🤖 AI Goes Rogue: Fake Presidents, Smart Containers & Jony Ive’s Big Bet

The Byte’s Bits

🎭 FBI Warning: AI Voice Scams Impersonating US Officials!

🛠️Docker MCP Catalog: Empowering AI Agents with Trusted Tools

🤖 AI Agents That Actually Work: 15 Workflows You Can Deploy Today

🚀 May 2025 ChatGPT Upgrades: Codex, Memory, and More!

🎭 FBI Warning: AI Voice Scams Impersonating US Officials!

Just when you thought scam calls couldn’t get any more absurd, AI decided to join the party, bringing a whole new level of creepy with it. The FBI is now waving the red flag: cybercriminals harness AI-generated voices and messages to impersonate high-ranking US officials. Your phone might now be “calling” you from the Pentagon, but plot twist — it’s some guy in a basement with ChatGPT and a dream.

Forget the old-school scams with broken English and dodgy grammar. These new fraudsters have leveled up, using sophisticated AI to create hyper-realistic voice clones and slick SMS campaigns (smishing and vishing, if you're nasty). And who are they targeting? Oh, just minor names like government officials and their associates. No big deal. It's phishing with flair — and your caller ID won’t save you this time.

• 📱 AI-Powered Deception: Voices so real, even mom would be fooled

• 🕵️ High-Profile Targets: Because if you’re going to scam, scam big

• 🔗 Fake Secure Links: Click here to “secure your data” — and instantly regret it

• 🚨 Phishing 2.0: Where truth and deepfake blur into one annoying ring tone

Scammers are getting smarter, and unfortunately, so are their tools. It’s like Mission: Impossible meets Craigslist fraud. Stay sharp, question everything, and maybe don’t answer calls from the “President”.

🛠️ Docker's MCP Catalog: Empowering AI Agents with Trusted Tools

Docker has launched the MCP Catalog, a curated collection of over 100 secure, high-quality MCP servers as Docker Images. This initiative aims to streamline the integration of AI agents with reliable tools, enhancing their capabilities in various domains.

I find the Docker MCP catalog is a step in the right direction to streamline MCP servers and their packaging. AI sometimes tries to reinvent the wheel of software development by installing everything locally. Now with MCP servers, it is easy to spin up a prepackaged approved MCP server in seconds. Now Docker’s motto expands to Build, Ship, Run, and AI

Docker MCP Catalog Features:

• 🔐 Security First: Each MCP server is vetted to ensure safe interactions with AI agents.

• 🧰 Diverse Toolset: The catalog includes various tools, from database solutions to developer utilities.

• 🚀 Seamless Integration: Designed for easy deployment, these Docker Images facilitate quick setup and execution.

• 🌐 Centralized Access: Developers can explore and utilize these tools directly from Docker Hub. Docker Hub

By providing a centralized, secure repository of MCP servers, Docker is enabling developers to enhance their AI agents' functionality with confidence.

I’m currently working on building a few MCP servers around the Google Pagespeed API. I will share the results next week.

🤖 AI Agents That Actually Work: 15 Real-World Workflows You Can Deploy Today

I first discovered n8n several years ago as an automation tool. Fast forward to today, n8n has now become the de facto open-source standard for building AI automation workflows. So if you want to build your own AI agent or automate sending data to ChatGPT, n8n is your tool, and again, it’s Open Source!

AI agents aren’t just theoretical anymore—they’re automating real business tasks, from scraping websites to writing blog posts. n8n’s latest blog post highlights 15 practical examples of AI agents that can save you time and reduce manual work.

• 💬 Conversational Agent: A basic chatbot using OpenAI and SerpAPI to handle user queries with memory buffer capabilities.

• 🕵️‍♂️ Vision-Based Web Scraper: An agent that extracts data from websites without dealing with CSS selectors, using ScrapeBee and Gemini.

• 📝 Blog Writing Assistant: An AI agent that generates SEO-optimized blog posts and images for WordPress sites.

• 📊 Data Analyst Agent: An agent interacting with a SQLite database to answer natural language queries.

• 📈 Marketing Team-in-a-Box: A single AI agent that creates and edits images, writes LinkedIn posts, and generates video content.

These examples demonstrate how AI agents can be integrated into workflows to handle complex tasks autonomously. Whether you're looking to enhance customer support, automate content creation, or streamline data analysis, there's likely an AI agent setup that fits your needs.

The barrier to building agents continues to drop. Now, with MCP and n8n, you can easily automate many of your current workflows.

🚀 May 2025 ChatGPT Upgrades: Codex, Memory, and Jony Ive!

OpenAI has rolled out powerful updates to ChatGPT this month, transforming it into a more capable and personalized assistant.

🧠 Codex: Your AI Coding Companion

I was super excited about Codex, but I was immediately disappointed when I saw it wasn’t available in my account. Next, my colleague mentioned that I’d seen the announcement about Codex and that it was now available in ChatGPT. I checked quickly, and Boom, we’re in business.

Once I had Codex set up, I connected it to my Docker Prometheus repo to see how it performed. I quickly had Codex scan the repo and report any findings or improvements. Codex found several YAML errors, spelling mistakes in the README, flags no longer supported, and updated my .gitignore, so nothing mind-blowing.

Next, I asked Codex to update all the components to the latest supported version. This is where Codex stumbled a lot. First, it only updated the components, for example, from 2.0 to 2.1, when the latest version was actually 3.4. I had to prompt quite a bit to tell Codex which version to upgrade to for each version. Codex wanted to pin each version to the tag “latest”, and as we know, this is not best practice, but Codex tried to convince me otherwise with several back-and-forths, which were frustrating until I put my virtual fist down and said pinning to a specific version is what I wanted.

So, what is Codex? The new AI agent, integrated with ChatGPT, is designed to assist developers with coding tasks. Built on the codex-1 model, a variant of OpenAI's o3 reasoning model, Codex can:

• Write, debug, and test code autonomously within a secure, cloud-based sandbox.

• Understand and adapt to your codebase, even matching your coding style.

• Handle multiple tasks in parallel, delivering results in minutes.

Is it ready to take over developers’ jobs? In my opinion, no, but it can help you triage issues, clean up, etc., which could definitely save time.

🧠 Enhanced Memory: Personalized Interactions

The new Enhanced Memory feature is mind-blowing or memory-blowing! Enhanced memory has been a game changer for my use with ChatGPT and has unlocked many new workflows for me.

ChatGPT can now remember details from past conversations, allowing for more personalized and context-aware interactions. This means:

• No need to reintroduce yourself or repeat preferences in each session.

• Improved continuity in long-term projects or ongoing tasks.

• Ability to recall and build upon previous discussions, enhancing

Users have full control over this feature, with options to view, edit, or delete memories, and to disable memory entirely if preferred.

How do I use Enhanced Memory? Glad you asked. Here are some examples:

I’ve started using the same chat and referencing other chats now. This has helped considerably with productivity in picking up the previous context.

Some examples of how to use enhanced memory or check the X post for more examples.

• Remember when we discussed X. Can you add XX to the conversation? (This helps build on previous conversations)

• If you are making changes, record each step into ChatGPT back into the prompt. So, if ChatGPT gives you code, tell the prompt what the outcome was. This way, you can have a log of changes and ask ChatGPT for a report of all changes.

• Create a report of all discussed items by date. Having a living changelog is really helpful from a coding perspective.

🤝 Jony Ive Joins OpenAI to Design the Future of AI Hardware

In a landmark collaboration, OpenAI has acquired Jony Ive's AI hardware startup, io, for $6.5 billion. This partnership aims to develop a new category of AI-powered devices seamlessly integrating into daily life.

• Innovative Design: The upcoming device is envisioned as a compact, screenless gadget, potentially wearable, designed to offer unobtrusive AI assistance.

• Mass Production Timeline: Production is slated to begin in 2027, with the device expected to be slightly larger than the Humane AI Pin and reminiscent of the iPod Shuffle in form factor.

• Strategic Shift: This move signifies OpenAI's transition from a software-centric company to a player in the consumer hardware market.

With Ive's design expertise and OpenAI's AI capabilities, this collaboration could redefine how we interact with technology!