Don't worry. Not everyone knows it...

Stop Guessing — Here's How to Pick the Right AI Model in Azure AI Foundry A quick guide for developers and AI builders who want a smarter, data-driven approach . Most people choose an AI model by reputation — "GPT-4o is popular, so I'll use that." But for AI agents that make real decisions, call external tools, and run multiple model calls per task, that guess can cost you in three ways: wasted money, sluggish responses, and broken workflows. Azure AI Foundry already has everything you need to choose smarter. You just have to know where to look. The Four Things That Actually Matter Before picking any model, evaluate it across four dimensions: Quality — benchmark scores across reasoning, coding, and Q&A tasks. Important, but don't over-index on general averages if your use case is specialized. Cost — not just per-token pricing, but total cost per completed agent task. An agent that makes 10 model calls per task spends 10× what a simple chatbot does. That math c...

Today’s post is a little different — and very close to my heart. My 10‑year‑old son, Advik, has been talking a lot about turning 11 soon. But instead of asking for gifts or planning a party, he surprised me with something completely unexpected. He said,  “I want to start earning my own pocket money. I want to work for it.” And honestly, that one sentence made me pause. At 10, I was definitely not thinking about earning or creating anything. But he has this spark — this excitement to build something on his own. So, he came up with an idea:  “What if I make activity books for kids?” And that’s exactly what he did. After days of choosing puzzles, testing pages, arguing about which ones were “fun enough,” and learning how books are made… he finally created his first one: 🕵️‍♂️ I Spy Secret Missions – 35‑Page Activity Book (Ages 5–7) It’s colorful, cute, and full of little challenges that kids love. And the best part? He is so proud of it. We’ve priced it at just $3 for the launch...

Have you ever spent hours trying to design a landing page, only to end up with something that looks... average? I've been there. That's exactly why I got excited when I discovered MotionSites.ai. What Is It? MotionSites is a library of premium AI prompts built specifically for web design. You pick a template, copy the prompt, paste it into an AI tool like Lovable or Claude, and within seconds you get a fully animated, professional-looking website section — no design skills required. What I Built I used one of their prompts to create an animated character carousel called TOONHUB — complete with smooth color transitions, depth effects, grain overlays, and responsive layouts. The whole thing came together on the first try. Watch me build it here: Why It Works The magic is in the prompts. Each one is packed with design decisions — fonts, animations, spacing, color palettes — that would normally take hours to figure out. You're not just saving time, you're borrowing expertis...

If you have spent any time building AI agents, you have probably run into this moment: you tell your agent something important, it responds perfectly, and then one message later it acts like the conversation never happened. Blank slate. No memory. Total amnesia. This is not a quirk. It is how large language models work by design. Every call to the model is stateless. Without explicit history being passed back each time, the model starts fresh on every turn. Context Providers are how the Microsoft Agent Framework solves this — and they do a lot more than just memory. What Is a Context Provider? A Context Provider is a class that wraps around every LLM call your agent makes. It gets two hooks: one before the call and one after. Before the call fires, the provider can inject anything into the conversation — previous messages, today's menu, user preferences, business rules, tool definitions. The LLM sees all of it before generating a response. After the call fires, the provider can ext...

Learn how to use Claude AI to audit your car insurance, find hidden discounts, and negotiate a lower premium — with 13 simple copy-paste prompts. Most people renew their car insurance every year without reading a single line of their policy. The number goes up slightly, they sigh, and they pay it. Insurance companies have built their entire business model around this habit. Here is what they do not tell you: they are legally required to justify every charge on your policy. Every coverage, every fee, every add-on. The only catch is that you have to ask. And now, with Claude AI, asking has never been easier. What Is Claude AI and Why Does It Work for Car Insurance? Claude is a free AI assistant at claude.ai. No technical skills needed. What makes it powerful for car insurance is simple — it reads complex documents, runs financial calculations, and generates professional scripts in plain language. Everything a good insurance broker would do, without the commission motive. You upload your ...

✨ “Wait… me? In the top 5? I had to blink twice. Honored!” Just spotted my name in a "Top 5 list for Claude Code projects" — and I genuinely had to blink twice. Moments like this remind me why I keep building, experimenting, breaking things, fixing things… and sharing every step of the journey. From custom MCP servers  to Claude specs  to local AI workflows —  it’s surreal to see the work resonate. Grateful. Energized. And absolutely fired up for what’s next ✨ Onwards — the star era continues 🌟 5 Fun Projects Using Claude Code - KDnuggets

AI tools are getting smarter every day — but one question keeps coming up: Can I run powerful AI models locally without sending my data to the cloud   The answer is finally yes. In this short guide, I break down how you can run Claude Code directly on your own machine, with zero cost, zero cloud dependency, and complete privacy .  Whether you’re a developer, a researcher, or someone exploring local LLMs for the first time, this setup gives you full control over your workflow. Running Claude locally means your code, prompts, and files stay exactly where they belong — on your device. No external servers. No data sharing. No API bills. Just fast, private AI coding assistance that works the way you want. You’ll learn how to install the tools, configure the environment, and start using Claude Code inside your favorite editor. It works beautifully with VS Code, Python projects, and local automation scripts. And the best part: the entire setup is free. If you’ve been curious abo...

Most AI systems can read your documents — but very few can understand them. Traditional RAG pulls text chunks and hopes the answer makes sense. It works for simple questions, but the moment you ask something relational —  Who works with whom? How are these teams connected? Why does this project matter? — it breaks. Graph RAG fixes that by turning your documents into a knowledge graph. People become nodes. Teams become nodes. Projects, locations, responsibilities — all nodes. And the relationships between them become edges. Suddenly, your data isn’t flat anymore. It’s connected. Graph RAG doesn’t just retrieve text. It retrieves meaning. It understands relationships, clusters related entities into communities, and uses LLM ‑generated summaries to explain what each group represents. So when you ask a complex question, the system doesn’t guess — it reasons. If you want to see this visually — with examples, and a full breakdown of how the pipeline works — I’ve created a detailed vide...

Are you trying to build an app with Claude AI but not getting the results you want? Your prompts might be the problem — not Claude . Introducing App Brief Wizard — a simple, offline HTML tool that helps you generate precise, structured Claude prompts for any app idea. Answer 5 quick questions → get a tailored prompt → instantly better UI and app concepts. No installs. No accounts. No learning curve. Just better prompts — instantly. 🎯 Perfect for:   Freelancers • Designers • Product Managers • Entrepreneurs • AI Builders • Claude Users 💡 Why it works:   App Brief Wizard helps Claude understand your app’s purpose, audience, and design personality — so your outputs look professional, not generic. 🔥 Early Bird Offer:   First few copies are just $5 (use code FIRST10) After that, it’s $9 — grab yours before it’s gone! 👉 Get App Brief Wizard here: App Brief

If you’re anything like me, you probably spend a big part of your day typing — emails, stories, books, blogs, notes, messages, ideas, reminders… it never ends. And honestly, it gets tiring. Our hands get tired. Our minds get tired. And sometimes, typing just slows us down. That’s when I realized something simple:   Talking is easier than typing. We speak faster, more naturally, and with less effort. But here’s the problem — most voice‑to‑text tools send your voice to the cloud.   Some store it.  Some analyze it.   Some require monthly subscriptions.   And many people don’t feel comfortable with that.  So, I decided to build something different. Meet Speakly — A Private, Offline Voice‑to‑Text App for Windows Speakly is a small, simple tool that lets you talk , and your computer types everything for you .   But the best part is what it doesn’t  do: It doesn’t send your voice to the internet   It doesn’t store your...

I’m excited to share my most awaited video, where I break down how to build an MCP server completely from scratch using FastAPI —with no SDKs, no decorators, and absolutely no hidden magic.  This tutorial walks through the real MCP protocol step‑by‑step. You’ll see how the manifest, tools list, execution flow, and even streaming responses come together using pure Python. It’s a simple, transparent way to understand MCP at its core. 🎥 Watch the full video: 

Over the past few weeks, I’ve been exploring how to build practical, privacy‑first agentic AI workflows that run entirely on a local machine. In my latest project, I combined GitHub Copilot SDK with Foundry Local to create a fully offline agent capable of choosing and executing tools intelligently — without relying on any cloud model. In this demo, I walk through how I built: - A Foundry Local LLM tool for on‑device inference - Three lightweight Python tools  - A router prompt that lets Copilot SDK decide which tool to invoke - A clean async loop that ties everything together The result is a flexible, extensible agent that can reason, select tools, and produce polished answers — all running locally. If you’re interested in agent design, local LLMs, or practical orchestration patterns, this walkthrough will give you a clear, end‑to‑end example you can adapt to your own projects. 🎥 Watch the full video here:

If you’ve ever wanted to run an LLM directly on your laptop without relying on the cloud, this new video is for you. I just released a hands‑on walkthrough of Foundry Local, where I show you exactly how to download an AI model to your machine and use it completely offline. In the video, I break down both methods developers use: - CLI workflow — install Foundry Local, pull a model, and run inference offline - Python SDK workflow — load the model in your code and build real offline AI features Whether you're a developer, an AI enthusiast, or someone who wants more privacy and zero token costs, this tutorial will help you get started in minutes.

Let’s be honest: most AI demos feel like magic tricks.  You type something, it replies. Cool!  But what happens when you want your AI to actually do something?  Like read a file, s ummarize a document, c all an API, r un a script, t rigger a workflow, etc. That’s where the GitHub Copilot SDK and MCP servers come in. They let you build real AI agents — ones that can reason, call tools, and interact with your environment like a tiny software teammate. In my latest video, I show you how to wire up local and remote MCP servers into a GitHub Copilot SDK agent. You’ll learn: - How MCP works  - How to build your own MCP server in Python - How to plug it into your agent - How to mix local and remote tools like a pro 👉 Watch the full walkthrough here If you’ve ever wanted to build an AI agent that feels like a real part of your stack — not just a chatbot — this is the video to watch. Let me know what you build after watching. I’m curious.

AI agents are quickly becoming a core part of modern development workflows, and the GitHub Copilot SDK makes it surprisingly straightforward to build your own. Instead of relying on prompt engineering alone, the SDK lets you define structured tools, give your agent explicit capabilities, and execute real code through LLM‑driven reasoning. In my latest demo, I walk through the full process of creating an agent from scratch — setting up the project, defining the agent, building custom tools, and running everything locally. You’ll see how the SDK handles tool invocation, schema validation, and natural‑language responses, all while keeping your logic deterministic and maintainable. If you're exploring agentic workflows or want to understand how Copilot can power real execution paths, this walkthrough will give you a clear, practical starting point. 🎥 Watch the full step‑by‑step video here: 👉  This is just the beginning — once you understand the pattern, you can extend your agent with...

The GitHub Copilot SDK just dropped, and it’s a game-changer for developers. You can now build your own Copilot-style AI features directly inside your apps, tools, and workflows — no more waiting for GitHub to do it for you. In my latest video, I break down exactly what the SDK is, how it works, and why it’s the future of developer productivity.  🎥 Watch now: Introducing GitHub Copilot SDK — Step-by-Step Demo If you’re serious about AI + dev tools, this is the video to start with.

Every once in a while, a new AI tool appears that doesn’t just improve on what already exists — it completely changes your expectations. That’s exactly what happened when I tested Awaz AI , a voice agent designed to handle real phone calls with natural, human‑like conversation. I’ve tried many voice systems before, and most of them sound robotic, interrupt at the wrong time, or fall apart when you ask something unexpected. But Awaz AI surprised me from the very first “hello.” The pacing, the tone, the timing — everything felt unusually natural. It didn’t rush. It didn’t freeze. It didn’t sound scripted. It actually felt like a real conversation. To make sure I wasn’t imagining it, I recorded the entire interaction. No edits. No retakes. Just a raw, real phone call between me and the Awaz AI agent. If you’re curious about how far voice AI has come — or if you simply want to hear an AI that sounds more human than most customer service lines — you should watch this demo. It’s one of the ...

AI workflows were a great starting point. They helped us build early prototypes, automate simple tasks, and experiment with LLMs. But the future of AI is not workflows - it's agents. Agents are more flexible, more intelligent, and more aligned with how real‑world tasks work. If you want to understand this shift - and learn exactly how to convert your existing workflows into agents - my video will walk you through the entire process step by step. Watch it. Learn it. Build with it. Your future AI systems will thank you.

If you’ve been curious about how to take your AI agent workflows from development to deployment, this new tutorial is for you. In my latest video, I walk you through the process of building declarative & hosted agents inside Visual Studio Code and then show you how to publish them directly to Microsoft Foundry (Preview). 🎯 What you’ll learn in the video: - How to set up declarative agents in VS Code - How to set up hosted agents in VS Code - Hosting workflows for scalable deployment - Publishing agents seamlessly to Microsoft Foundry - Why Foundry is becoming the go-to platform for enterprise-ready AI agents Whether you’re a developer experimenting with agent-based systems or an AI enthusiast looking to understand Microsoft’s latest tools, this tutorial will give you a clear, step-by-step guide to get started. 👉 Watch the full video here 📌 Don’t forget to like, comment, and subscribe for more tutorials on building intelligent agents with Microsoft Foundry!

If you’ve been exploring Microsoft Foundry and wondering how to actually run those declarative agent workflows you keep hearing about… this video is for you. I break down exactly how to view, run, and test declarative workflows inside VS Code — no fluff, just practical steps. You’ll learn how to open workflow files, understand agent logic, and test everything with real-world examples. Whether you're building multi-agent systems or just curious about how Foundry works, this tutorial will help you get hands-on fast. 🎯 What’s inside the video: - How to open and explore workflows in VS Code - Running workflows with Microsoft Foundry - Testing agent logic and human-in-the-loop steps - Tips for debugging and refining your setup 👉 Watch now: Happy learning!