Learn Nightmare, Openai and Zag

OpenAI's technologies offer a range of AI-powered tools and capabilities, making it easier for developers to build intelligent, responsive, and user-friendly applications.

openai

We can teach an AI model to better understand the information we want to generate responses from. To do that, we can use training data in a particular format, JSONL.

Once we have our data generated and uploaded to OpenAI, we can now use it to fine tune or customize a model to provide better results.

We'll use the training data we generated and uploaded in previous lessons to create a new fine-tuned model based off of the OpenAI davinci model.


- [https://platform.openai.com/docs/api-reference/fine-tunes/create](https://platform.openai.com/docs/api-reference/fine-tunes/create)

Fine Tuning OpenAI Models with Training Data Files

Once we have our training data, we need to a way to deliver it to OpenAI.

We can do that by uploading our training data file to OpenAI directly where we can then reference that with our fine tuning.


- [https://platform.openai.com/docs/api-reference/files/upload](https://platform.openai.com/docs/api-reference/files/upload)


Uploading Training Data Files to OpenAI for Fine Tuning AI Models

Using the models that OpenAI provides will get you really far, but their knowledge base can be somewhat limited for particular information or if its past a certain date that the model doesn't support.

To improve the experience, we can train models to have a better understanding of the data we want to work with, whether that's documentation for a new developer tool or guiding the model to respond closer to how how we expect.

The problem—is you need a lot of data in a specific format to do this and its not very practical to write it all our by hand. So we can instead use GPT, where by providing our data, we can ask it to generate JSONL training data for us and format it in a way that we can use to fine tune our model.


- [https://platform.openai.com/docs/guides/fine-tuning/prepare-training-data](https://platform.openai.com/docs/guides/fine-tuning/prepare-training-data)

Generating & Preparing JSONL Training Data with GPT for Fine Tuning an OpenAI Model

Node.js is a platform built on Chrome's JavaScript runtime for easily building fast, scalable network applications. Node.js uses an event-driven, non-blocking I/O model that makes it lightweight and efficient, perfect for data-intensive real-time applications that run across distributed devices.

node

With a reliable data API, we can start using our generative output in creative ways, including feeding that directly into a new generation like creating an image based off of a blog post.

We'll look at taking a GPT-generated blog post and feeding it into DALL-E along with some stylistic prompt engineering to generate an image for our post.

Use Chat Completions for Image Generation Prompts with DALL-E

An important part in creating a data API whether based on generative systems or not is being able to reliably generate responses in the format you need. Variable data structures and irrelevant information could break your application or frustrate the users of your API, resulting in downtime or lost customers.

Using the messaging API, we can train GPT to become more aware of our needs by using System roles to instruct GPT on behavior and User to Assistant examples to give GPT precedent to base its next response on.

[https://platform.openai.com/docs/guides/chat/introduction](https://platform.openai.com/docs/guides/chat/introduction)

Training GPT Using Roles & Message Examples for Better Chat Completions

Getting a long string as a response makes a lot of sense if you're building an app or tool that appears conversational or if you just want to display a little bit of information as a response, but if you want to do something programmatic with the information, such as using different datapoints from the information in different parts of the UI, you'll have a hard time parsing out that information from a single request.

To get around this, we can simply ask GPT to return our data formatted as JSON! Further, we can ask it in a particular shape. You'll learn how to format a chat completion prompt to request JSON data and gain confidence in that data by defining the shape in the request.

Requesting JSON Chat Completions from OpenAI for Programmatic Use


Requests that we typically make in an application come all at the same time. We make the request, we get a response. But using a streaming mechanism, we can get our information as it comes resulting in a quicker initial response, as well as set ourselves up for different user experiences including conversational-like messaging.

Using Edge Functions in Next.js, we can stream our response back to the application and create an experience similar to the official ChatGPT interface. You'll learn how to set up a new Edge function, make a request directly to the OpenAI API, and listen for the stream events in Next.js with [eventsource-parser](https://www.npmjs.com/package/eventsource-parser).

- [https://vercel.com/docs/concepts/functions/edge-functions/edge-functions-api](https://vercel.com/docs/concepts/functions/edge-functions/edge-functions-api)

Stream OpenAI Chat Completions to a Next.js App Using Vercel Edge Functions

Get Started with AI-Driven App Development Using the OpenAI Node.js SDK

ChatGPT and similar conversational AI tools have quickly made impressions in the tech industry and beyond by giving the people using those tools a more natural and a potentially more helpful means to get answers to the questions they have. Bringing GPT into applications can unlock new ways to solve problems and quickly help customers to solve theirs.

We can create ChatGPT-like conversational messages with the OpenAI SDK in a Node.js project. You'll learn how to use the [createChatCompletion](https://platform.openai.com/docs/api-reference/chat/create) method and configure the gpt-3.5-turbo model to generate a response based on a message prompt.

[https://platform.openai.com/docs/api-reference/chat/create](https://platform.openai.com/docs/api-reference/chat/create)


Generate Chat Completions with AI in Node.js Using the OpenAI SDK & GPT

The more descriptive an image prompt, the better the results will be. That includes being able to take user input to help articulate what the image should look like.

Dynamic prompts in Next.js serverless functions can help us generate images based off of application input with the OpenAI SDK. You'll learn how to configure an API Endpoint to parse a body of a POST request to grab a dynamic prompt value and use it to configure the prompt for the createImage method.

[https://nextjs.org/docs/api-routes/introduction](https://nextjs.org/docs/api-routes/introduction)

Create Dynamic Image Generation Prompts for the OpenAI SDK in Next.js API Endpoints

Image generation provides us a way to bring life to natural language through a visual medium. Through a simple or detailed description, we can generate a completely unique image that tells the story we want whether that's for dynamically creating blog images or exploring new forms of art.

We can use AI to generate images with DALL-E and the OpenAI SDK in Node.js. In this lesson you'll learn how to use the [createImage](https://platform.openai.com/docs/api-reference/images/create) method generate an image using a descriptive prompt, the number of images you want to generate, and the size of the image. 

[https://platform.openai.com/docs/api-reference/images/create](https://platform.openai.com/docs/api-reference/images/create)


Generate Images with AI in Node.js Using the OpenAI SDK & DALL-E

OpenAI is an AI research lab that has developed a number of tools for researchers and developers including natural language processing, robotics, and reinforcement learning. By leveraging the OpenAI APIs, we can build faster and smarter, while bringing new and interesting capabilities to help enhance existing solutions and solve new challenges.

Let's get started building with AI using the [OpenAI SDK](https://platform.openai.com/) by installing and configuring the SDK in a Node.js project. Here you'll learn how to install the [openai npm package](https://www.npmjs.com/package/openai) into an existing project, import the openai dependency, and configure the SDK with your OpenAI API Key to make authenticated requests to the OpenAI API.

[https://platform.openai.com/overview](https://platform.openai.com/overview)
https://platform.openai.com/overview

Install and Configure the OpenAI SDK in a Node.js Project

Learn how to build a pin-input component with state machines, Zag.js and React.

We’ll look at modeling a UI component with state machines, implementing those in a framework agnostic way using Zag.js, and connect the logic to the UI component.

This course is a must-watch if you’re unfamiliar with State machines, and want to learn how to use them to build UI components with ease.

In this course, you will:

- Learn fundamental concept of state machine, event and transition
- Model a UI component with state machine
- Connect state machine logic to UI
- Manage DOM events in state machines
- Create custom React hooks with state machines

A collection of framework-agnostic UI component patterns like accordion, menu, and dialog that can be used to build design systems for React, Vue and Solid.js

Statechart Driven UI Components with Zag.js

React is one of the web’s most popular frameworks for building JavaScript applications. 

If you know what you’re doing, React can drastically simplify how you build, use, and maintain code.

Whether you’re a React newbie or you’re ready for advanced techniques, you can level-up with egghead. Explore our in-depth courses, lessons, and community resources to build more powerful applications and crack open your career.

react

In many projects, it is useful to separate the logic that you right with the view that you create so that you can easily share that logic throughout your application. In React we do this with custom React hooks.

We'll improve on the reusability of the pin input component by refactoring it to a custom React hook. To do this we will utilize the prop getter pattern 

This Headless Component pattern is a very useful pattern when building reusable components. It allows you separate functionality from presentation, making it easy to style them as needed.


Refactor Zag.js State Machine Powered Pin Component Functionality to a Custom React Hook

In some cases you'll want to actually have your pin input in a form to handle submitting the values. What's awesome about our implementation is that it doesn't get in the way of adding functionality like this.

First we'll extend our state machine to focus the first input when we click on the input label. Then we'll wrap our Pin Input component in a form element and implement an onSubmit. We can submit values of the form by creating a hidden input and set it's values to the joined values of the other inputs.


Submit Pin Input Values in a Form with a Hidden Input

When the pin input values are full we will want to do something with those values such as sending an HTTP request.

To do this we'll implement a callback in context that will get invoked when all the values are computed. We'll do this by creating a computed property that returns true when all values in context are full. From there we will watch that computed property to invoke our callback.

Trigger Side Effects using Zag.js computed and watch Properties

We have a bug in our pin input. Only 1 value should be allowed per input but if we focus an input that already has a value we can still input another value. 

Let's fix that.

We'll need to implement a function `getNextValue` that will take in the focused value of the current input and the event input that is coming in. After a set of conditionals we will select the value that wasn't the current focused value to update the field.


Implement Conditional Logic in a Zag.js Action to Enforce a Single Character Per Input

The paste event is a unique input event that we need to handle. When we paste a value in from the clipboard, we need to split that value and input single characters to the individual inputs in our Pin Input.

To do this we will need to `preventDefault` in the `onPaste` event so that we can implement our custom logic here. Next we'll implement `setPastedValue` that splits and iterates over the input value we receive and sets it into context. 

Finally we'll implement `focusLastEmptyInput` that looks up the index of the first empty value with `findIndex` and focuses that index.


Spread Values Across a Pin Input Field in a Zag.js State Machine 

We'll be implementing the backspace event which triggers two actions: `clearFocusedValue`, and `focusPreviousInput`. 

These implementations you'll be familiar with at this point but what you'll notice when we actually try to hit backspace that the focus value will get cleared but we won't be setting focus to the previous input. This is because there are several DOM mutations happening at the same time. We'll fix this by scheduling the focus of inputs to the next animation frame with `requestAnimationFrame`


Schedule Input Focus with requestAnimationFrame to Handle Multiple DOM Mutations in Zag.js

Let's actually implement inputting values into our Pin input. We have 3 functions that need implemented. They are `setFocusedValue`, `focusNextInput`, and `executeFocus`. The first two will be pretty straight forward in grabbing values from context and our events and setting new values to our context. 

When we finish implementing `setFocusedValue` and `setFocusedInput` we'll need a way to actually trigger our focus to be set in the DOM. This is where the watch property in a Zag.js State Machine comes into play. We will watch the `focusedIndex` context value and run `executeFocus` any time the focus index changes in our machine (e.g. when `focusNextInput` is triggered after an input event).


Trigger Side Effects in a Zag.js State Machine with the watch Property

Now that we have a machine and our UI is sending it events, we need to actually do something with events that are sent. This is where actions come into play.

We'll implement focus and blur events for the input by setting the context `focusedIndex` to the index that is coming in from the UI. We can mutate context directly in Zag.js because it handles mutations through proxies to stay reactive.


Implement a State Machine Action for Setting Input Focus in Zag.js

Hey Everyone 👋,

John Lindquist here with the third issue of AI Dev Essentials! Another week, another whirlwind of updates in the AI development space. We saw Google drop Gemini 2.5 Flash, OpenAI counter with o3/o4-mini and a new Codex CLI, Anthropic beef up Claude with research tools, and continued evolution in the MCP and agent framework landscape. Personally, I've been digging into Gemini 2.5 Flash – it's impressive to see Google pushing the performance frontier for smaller models and comparing it directly against the competition like O4-mini. Let's dive in!

___

### New egghead.io Lesson This Week:

* **[Generate and Debug MCP Servers from Scratch using GitHub Examples and AI](https://egghead.io/generate-and-debug-mcp-servers-from-scratch~k98ad):** Building complex components like Model Context Protocol (MCP) servers from scratch can be challenging, especially without good reference implementations. This lesson demonstrates a powerful workflow combining the `ghx` CLI tool (a wrapper I built for the GitHub CLI - find it [here](https://www.npmjs.com/package/@johnlindquist/ghx)) with AI assistants (like Google's AI Studio and Cursor's agent) to overcome this hurdle. Learn how to scrape real-world examples from GitHub using `ghx` ([see scraped examples here](https://gist.github.com/johnlindquist/b19673bbf6a396406dff869beecace1d)), feed them to an AI to generate a new server, and then integrate and debug it within Cursor. It's a practical look at bootstrapping complex tools using AI and targeted code searching.

___

Here are the highlights that caught my eye this past week:

## Model & Major Platform Updates

* **Gemini 2.5 Flash Arrives:** Google released Gemini 2.5 Flash ([Google Developers Blog](https://developers.googleblog.com/en/start-building-with-gemini-25-flash/)), their first unified reasoning model with thinking budgets. It's positioned as a high-performance, cost-effective option. Notably, Google's benchmarks included comparisons against OpenAI's very recent O4-mini, drawing praise for transparency ([Nathan Lambert](https://x.com/natolambert/status/1912958672881303943)). It's available via the Gemini API, AI Studio, Vertex, and the Gemini app. The [Live API docs](https://ai.google.dev/gemini-api/docs/live) are also updated.
* **OpenAI Counters with o3/o4-mini & Codex CLI:** OpenAI launched [o3 and o4-mini](https://openai.com/index/introducing-o3-and-o4-mini/), highlighting their agentic capabilities and tool use during Chain-of-Thought. Alongside these, they released the open-source [Codex CLI](https://github.com/openai/codex), a coding agent designed to leverage these new models. Interestingly, the OpenAI dev behind Codex CLI was spotted [using Cursor](https://x.com/chuhang1122/status/1912786904812294312).
* **Claude Gets Research Superpowers:** Anthropic rolled out a major [Claude upgrade](https://x.com/jowettbrendan/status/1912446262756151732) introducing a "Research" agent capability. It can search the web and internal docs, chain queries, and explore topics agentically, building citations along the way. They also added Google Workspace integration (Gmail, Calendar, Docs) for paid users. (Official blog post link pending confirmation).
* **Cursor 0.49 Update:** The latest Cursor update ([Changelog 0.49](https://www.cursor.com/changelog/0-49)) brings several enhancements: improved Cursor Rules generation and attachment, better chat history access, built-in diff reviews, image support in MCP context, agent command editing/skipping, global ignore files, added models (Gemini 2.5 Pro, Grok 3, GPT-4.1), and project structure context for the agent. [AGIdevice](https://x.com/agidevice/status/1913575514670776520) shows how to enable the beta features.

## Tooling & Protocol Developments

* **MCP Nuances:** David Cramer cautioned against simply [proxying OpenAPI specs via MCP](https://x.com/zeeg/status/1913289979368796671), arguing it misuses the protocol and ignores LLM limitations in tool selection and JSON handling.
* **MCP Transport Clarification:** Harald Kirschner clarified the newer "Streamable HTTP" transport ([MCP Spec](https://modelcontextprotocol.io/specification/2025-03-26/basic/transports)), noting it's primarily HTTP-based, with Server-Sent Events (SSE) as an optional upgrade for streaming server messages, deprecating the older standalone SSE transport. Discussions on implementation and future streaming enhancements are ongoing ([TS SDK Issue #220](https://github.com/modelcontextprotocol/typescript-sdk/issues/220), [MCP Roadmap](https://modelcontextprotocol.io/development/roadmap)).
* **Agent2Agent (A2A) Protocol Announced:** Google unveiled the Agent2Agent (A2A) protocol ([Google Developers Blog](https://developers.googleblog.com/en/a2a-a-new-era-of-agent-interoperability/)), an open standard aiming to allow AI agents from different frameworks and vendors to communicate and collaborate. This is a significant step towards agent interoperability. Akshay Pachaar also provided a clear explanation contrasting [MCP vs. A2A](https://x.com/akshay_pachaar/status/1912123161883345321). ([GitHub Repo](https://github.com/google/A2A))
* **New MCP Tools:** Ishan Goswami released an open-source [Twitter MCP Tool using Exa](https://x.com/TheIshanGoswami/status/1914226695118168331) (GitHub: [exa-labs/exa-mcp-server](https://github.com/exa-labs/exa-mcp-server)) for searching tweets/profiles without the official API. Eyal Toledano shipped [TaskmasterAI v0.12](https://x.com/EyalToledano/status/1914084622918455572) with multi-PRD parsing, smarter breakdowns, and easier MCP setup, alongside a new [website](http://task-master.dev).
* **Cursor Workflows & Tips:** Ryo Lu shared [12 tips for using Cursor effectively](https://x.com/ryolu_/status/1914384195138511142), emphasizing clear rules, specific prompts, and iterative testing. Prajwal Tomar detailed his ["Context Boundary Method"](https://x.com/PrajwalTomar_/status/1913961327711633601) using structured docs (PRD, Flow, Tech Stack etc.) to guide AI coding tools like Cursor and reduce errors. Andrew Jefferson noted potential issues with [Markdown tab completion](https://x.com/EastlondonDev/status/1913967886521053688) in Cursor.
* **AI UI Builders Compared:** Damien shared a comparison of [Replit, Bolt.new, Lovable.dev, and v0](https://x.com/damienghader/status/1913248328302608482) for generating a multi-step onboarding flow from the same prompt.

## Building AI Agents: Frameworks & Best Practices

* **Agent Framework Discussion:** Harrison Chase (LangChain) published a deep dive on [how to think about agent frameworks](https://blog.langchain.dev/how-to-think-about-agent-frameworks/), sparking discussion comparing approaches (agents vs. workflows, declarative vs. non-declarative) and frameworks like LangGraph, OpenAI's Agents SDK, CrewAI, and AutoGen. Key themes include balancing flexibility vs. predictability and choosing the right abstraction level. OpenAI also released its own [practical guide to building agents](https://cdn.openai.com/business-guides-and-resources/a-practical-guide-to-building-agents.pdf).
* **Production Agent Principles:** Dex Horthy shared principles for production-grade LLM apps in ["12-factor-Agents"](https://github.com/humanlayer/12-factor-agents), arguing successful systems often use focused LLM loops within well-engineered software rather than fully autonomous, monolithic agents. This resonated with many builders facing reliability challenges.
* **Agent Memory:** Taranjeet from [mem0.ai](https://x.com/taranjeetio/status/1912185506559053855) emphasized the importance of persistent memory (beyond session history) for truly intelligent agents that learn and adapt over time ([Blog Post](https://mem0.ai/blog/memory-in-agents-what-why-and-how/)).
* **Key Best Practices:** Common advice emerging from various discussions includes: start simple (maybe not even an agent), use structured outputs (like JSON schemas) to reduce hallucinations, prefer multiple specialized agents over one complex agent, own your control flow rather than relying solely on the framework, and carefully manage context windows. ([Anthropic Guide](https://www.anthropic.com/research/building-effective-agents), [Dev.to Summary](https://dev.to/isaachagoel/read-this-before-building-ai-agents-lessons-from-the-trenches-333i)).

___

**✨ Workshop Spotlight: Conquer the Complexity of Cursor ✨**

Master practical AI development workflows in Cursor. This hands-on workshop covers Agents, Ask, Custom Modes, multi-file analysis, effective prompting, Cursor rules, and strategies for handling AI failures.

* **When:** **This Thursday!** April 24, 2025, **9:00 AM - 2:00 PM (PDT)**
* **Where:** Zoom (Live Q&A included)
* **Investment:** $249

Unlock Cursor's potential and streamline your development process.

➡️ [**Read More about the Workshop**](https://egghead.io/workshop/cursor) ➡️ [**Register Now ($249)**](https://buy.stripe.com/9AQaG5ftF5P22eAfZ3)

_(Team training also available)_

___

## Notable Prompts

* **o3 Fact-Checking:** A prompt structure for using o3 to thoroughly fact-check an entire article by parsing individual facts. (See [Matt Shumer's Tweet](https://x.com/mattshumer_/status/1914412812891570199) for the prompt start).
* **Multiple Experts Method:** A system prompt that instructs the AI to analyze problems from multiple expert perspectives (e.g., Systems Thinker, Risk Assessor) and synthesize their insights. (See [God of Prompt's Tweet](https://x.com/godofprompt/status/1914370848032497906) for the full prompt and example).
* **Personality-Infused Prompt:** A detailed persona prompt instructing the AI to adopt a specific, informal, and slightly critical tone using modern slang and abbreviations. (See [Kitze's Tweet](https://x.com/thekitze/status/1913719205007937825) for the full prompt, originally from [@threepointone](https://x.com/threepointone)).

## Community Buzz & Resources

* **New Community Site:** [AGIdevice](https://x.com/agidevice/status/1913666033870647325) is building [cursorcommunity.com](http://cursorcommunity.com) in public.
* **Industry Chatter:** Deirdre Bosa discussed the potential [OpenAI bid for Windsurf](https://x.com/dee_bosa/status/1912906057791447294) and the shift towards the app layer. AI Native Dev reported on perceived [friction between Microsoft and Cursor](https://x.com/ainativedev/status/1912869556973949430) ([Article](http://ainativedev.co/spx)).
* **Android XR Demo:** Google showed off prototype smart glasses with Android XR ([Demo Video via 9to5Google](https://9to5google.com/2025/04/17/google-android-xr-smart-glasses-live-demo/)).
* **New Open Dialogue Model:** A new small open model, [Dia-1.6B](https://x.com/altryne/status/1914421817139454341), was released. It's focused on dialogue disentanglement – figuring out who's talking to whom in complex conversations. ([GitHub](https://github.com/nari-labs/dia), [Hugging Face](https://huggingface.co/nari-labs/Dia-1.6B)). I was extremely impressed with the demos – definitely worth checking out.
* **Video Generation with Veo 2:** Google's Veo 2 video generation model is now available in Gemini Advanced and Google AI Studio ([Google Blog](https://blog.google/products/gemini/video-generation/), [Developers Blog](https://developers.googleblog.com/en/veo-2-video-generation-now-generally-available/)). I've tried out a few videos and was really impressed with the quality. Pro tip: In AI Studio, you can use an image as a reference when generating video ([API Docs](https://ai.google.dev/gemini-api/docs/video#generate_from_images)), though surprisingly, Gemini doesn't support that yet. It's heavily rate-limited at the moment, but still fun to play with for generating short 8-second clips.

___

That’s the scoop for this issue! The pace isn't slowing down, particularly around agent capabilities, specialized models, and the tooling ecosystem supporting them.

If you have any feedback or questions, hit reply! Always happy to chat about the latest in AI dev tools.

John Lindquist


John Lindquist

AI Dev Essentials #3: Gemini Flash, o4-mini, Codex CLI, A2A Protocol & Agent Updates

AI Dev Essentials covers Gemini 2.5, O4-mini, Claude’s research tools, Codex CLI, and new workflows for building AI agents with MCP and Cursor.

Establish an automated, AI-driven workflow for implementing UI designs by comparing live development versions against target screenshots. This lesson integrates the Playwright MCP server for browser automation, a custom script using Pixelmatch for visual diffing, and Cursor Rules for persistent context. See how to instruct the Cursor Agent to iteratively implement design changes based on visual diff feedback, creating a robust loop where the AI refines the UI towards the target design.

**Workflow demonstrated in this lesson:**

*   Configure the Playwright MCP server within Cursor for browser control.
*   Use the Agent and Playwright MCP tools (`browser_navigate`, `browser_resize`, `browser_take_screenshot`) to capture baseline and current state PNG screenshots.
*   Install `pixelmatch` and `pngjs` dependencies.
*   Guide the Agent to create and run a TypeScript script (`compare-images.ts`) to generate visual diff images between screenshots.
*   Establish Cursor Rules (`visual-testing.mdc`, `project-setup.mdc`, etc.) to provide the Agent with necessary context about directories, scripts, dependencies, and the overall workflow.
*   Initiate an AI-driven design implementation loop:
    *   Agent captures the current state via Playwright.
    *   Agent runs the comparison script to get a visual diff.
    *   Agent analyzes the diff and implements code changes in the project (e.g., `app/page.tsx`, components, `globals.css`).
    *   Repeat the capture-compare-implement cycle, using the latest diff as feedback.
*   Use context files and prompts like "Please continue" to manage the multi-step process and guide the AI through potential interruptions.

**Key benefits:**

*   Creates an iterative, AI-driven workflow for implementing designs based on visual comparisons.
*   Combines browser automation (Playwright) and pixel-level diffing (Pixelmatch) within the Cursor environment.
*   Utilizes Cursor Rules to effectively manage context and guide the AI through complex, multi-step tasks.
*   Demonstrates how visual diffs can serve as direct feedback for AI-driven UI code generation and refinement.

AI-Driven Design Workflow: Playwright MCP Screenshots, Visual Diffs, and Cursor Rules

[Scraped MCP Examples](https://gist.github.com/johnlindquist/b19673bbf6a396406dff869beecace1d)
[ghx tool](https://www.npmjs.com/package/@johnlindquist/ghx)

Building complex components like Model Context Protocol (MCP) servers from scratch can be challenging, especially without good reference implementations. This lesson demonstrates a powerful workflow combining the `ghx` CLI tool with AI assistants (like Google's AI Studio and Cursor's agent) to overcome this hurdle.

You'll learn how to:

1.  **Use `ghx`:** Leverage the `ghx` command-line tool (which wraps the GitHub CLI) to search GitHub for real-world code examples based on specific import paths (e.g., `modelcontextprotocol/sdk/server/mcp.js`).
2.  **Refine Search Results:** Use `ghx` options (`--help`, `--limit`, `--max-filename`, `NOT demo`) within Cursor's chat to filter results, exclude simple demos, and gather a rich set of relevant MCP server implementations.
3.  **Provide AI Context:** Copy the aggregated code examples found by `ghx` into an AI environment like Google AI Studio (or ChatGPT).
4.  **Handle Context Limits:** See techniques for trimming large context files (like using Cursor's `Cmd+K` inline edit with `head -n` commands) to fit within model limits while retaining valuable examples.
5.  **Generate Code:** Prompt the AI (Gemini 2.5 Pro in AI Studio or ChatGPT O3) to generate a *new* MCP server (e.g., with five distinct tools) based *solely* on the provided real-world examples.
6.  **Integrate and Debug:** Set up the generated TypeScript code (`mcp_server.ts`) locally in Cursor, install dependencies (`pnpm install`), and use Cursor's agent (`Cmd+I`) or inline fixes (`Cmd+K`) to resolve TypeScript errors and path issues.
7.  **Configure in Cursor:** Add the newly created MCP server (running via `bun mcp_server.ts`) to Cursor's MCP settings, using the absolute path and the `bun` command.
8.  **Test the MCP Server:** Interact with Cursor's chat agent to call the custom tools (e.g., `add_note`, `list_notes`, `write_managed_file`) defined in the AI-generated MCP server, verifying its functionality.

This lesson showcases how combining targeted code searching (`ghx`) with the generative power of AI, fueled by relevant examples, can dramatically accelerate the process of building and bootstrapping complex tools and applications.

Generate and Debug MCP Servers from Scratch

Hey Everyone 👋,

John Lindquist here with the second issue of AI Dev Essentials! It's been another busy week in the AI tooling world. The biggest splash was probably GPT-4.1 hitting the streets, but there's plenty more action around MCP, Vercel's v0, Google's developer tools, and interesting community discussions. I've spent some time with GPT-4.1 already (specifically in Cursor), and while it shows promise, it didn't quite convince me to switch from Gemini 2.5 Pro yet. It stumbled on a few basic requests where Gemini usually shines on the first attempt.

---

### New egghead.io Lesson This Week:

* **[Create a Custom MCP that Calls Cursor Tools](https://egghead.io/create-a-custom-mcp-that-calls-cursor-tools~r77sw):** Learn how to build custom MCP server tools that allow Cursor's AI Agent to access your local project context. This lesson shows how to use Zod schemas and the crucial `.describe()` method to explicitly request information (like Git status or version) from the Cursor environment *before* your tool runs, enabling powerful, context-aware automation.

---

Here are the highlights that caught my eye this past week.

## Model Mania: GPT-4.1 Arrives, Plus Grok & Llama Updates

* **GPT-4.1 Lands:** OpenAI announced GPT-4.1, and it quickly became available via the [API](https://openai.com/live/) and showed up in [Cursor](https://x.com/cursor_ai/status/1911835651810738406). Cursor is offering it free for now to let users get a feel for it, though they don't recommend it as a default just yet. Curious to hear initial impressions from folks testing it out for coding tasks.

* **Grok 3 in Cursor:** Following other model additions, [Cursor also enabled Grok 3 models](https://x.com/amanrsanger/status/1910434915419906141), with the mini version available for free users. Aman Sanger noted it seems to "think upfront more than other models."

* **Community Hack: Llama 4.0 via Groq in Cursor:** Levi Dan shared a [detailed (and slightly hacky) guide](https://x.com/lxvi_brg/status/1908865544721154455) on how to wire up Groq's fast Llama 4.0 API into Cursor by using the OpenAI API key slot and intercepting requests. A clever workaround for the adventurous!

## Tooling & Platform Pulse

Lots of movement across different platforms and tools this week:

* **Vercel Ecosystem:**

  * Guillermo Rauch teased [v0 coming to VS Code](https://x.com/rauchg/status/1911524154458984515), hinting at tighter integration for the UI generation tool.

  * v0 also [added AI integrations](https://x.com/v0/status/1911912183149256943) with models from xai (Grok), FAL, Groq, and DeepInfra, including automatic API key setup and pay-as-you-go billing via Vercel Marketplace. They're offering a $10 credit for the Grok free tier.

  * Vercel shared insights on the [grep.app migration from CRA to Next.js](http://grep.app/), showcasing performance improvements. ([Source](https://x.com/rauchg/status/1911920410020757881)) I'm personally excited about this one – I rely heavily on examples when working with AI, and for me, a good example is often worth ten prompts. Tools like `grep.app` that make code discovery easier are invaluable.

* **Google's Growing Stack:**

  * Project IDX has officially become [Firebase Studio](https://x.com/_davideast/status/1909990728593453455), positioning itself as a cloud-based, agentic dev environment for full-stack AI apps, deeply integrated with Firebase. I've tried spinning up a few sites with it, and the speed is impressive. However, for more complex tasks like designing from images or intricate interactions, I still find myself leaning towards v0 for now.

  * Google quietly released the [Agent Development Kit (ADK)](https://x.com/omarsar0/status/1910004370864742757), a Python library (`pip install google-adk`) for building multi-agent systems with features like state management, tool usage, and orchestration. I haven't had a chance to dive into this yet, but it's right at the top of my list. Building robust agent systems is a huge area of interest, so this is probably the announcement I'm most excited about exploring this week.

  * Ramón Medrano Llamas shared his experience [moving back to VS Code from Cursor](https://x.com/rmedranollamas/status/1909516749612999149), highlighting the effectiveness of GitHub Copilot combined with the Gemini Code Assist extension (using Gemini 2.5 Pro in Agent Mode). I gave the Gemini Code Assist extension a try myself, but as someone deeply embedded in the Cursor workflow with established keyboard shortcuts and patterns, I struggled to find a flow I liked and didn't stick with it for long. It's interesting to see different perspectives, though!

  * AI features are rolling out more broadly in [Chrome](https://x.com/iam_chonchol/status/1909942954686636224), including text grabbing from YouTube videos and Google Lens integration.

* **Cloudflare Corner:**

  * Big news for edge compute: [Cloudflare Containers are coming to Open Beta in June](https://x.com/CherryJimbo/status/1910682562499850405), enabling running containerized apps (Next.js, Go, PHP confirmed) directly on the edge. This is huge news, in my opinion. Think about all the WordPress sites and other applications that rely on containerization. Opening this up on the edge unlocks incredible possibilities, including for AI applications. I'm super hyped to see what people build with this come June.

  * A new tool, [MCPify.ai](http://mcpify.ai/), launched for building MCP servers conversationally, built entirely on Cloudflare Workers and Durable Objects. ([Source](https://x.com/pbteja1998/status/1911723180445610149))

  * Securing MCP servers is getting easier, with examples shared using [WorkOS AuthKit](https://x.com/WorkOS/status/1910722639036874943) and [Auth0](https://x.com/dschenkelman/status/1910693265566728319).

* **IDE Updates:**

  * JetBrains shared the public [WebStorm roadmap for March-April](https://youtrack.jetbrains.com/issue/WEB-72310). ([Source](https://x.com/WebStormIDE/status/1909291367504748937))

> **✨ Workshop Spotlight: Conquer the Complexity of Cursor (Europe Friendly Time!) ✨**
>
> Missed the last workshop or found the timing tricky? This one's for you! Master practical AI development workflows in Cursor in a session scheduled for **12:00 PM UTC** to better accommodate European attendees.
>
> This hands-on workshop covers Agents, Ask, Custom Modes, multi-file analysis, effective prompting, Cursor rules, and strategies for handling AI failures. Unlock Cursor's potential and streamline your development process.
>
> * **When:** \[Check website for upcoming dates - This session is Europe-friendly at 12:00 PM UTC\]
>
> * **Where:** Zoom (Live Q&A included)
>
> * **Investment:** $149
>
> ➡️ [**Read More about the Workshop**](https://egghead.io/workshop/cursor)
> ➡️ [**Register Now ($149)**](https://buy.stripe.com/7sI7tT81dfpC06s28c)
>
> *(Team training also available)*

## Community Buzz & Resources

* **MCP Explained:** Kent C. Dodds wrote a piece explaining [why he's excited about Model Context Protocol (MCP)](https://www.epicai.pro/the-game-changing-potential-of-model-context-protocol) and its potential. ([Source](https://x.com/kentcdodds/status/1911840456432034112))

* **Prompting Guide:** Ian Nuttall surfaced [v7 of Google's Prompt Engineering Guide](https://x.com/iannuttall/status/1911086164196606314), a deep dive into techniques and best practices. Definitely worth a look.

* **Discussion Starter:** Eric Zakariasson asked the community about their [favorite CLIs and MCP servers to use with the Cursor agent](https://x.com/ericzakariasson/status/1910808935961088346). What are you all using?

* **Gemini 2.5 Pro Praise:** The positive buzz around Gemini 2.5 Pro continues to grow in the community. It's getting rave reviews and even achieving 'S-tier' status on social platforms like Reddit for its coding capabilities. ([Source](https://x.com/suchenzang/status/1911251230317490416))

That’s a wrap for this week! There's always so much happening, it's hard to keep up, but hopefully, this helps surface some of the more interesting bits for developers working with AI.

If you have any feedback or questions, feel free to respond to this email. Always happy to talk shop about the latest AI tools.

John Lindquist

[egghead.io](https://egghead.io)

AI Dev Essentials #2: GPT-4.1 Arrives, MCP Deep Dive & Google's Agent Dev Kit

AI Dev Essentials #2: GPT-4.1 & Gemini updates, MCP tools, Vercel v0, Google ADK, Cloudflare Containers news. Latest AI dev tooling highlights.

In this quick walkthrough, you’ll learn how to drop Nx into an existing npm or pnpm workspace and immediately benefit from project graph visualization, caching, and smart task orchestration, all without rewriting your setup.

You’ll start with a standard React monorepo containing a Vite app, a UI package, and utility functions. Then, using a single command (`nx init`), you’ll initialize Nx and see how it auto-detects your existing scripts, builds a dependency graph, and enables advanced features like affected commands and local caching.

We're also going to explore optional enhancements using Nx plugins for Vite and React. You’ll see how they help Nx infer tasks directly from your Vite config files, improve DX with the Nx Console VS Code extension, and leverage Nx plugin generators to scaffold code with a consistent structure.

What you’ll learn
- How to initialize Nx in an existing npm or pnpm monorepo
- How Nx understands your workspace structure and scripts
- How local caching and affected builds speed up development
- How optional Nx plugins improve DX without adding bloat
- How to scaffold new packages

Using Nx on an existing React and NPM workspace based monorepo

Applying Nx to an existing React monorepo to enable caching, task orchestration, and project graph analysis without changing your current setup.

Typing repetitive or complex instructions for Cursor's AI agent can slow down your workflow. This lesson teaches you how to create custom slash commands (e.g., `/s term`, `/l`, `/c`) using Cursor's "Rules" feature. Learn how to map these simple shortcuts to specific AI actions, instructing the agent to use precise tools like `codebase_search`, `edit_file`, or `run_terminal_cmd` based on your command and the current conversation context.

**Workflow demonstrated in this lesson:**

*   List available AI tools (built-in and MCP) to get their exact names.
*   Navigate to Cursor Settings -> Rules.
*   Define custom rules under a heading (e.g., `## Slash Commands`).
*   Create rules mapping a slash command pattern (like `/s <term>`) to a detailed instruction specifying which tool to use and how (e.g., "Use the codebase_search tool to search for <term/>").
*   Define rules that leverage conversation context (e.g., for summarizing notes or creating commit messages).
*   Save the rules (potentially in `.cursor/rules/project.mdc` for team sharing).
*   Invoke the custom slash commands (e.g., `/s handle`, `/l`, `/c`) in the chat/agent window.
*   Observe the AI understanding the shortcut and executing the intended tool call.

**Key benefits:**

*   Creates efficient, reusable shortcuts for common AI interactions.
*   Reduces the need for typing lengthy, detailed prompts repeatedly.
*   Allows mapping simple commands to specific Cursor tool invocations.
*   Enables commands that leverage the ongoing conversation context.
*   Improves workflow speed and consistency when working with the AI agent.

Speed Up Your Agents with Cursor Slash Commands

Hey Everyone 👋,

John Lindquist here. Since I've been diving deep into AI tooling and workflows over the past year, I'm starting this weekly newsletter to share updates on what interests me and could be helpful to the community. The goal is to help everyone stay up-to-date with the latest in essentials tools for AI devs.

I'll focus on relevant community news, model evaluations for coding, notable tools like Cursor and Windsurf, and practical workflows (including tools I build myself 😇). This week kicks off Cloudflare Developer Week with major announcements around agent/MCP services that many of us will likely use soon (I'll dig in and share my thoughts next issue). There are also lots of hints about new coding models possibly coming to Cursor (though my preference remains Gemini 2.5). Personally, I'm hoping for a Gemini Flash 2.5 variant. So let's hop into the latest! For my detailed, up-to-date workflows, check out the workshop info below.

Recent egghead.io lessons about AI tooling:

* **[Swarm Intelligence: Massively Parallelize Your AI Agents in Cursor](https://egghead.io/swarm-intelligence-massively-parallelize-your-ai-agents-in-cursor~g7u2g):** A workflow using `file-forge`, AI scripting, and a custom `git worktree` tool (`wt`) to launch parallel Cursor agents on different branches.
* **[Build a Better Edit Tool in Cursor with Custom Agent Modes](https://egghead.io/build-a-better-edit-tool-in-cursor-with-custom-agent-modes~21rr0):** Create a focused 'edit mode' in Cursor for precise AI refactors without conversational overhead, overcoming Command-K limitations.
* **[Visually Clone Websites with Gemini 2.5 Canvas for Cursor](https://egghead.io/visually-clone-websites-with-gemini-2-5-canvas-for-cursor~s5ugk):** Use Gemini 2.5 Pro's Canvas feature within Cursor to analyze and replicate website structures.

---

## Cloudflare Developer Week: Agent & Platform Updates

Cloudflare announced several updates relevant to AI application development:

* **Durable Objects Free Tier:** Lowers the barrier for experimenting with stateful serverless and the Agents SDK. ([Rita Kozlov](https://x.com/ritakozlov_/status/1909231625855983939))
* **MCP Enhancements:** BYO Auth (incl. [Auth0](https://x.com/auth0/status/1909307220816228354)) and Hibernation for MCP servers aim to improve agent security and efficiency. Exercise caution with MCP, but these are steps toward production viability.
* **Workflows GA & autoRAG:** Orchestration tool is now GA; a new autoRAG product indicates more integrated AI solutions.
* **Outerbase Acquisition:** Expect Durable Objects usability improvements following this acquisition. ([Rita Kozlov](https://x.com/ritakozlov_/status/1909232149246492887))
* **OpenNext Adapter Beta:** v1.0-beta for deploying Next.js apps on Cloudflare Pages adds caching, PPR, middleware support. ([CloudflareDev](https://x.com/CloudflareDev/status/1909598450871652398))

**Summary:** Cloudflare is actively enhancing its platform for AI/agent development, particularly reducing friction with Durable Objects and MCP.

## Llama 4 & Nemotron-Ultra Released

Meta's Llama 4 models began rolling out, with initial reports of variable quality across implementations. Meta attributes this to implementation stabilization, denying training on test sets ([Ahmad Al-Dahle](https://x.com/Ahmad_Al_Dahle/status/1909302532306092107)). Nvidia subsequently released **Llama-Nemotron-Ultra** (253B parameters), derived from Llama-405B via pruning and reasoning-focused post-training. Open weights and data are available ([Oleksii Kuchaiev](https://x.com/kuchaev/status/1909444566379573646)).

**Summary:** More powerful open models are available. Llama 4's initial rollout highlights the importance of stable implementations. Nemotron-Ultra adds a high-end specialized option.

---

## Quasar Alpha: Stealth Model on OpenRouter

OpenRouter provides early access to "Quasar Alpha," a model with limited public details but notable claims: 1M token context, benchmarks reportedly surpassing Claude 3.7 Sonnet, and >4x faster speeds. Reports suggest strong coding performance comparable to O3 Mini/DeepSeek R1 but significantly faster ([Matthew Berman's Thread](https://x.com/TheMattBerman/status/1908286423976886371)). The originating lab remains anonymous; speculation includes GPT-5, Claude 4, or Gemini 3.

**Summary:** The combination of large context, speed, and reported coding strength warrants attention for agentic workflows and large codebase analysis, despite benchmark caveats. Worth testing via OpenRouter if accessible.

---

**✨ Workshop Spotlight: Conquer the Complexity of Cursor ✨**

Master practical AI development workflows in Cursor. This hands-on workshop covers Agents, Ask, Custom Modes, multi-file analysis, effective prompting, Cursor rules, and strategies for handling AI failures.

* **When:** **This Thursday!** April 10, 2025, 9:00 AM - 2:00 PM (PDT)
* **Where:** Zoom (Live Q&A included)
* **Investment:** $149

Unlock Cursor's potential and streamline your development process.

➡️ [**Read More about the Workshop**](https://egghead.io/workshop/cursor)
➡️ [**Register Now ($149)**](https://buy.stripe.com/dR67tT4P17Xa8CYfZ1)

*(Team training also available)*

---

## Quick Links

* **Nightwhisper Buzz:** Highly-rated coding model "Nightwhisper" (rumored Gemini 2.5 variant?) left LMSys Arena; release may be imminent. ([ʟᴇɢɪᴛ Tweet](https://x.com/legit_api/status/1908264443827351913), [AiBattle Tweet](https://x.com/AiBattle_/status/1907519430961025476))
* **AI SDK Speech-to-Text:** Vercel AI SDK added a `transcribe` function. ([AI SDK Tweet](https://x.com/aisdk/status/1909603418639155460))
* **Gemini 2.5 Pro Updates:** Supports YouTube URLs as input in AI Studio (transcribe, summarize, analyze); reported improvements reducing stop issues/displaying thinking tokens. ([Jean Gatt Tweet](https://x.com/jean__gatt/status/1909256747518226829), [Eric Zakariasson Tweet](https://x.com/ericzakariasson/status/1908619515023548880))
* **Baby Cursor Project:** Cursor's Ryo Lu is building a rapid prototyping environment for IDE ideas. ([Ryo Lu Tweet](https://x.com/ryolu_/status/1908328050154233863))
* **(Rumor) Qwen3 Soon?:** Hints suggest Alibaba's Qwen3 model release is near. ([Elaina Tweet](https://x.com/Elaina43114880/status/1909572798944841859))

---

If you have any feedback or any questions, feel free to respond to this email. Always happy to talk shop about the latest AI tools.

John Lindquist

[egghead.io](https://egghead.io)

AI Dev Essentials #1: Cloudflare agent updates, Llama 4/Nemotron/Quasar models, Cursor workshop

Cursor's AI Agent has the abilitly to invoke MCP tools, but they lack direct access to your project. This lesson shows how to build custom MCP (Model Context Protocol) server tools that bridge this gap. Learn to define tool parameters using Zod schemas and the crucial `.describe()` method to explicitly request necessary context from the Cursor environment *before* your tool runs. See how the AI gathers this information (like Git status or version via terminal commands, or web search results) and passes it to your tool, enabling powerful, context-aware automation.

**Workflow demonstrated in this lesson:**

*   Scaffold a basic MCP server using a Cloudflare template (`pnpm create cloudflare`).
*   Run the MCP server locally and configure it in Cursor settings.
*   Define MCP tools (e.g., `status`) within the server code (`index.ts`).
*   Specify required context for tools using Zod schema parameters and `.describe()` (e.g., requesting "The status of the git repository").
*   Interact with the Cursor Agent using natural language (e.g., "Please check the status").
*   Observe the Agent's thought process: identifying the need for specific context based on the `.describe()` text.
*   See the Agent automatically using its *own* tools (like `run_terminal_cmd` or `web_search`) to gather the requested context *before* calling your MCP tool.
*   Watch the Agent call your MCP tool, passing the gathered local context (Git status, version, etc.) as arguments.
*   Receive the final response from the Agent, incorporating the information provided by your local MCP tool.

**Key benefits:**

*   Allows Cursor's AI to access and reason about information from your local machine (Git status, versions, file contents, etc.).
*   Enables the creation of AI tools that directly interact with or report on your local development environment.
*   Demonstrates the power of using `.describe()` in Zod schemas to guide the AI in gathering necessary context automatically.
*   Unlocks advanced automation possibilities by combining AI reasoning with local machine access.

Create a Custom MCP that Calls Cursor Tools

Vimium's Vomnibar is a "command palette"-like search tool for your open tabs, browser history, and bookmarks. We can search through and jump to any tab open across browser windows. We can also do a generalized search with results that also include browser history and bookmarks which can either be opened in the current tab or a new tab. If you're anything like me, you open tons of tabs and visit hundreds of webpages over the course of a week. The vomnibar puts all of that at your fingertips.

Josh Branchaud

Open Tabs, Browser History, and Bookmarks with the Vomnibar

Vimium's Vomnibar is a "command palette"-like search tool for your open tabs, browser history, and bookmarks.

Another aspect of Vimium's configurability is customizable key mappings. We can add new mappings and override existing ones. If there is a certain Vimium feature you want to make easier to trigger, adjust the key mapping to suit your needs.

Configure Custom Vimium Key Mappings

Another aspect of Vimium's configurability is customizable key mappings. We can add new mappings and override existing ones.