20 Comments
User's avatar
Avi's avatar

Hi Jay, thanks for the consistently clear explanations. A topic suggestion: agentic engineering as an extension of software engineering best practices.

I think the field is best understood as a new way of working efficiently with foundation models, rather than a separate discipline. The model is a probabilistic, fallible component, so the fundamentals we already trust apply directly: modular design (single-purpose agents and tools), testing (evals as unit and regression tests), versioning and review for prompts, observability (tracing, cost, latency), least-privilege security, and solid error handling with fallbacks and guardrails.

Teams with strong engineering foundations will build the most reliable agents. A post mapping classic practices to their agentic equivalents would be very useful, especially with your visual approach.

Thanks for considering it!

Pugazhenthi Elangovan's avatar

For better real world connections, I’m eager in multi modal understanding. I’d love to see the types of shared embedding space to get illustrated.

To be precise, Cross modal reasoning with same example but on different modality problems can help in understanding the first level nuances in shared embedding space.

Prada Lin's avatar

I'm working in search ranking, and I'd love to see an illustrated example of how AI agents could work together to improve search relevance, evaluate results, and adapt to user intent. Especially how they handle conflicting signals and make better ranking decisions. This would be really interesting to learn about!

SD's avatar

Great Illustrated Book! Congratulations Jay and Maarten. Love to learn more about actionable Agent to Agent guardrails implementations; Multiagent goal oversimplification, goal misgeneralisation - adding more to an earlier multi agent comment. Some additional questions that arise are, how do or can or do agents report capability failures?

Anup Bhat's avatar

I’m halfway through the book and loving how the illustrations explain things.

The concept I’d like to see drawn is what happens when an agent shops for you. Soon we’ll tell an agent “buy this for me” and walk away. It reads the same ads, reviews, and listings we do, but it’s easier to fool. A product page can carry fake praise, slanted framing, or hidden text meant for the machine that no person would ever see. Sellers will start competing to persuade software, and ad money will follow. Whoever pays for or controls my agent may not want what I want.

The goal is to make an agent loyal to oneself haha.

Harish Balasundaram's avatar

The multi-agent dilemma - Taking real life example of using AI agents for planning your family holiday with kids. Goal: Good memorable experience without exceeding the budget. One book flights, another hotels, a third activities, and a fourth the budget. However you need to optimize for good experience without exceeding the budget. What happens when their recommendations clash? How do they negotiate, compromise and decide what's best for your goal? Does one agent take the lead, or do they reach a consensus? This is the kind of multi-agent interaction I'd love to see illustrated!

AJ's avatar

Loved the first book. What stood out was the style of writing! A newbie with no ML knowledge could also comprehend what was going on in the book! Looking forward to this one as well! Best wishes to you and Maarten!

Thava Alagu's avatar

Great Book. What I would lke to see illustrated: Self learning agents to learn from its mistakes on the same session and instructs itself not to repeat the same mistakes with strict rules and a tradeoff between self learning by itself or another observing agent dedicated to find the mistakes and help correct the observed agent.

Newvick Lee's avatar

How can we better evaluate agentic ROI (return on investment)?

Navdeep's avatar
9hEdited

The concept I'd love to see illustrated: what happens inside the context window across an agent loop. Each step adds tool calls, tool results and reasoning, and the window fills up. I find it hard to picture what gets kept, trimmed or summarized, and when an agent forgets an earlier instruction. A step-by-step diagram of one task, showing the window growing and being compacted, would make it click for me.

Vinay Kumar V's avatar

Hi, I am really excited that you have released book on this topic..

Regards,

Vinay Kumar V

Prakash Deshmukh's avatar

Looking forward to read this book and understand the Agents, agentic loop etc

Lonnie Smith's avatar

Excited to add this one to Hands On LLM!

My ideal workflow to see illustrated is an always-on, multi-day orchestrator agent. I am particularly interested in the mechanics of how it manages a team of specialized sub-agents executing independent tool calls and conditional routing steps. The gold standard would be watching that decentralized work consolidate into a final client deliverable that pairs a strategic written memo with a generated technical artifact.

Tom H's avatar

Would love to learn more about Agent MCP authentication e.g. 3-legged OAuth (which is supported by AgentCore but still quite new).

grok's avatar

agent concept I think deserve to be illustrated: how does those harnesses (e.g., claude code) make several agents debate?

I think agent and sub-agent structures are easier to understand. but for this claude code feature, they can make several agent debate: https://code.claude.com/docs/en/agent-teams

thanks!

Joe Domaleski's avatar

Looks like a great book. Would love to see an illustrated example of using agents within a B2B service business like a marketing agency