OpenAI has greatly enhanced the Work with Apps functionality in the ChatGPT app. Users can now incorporate the outputs from these apps as context for their prompts in ChatGPT and implement code changes in Cursor, VS Code, and various other IDEs. This feature positions the ChatGPT app as a competitor to Cursor and GitHub Copilot in VS Code. Furthermore, outputs from well-known terminal applications such as Warp and iTerm2 can be used as context within ChatGPT.

Source

Zach Lloyd, a former Principal Engineer at Google, introduces new features at warp speed to the best terminal app in the galaxy, with the latest Warp release introducing cutting-edge AI technologies from The Next Generation. The standout addition is the Dispatch mode, which I prefer to call reasoning mode, enabling administrators to plan complex automation tasks with AI assistance. Prompt Suggestions might better be described as big brother or creepy mode since the AI continually observes your actions and intervenes when it believes you need assistance. Additionally, AI memory functions similarly to the personalization feature in ChatGPT, allowing you to save preferences as Knowledge across sessions.

Source

We have reviewed Warp's Mac and Linux versions multiple times, as it stands out as the top terminal application. The newly released Windows edition introduces AI-driven automation for PowerShell administrators. Even if you have no use for natural language automation because PowerShell is your primary language anyway, you should give Warp a try, as no terminal app can compete with it on Windows.

Source

When I first watched Anthropic's demo of Claude Code, I was thrilled because I believed this tool could be a valuable asset as an AI-driven CLI assistant akin to Warp's AI agent. After all, who even codes in a terminal anymore?

Source

A few weeks back, I reviewed the new AI features of Warp, the best terminal app out there. Now, Warp has taken the next step, transforming the terminal into a fully autonomous AI agent.

Source

In my previous post, I detailed the process of importing open-source models, such as DeepSeek R1 from Hugging Face, into Amazon Bedrock, a fully managed service offered by AWS that gives users access to foundational models from top AI firms and Amazon. This post will explain how to access the AI models through the Bedrock API from third-party apps like VS Code or Cursor. Specifically, when using AI models like Claude Sonnet 3.5 for coding, accessing it through Amazon Bedrock offers enhanced security compared to Anthropic or model hub APIs such as OpenRouter.

Source

In a previous post, I demonstrated how to run DeepSeek R1 on your Mac or PC. Today, I will guide you through importing the Chinese open-source model that has outperformed OpenAI’s top models in various benchmarks from Hugging Face—the GitHub equivalent for hosting, sharing, and deploying open-source AI models. Hugging Face provides the hf-bedrock-import tool for this purpose, which is more straightforward than the aws bedrock create-model-import-job command.

Source

To keep pace with Cursor, Microsoft has integrated several valuable features from the VS Code fork into its GitHub Copilot extension. The February update introduces three new features: Agent Mode, Vision, and Next Edit Suggestions (NES). You'll appreciate these latest additions if you prefer VS Code over Cursor for some reason.

Source

In a prior article, I showcased automating PowerShell management in natural language using the VS Code extension Cline. Roo Code (formerly Roo Cline) is a Cline fork that provides extra valuable features for the AI-enhanced IT pro. Roo Code remains comparable, and the insights I shared about working with MCP servers in Cline also apply to Roo Code. The main difference is that Roo Code serves system administrators more effectively, as it provides a mode more suitable for natural language use in a CLI. Additionally, Roo Code's support of VS Code's Code Actions feature makes it perfect for understanding and modifying a colleague's scripts.

Source

In my previous posts, I introduced Anthropic's Model Context Protocol (MCP) and explained how to install an MCP server for the Claude Desktop app. Today, I show you something more relevant for admins. My introductory article about Cline mentioned that the VS Code extension supports MCP. You will see that installing an MCP server is easier with the assistance of an AI. In this post, I will explain how to install AWS MCP, an MCP server that allows you to manage AWS resources with the help of an AI. If you don't want to be left behind, I highly recommend getting started with learning how to use AI to get your job done. This tech revolution is advancing faster than any before it. And we are only at the very beginning.

Source

In my last post, I introduced Anthropic's new Model Context Protocol (MCP), which enables AI agents to access and manipulate external data sources. In today's article, I will illustrate the MCP concept by installing MCP Server Fetch, which allows the Claude desktop app to retrieve webpages and convert them to Markdown. I believe that managing MCP servers will likely be one of the main tasks for AI ops admins.

Source

The Model Context Protocol (MCP) is an open standard developed by Anthropic to facilitate secure, two-way connections between AI models and external data sources, such as databases, business tools, and development environments. It enables AI agents to make informed decisions to perform tasks autonomously. Given that Anthropic's announcement is just a few weeks old, the number of available MCP servers is impressive. In the near future, installing and configuring MCP servers will likely be one of the primary tasks for AI Ops admins responsible for deploying AI agents

Source

Since DeepSeek’s R1 model is built for efficiency during training and inference, I wanted to test how one of the distilled R1 models with fewer parameters performs on Ollama on my MacBook. Of course, you can also experiment with DeepSeek R1 on your Windows PC.

Source

When deploying a customized Large Language Model (LLM), your first decision is whether to use fine-tuning or embeddings-based Retrieval-Augmented Generation (RAG). Many online introductions explain the difference in a highly abstract manner. This post concentrates on the practical implementation differences between fine-tuning and RAG, which are crucial for IT professionals tasked with LLM adaptation.

Source

In a previous post, I discussed the Pushover app, which can send push notifications to admins if a server encounters issues. Today's post features a sample script that monitors a server and alerts the admin if any issues arise. Linux and its applications are incredibly reliable. However, a rogue script running into an infinite loop might deplete resources, disrupting services. In such cases, you can automate the restart of services with a Bash script, and if that fails, automatically reboot the server.

Source

In a previous post about the new AI features in the Warp terminal, I suggested that AI interfaces would eventually replace the CLI and facilitate the automation of script generation. The Cline extension for Visual Studio Code is a significant step in this direction, allowing you to create scripts using natural language, execute them with AI support in the terminal, rely on AI to analyze shell output and respond to errors. While the extension supports various shells, I'll demonstrate how to utilize Cline to automate PowerShell administration.

Source

Google just announced that Gemini integration in Google Workspace is now available for free. Users in Europe and Japan must first turn on this feature. Workspace admins can enable or disable Gemini in the Google Admin Console. You can also use Gemini to access Gmail if you have a personal Google account, but there are limitations. I tested the Gemini integration in Gmail and shared my experiences in this post.

Source

Warp, the best terminal app by far, has introduced impressive new AI features: Active AI and Agent Mode. I believe Agent Mode is a real game changer, showcasing the future of system administration, as it will likely make learning a shell language superfluous.

Source

Pushover is a versatile push notification service for iPhone, Android, Mac, and Windows (via browser plugin), providing real-time alerts from various sources like web applications, network monitoring systems, and IoT devices. It integrates with platforms like IFTTT and Zapier and supports custom scripts, making it useful for admins and developers.

Source