Newsletter

Issue #67 | All you need to know about the updated MCP spec - stateless core, extensions, task runtime, current SDK support and integration considerations.

Issue #66 | How do we get LLM Agents to Plan, Review and Monitor Progress on complex tasks?

Issue #65 | Notes from building an interactive reader map for my book site with Claude Design and Claude Code in about 45 minutes.

Issue #63 | What 500,000 lines of production agent code teach us about building and optimizing agents

Issue #61 |Three Extensions That Turn your For Loop into an Agent like Claude Code

Issue #60 | When AI Agents Flood One Side of a Platform, Everyone Loses

#59 | Managing context growth, preventing early stopping, and measuring what works

#58 | The pattern behind Claude Code's extensibility - and a working implementation in 100 lines of Python.

#57 | On where general-purpose coding agents like Claude Code shine, where they break, and how to integrate them into your workflow.

#56 | The bitter lesson isn’t dying. It’s being amortized.

Issue #55 | Early agents (LIDA, AutoGen) supported actions via code. Then the industry moved to action via tools/structured output. And now we are back again to code. What happened?

#54 | Five practices to make your AI-assisted coding more production-ready

Issue #53 | How to intercept agent operations for logging, safety, rate limiting, and compliance - the middleware pattern for production AI agents.

Issue #52 | How AI agents work under the hood - the execution loop that enables reasoning, tool calling, and iterative problem solving.

Issue #51 | Sometimes the arrangement is the problem

Issue #50 | A list of books I'd recommend to folks interested in understanding and building AI and Agents in 2026

Issue #49 | It took only 2 years-ish, but I finished writing the multi-agent systems book and it is now available!

Issue #48 | somber reflections on how AI coding tools are changing developer behavior, cognition, and dependency (the slope we are all sliding down)

#47 | Add reusable validation, security, and logging without touching your core agent logic

#46 | If Semantic Kernel and AutoGen joined forces …

#45 | A deep dive into implementing Tools, Resources, Prompts, Roots in the MCP

#44 | New to MCP? Heres how to build your first MCP server and a Host Application that integrates and uses the server.

#43 | 6000 subscribers, and my reflections on why I write about AI .. and why you should too!

#42 | A summary from a keynote talk I gave at the 2025 AI.Engineer World's Fair Conference.

#41 | An in-depth analysis on Agentic AI startups using data from the YCombinator startup accelerator.

#40 | Agents grow up!

#39 | A thought experiment on how advances in strong AI could completely drive the scientific research process in the near future.

#38 | AI that is both powerful but also supplicant (agrees with you irrespective of truth) can have serious unintended negative side effects

#37 | And the winner is ...

#36 | What makes a multi-agent framework good or useful?

#35 | A walkthrough on the newly released Google ADK

# 34 | How to enable "voice mode" with your AutoGen Agents!

#33 | With all the frenzied articles on MCPs, I think it is important to reflect on the current limitations - developer experience, application deployment and security.

#32 | What is the Model Context Protocol (MCP) from Anthropic and how can you use it with AutoGen?

#31 | I tested the recently released Phi-4 multimodal model - it's impressive!

#30 | AutoGen Studio has been completely rewritten on the new AutoGen v0.4 API, with a few new features

#29 | An overview of the AutoGen v0.4 API, and how to build your first multi-agent app.

Issue # 28 | How to build a multi-agent system that addresses tasks using a browser (like the OpenAI Operator agent)

When Progress Becomes a Sprint: The Human Side of AI's Rapid Advancement

Listen now (14 mins) | What were the top 5 trends in AI Multi-Agent Systems from 2024? What can we expect in 2025?

Issue #25 | Instead - Here are 7 ways to get the most of Generative AI pair programming tools like Gemini, ChatGPT or Claude

Issue #24 | Review of the NV-Embed paper where a generative model (Mistral 7B base) shows state of the art performance as a generalist text embedding model.

Issue #23 | How will software engineering change in the age of strong AI models that can write code and what should individual engineers and teams do to adapt?

Issue #22 | I am writing a book on multi-agent systems. You can pre-order it today!

Issue #21 | An in-depth analysis and guide on emerging tools/practices that apply large language models (LLMs) like ChatGPT/GPT4 to automate data visualization

Issue #20 | LLM-Enabled agents that act by driving interfaces (e.g., fill in and submit a form, conduct search). Current approaches, challenges and use cases.

Issue #19 | How to build your own web application that integrates Generative AI Agents

Issue #18 | Build Multi-Agent Apps with AutoGen, Write Zero Code

Issue # 17 | Understand key concepts with AutoGen and write your first multi-agent GenAI app!

Issue # 16 | Autonomous Agents are the next step in the evolution of LLM Apps. What is the state of the art, applications and current limitations?

Issue #15 | Cost reduction, new models, new developer APIs. What does this mean for developers, and the types of new applications that can now be built with ease?

Issue # 14 | Deep dive on how to access/call the PaLM API (MakerSuite, Vertex AI client libs, REST api), implementing a useful task (structured data extraction), developer notes.

Issue #13 | With multiple LLM providers (GPT4, PaLM) and 1000s of OSS Models, how do you select the right model?

Issue #12 | How to build tools for automatic data exploration, grammar-agnostic visualizations and infographics using Large Language Models like ChatGPT and GPT4.

Issue #11 | How can we make systems that integrate LLM's like ChatGPT more reliable? Here are practical techniques (and research) to mitigate hallucination and improve overall performance.

The ability to generate high quality novel images has been the target of extensive research in the deep learning and computer vision field. The promise of simply describing an image using natural language, and a tool generates high quality images that are faithful to that description is .. mind blowing. Importantly, this could supercharge creativity everyone!! Think futuristic photoshop.Well .. that technology is already here!In this newsletter, I'll walk through some examples of how a recent class of machine learning models (diffusion models) can be used in generating compelling images.Hint: the images below were generated by a model purely based on text.

In this issue we will discuss how to run machine learning models locally on an Android mobile devices. We will explore :Why on-device machine learning is interestingBuilding Android apps with Jetpack ComposeHow to build and train a simple neural network using TensorflowHow to convert the model to the Tensorflow lite format which is optimized to run on resource constrained devices (microcontrollers, microprocessors)How to integrate the Tensorflow lite model in your Android app and make predictions!Let's dive in!Note: This newsletter does not include code snippets for the tasks above. Please see the full post and links to sample code here.

Hi all, hope you are well!In this issue I'll talk about some of my recent experiments building a simple text classification model (predict if a tweet is related to politics or not) and debugging/inspecting the model's behaviour using gradients.

Hi everyone! Hope you are all doing well! In this newsletter issue, I'll discuss some of my recent experiments with attempting to render beautiful, styled (custom themed) Jupyter notebooks in web pages. For context, I have written a few pedagogical notebooks and I wanted to explore ways to organize/integrate them into a web application where they can be browsed easily and then launched in an environment like Colab or Kaggle if needed. My requirements for this project included :Support a workflow where I can update/maintain the notebook directly and those changes are automatically synced with my web application;Support custom styling of the notebook to fit the web application design language and support custom behaviours (e.g., custom buttons, click events).This newsletter summarizes my process in addressing these requirements.

Consider the following scenario - You write a comment "create a function that fetches all machine learning tweets and saves it to a file", and a deep learning model generates a full python function that achieves this goal (see screenshot below). It imports a suitable library, creates the required boilerplate authentication code and call structure such that you have decent functionality with minimal effort!The research area, frequently referred to as deep learning for program synthesis has recently become quite active, with useful applications in improving developer productivity (e.g., reducing time to implement features), improving the experience for developers exploring a new language or framework or even helping novice developers understand existing code.

What is the state of consumer VR devices and how easy is it to write VR apps today? I was curious about these questions and this newsletter issue summarizes some of my high level findings.

Across many business and practical use cases, we might want to automatically generate short, abbreviated versions of some long text. It turns out you can apply advances in modern (BERT based) neural networks in solving this task! In this issue, we will start with an overview extractive and abstractive summarization and how to implement extractive summarization with BERT.

Google IO is Google's annual developer conference where they announce new products, projects and developer tools. I attended several of the sessions and this newsletter contains a list of my top 10 insights that border on Machine Learning and HCI/Design.

Real-Time High Resolution Background Matting, Fine Grained Segmentation

This edition is a roundup on the state of object detection models and self supervised multimodal models!