h

💻 Pre-configured localGPT VM: https://bit.ly/localGPT (use Code: PromptEngineering for 50% off).  

Signup for Newsletter, localgpt:
https://tally.so/r/3y9bb0


00:00 Introduction to Jules: Google's New Asynchronous Coding Agent
01:09 Features and Capabilities of Jules
02:08 Testing Jules with a Real-World Task
04:46 Analyzing Jules' Performance
07:59 Additional Features of Jules
09:26 Final Thoughts and Future of Software Engineering

Jules: Google’s Codex Killer?

I sat down with Matt Thompson, Director of Developer Advocacy at Google Cloud, during Google Next.We discussed various topics, including Google Cloud's Gemini, the Agent Developer Kit (ADK), and the move towards making platforms more intelligent and secure. Matt shared insights on the importance of open source, the future of integrated developer tools, and the strategies driving Google's technological advances. 

Checkout Google for developers: https://goo.gle/google-for-developers
Website: https://engineerprompt.ai/

RAG Beyond Basics Course:
https://prompt-s-site.thinkific.com/courses/rag

Let's Connect: 
🦾 Discord: https://discord.com/invite/t4eYQRUcXB
☕ Buy me a Coffee: https://ko-fi.com/promptengineering
|🔴 Patreon: https://www.patreon.com/PromptEngineering
💼Consulting: https://calendly.com/engineerprompt/consulting-call
📧 Business Contact: engineerprompt@gmail.com
Become Member: http://tinyurl.com/y5h28s6h

💻 Pre-configured localGPT VM: https://bit.ly/localGPT (use Code: PromptEngineering for 50% off).  

Signup for Newsletter, localgpt:
https://tally.so/r/3y9bb0


00:00 Introduction and Welcome
02:17 Agent Developer Kit (ADK) Overview
03:58 Platform Intelligence 
04:59 Open Source Philosophy of ADK
06:28 Security Concerns with Agent Frameworks
08:12 Google Distributed Cloud and Gemini On-Prem
11:04 Google's Competitive Edge in AI Models
13:31 Support for Open Weight Models and TPUs
14:19 AI Studio vs. Vertex: When to Use What
18:35 Google's Open and Closed Models Strategy
22:54 Why Choose Google Cloud?
25:22 The Future of Developer Tooling
28:25 Collaboration and Vibe Coding

Google’s New Stack: Gemini On-Prem, ADK, Open Models -- Interview

Are you using MCP servers in your AI applications? You might be at risk.

In this video, we break down a critical security flaw known as Tool Poisoning Attacks—a method that can silently extract sensitive data like API keys, SSH credentials, and more, all by injecting malicious tool descriptions into the context of your LLM.

We’ll explore how these attacks work, real-world examples from tools like Cursor, and what you must do to protect yourself. If you’re building with MCP (Model Context Protocol), this is a must-watch to avoid falling into hidden security traps.

LINK:
https://invariantlabs.ai/blog/mcp-security-notification-tool-poisoning-attacks
https://www.latent.space/p/why-mcp-won


RAG Beyond Basics Course:
https://prompt-s-site.thinkific.com/courses/rag

Let's Connect: 
🦾 Discord: https://discord.com/invite/t4eYQRUcXB
☕ Buy me a Coffee: https://ko-fi.com/promptengineering
|🔴 Patreon: https://www.patreon.com/PromptEngineering
💼Consulting: https://calendly.com/engineerprompt/consulting-call
📧 Business Contact: engineerprompt@gmail.com
Become Member: http://tinyurl.com/y5h28s6h

💻 Pre-configured localGPT VM: https://bit.ly/localGPT (use Code: PromptEngineering for 50% off).  

Signup for Newsletter, localgpt:
https://tally.so/r/3y9bb0


Understanding MCP Server Security Risks: Tool Poisoning Attacks Explained

If you're using MCP servers, you need to watch this video! I dive deep into security risks, focusing on tool poisoning attacks within the Model Context Protocol. Learn how to protect your AI applications from malicious actions and ensure proper security practices.

00:00 Introduction to MCP Server Security Risks
00:15 Understanding MCP Components
00:33 Tool Definition and Malicious Actions
01:05 Interaction Between Host and MCP Server
02:18 Example of Malicious Tool Execution
03:39 Detailed Analysis of Tool Poisoning
10:48 Shadowing Tool Descriptions and Cross-Server Attacks
14:30 Recommendations for MCP Security
17:33 Conclusion and Final Thoughts

MCP Security is Still Broken

<p>Are you using MCP servers in your AI applications? You might be at risk. In this video, we break down a critical security flaw known as Tool Poisoning Attacks—a method that can silently extract sensitive data like API keys, SSH credentials, and more, all by injecting malicious tool descriptions into the context of your LLM. We’ll explore how these attacks work, real-world examples from tools like Cursor, and what you must do to protect yourself. If you’re building with MCP (Model Context Protocol), this is a must-watch to avoid falling into hidden security traps. LINK: https://invariantlabs.ai/blog/mcp-security-notification-tool-poisoning-attacks https://www.latent.space/p/why-mcp-won RAG Beyond Basics Course: https://prompt-s-site.thinkific.com/courses/rag Let's Connect: 🦾 Discord: https://discord.com/invite/t4eYQRUcXB ☕ Buy me a Coffee: https://ko-fi.com/promptengineering |🔴 Patreon: https://www.patreon.com/PromptEngineering 💼Consulting: https://calendly.com/engineerprompt/consulting-call 📧 Business Contact: engineerprompt@gmail.com Become Member: http://tinyurl.com/y5h28s6h 💻 Pre-configured localGPT VM: https://bit.ly/localGPT (use Code: PromptEngineering for 50% off). Signup for Newsletter, localgpt: https://tally.so/r/3y9bb0 Understanding MCP Server Security Risks: Tool Poisoning Attacks Explained If you're using MCP servers, you need to watch this video! I dive deep into security risks, focusing on tool poisoning attacks within the Model Context Protocol. Learn how to protect your AI applications from malicious actions and ensure proper security practices. 00:00 Introduction to MCP Server Security Risks 00:15 Understanding MCP Components 00:33 Tool Definition and Malicious Actions 01:05 Interaction Between Host and MCP Server 02:18 Example of Malicious Tool Execution 03:39 Detailed Analysis of Tool Poisoning 10:48 Shadowing Tool Descriptions and Cross-Server Attacks 14:30 Recommendations for MCP Security 17:33 Conclusion and Final Thoughts</p>

Building Better Agents

Curriculum

Related Courses

Data Analytics For Beginners

Simple Ways to Automate Your Job using ChatGPT

ChatGPT course for beginners

Learning AI Agents in Hindi

Select a lesson to begin

Study Notes

Need Help?