AI

Everyone wants to ship AI into production. Almost no one wants to own what happens when it goes wrong. I’ve been in enough rooms with financial services clients to know how this plays out. A team builds something impressive on Bedrock — a RAG-powered knowledge assistant, an internal compliance copilot, a customer-facing chatbot. The demo looks great. Then someone in Legal raises their hand. What happens if it leaks a customer’s SSN? What if it makes a recommendation that sounds like investment advice? What if a clever user tricks it into ignoring your system prompt? ...

I wrote this article for Ippon on February 10, 2026. The enthusiasm around projects like OpenClaw—an open-source framework enabling autonomous task execution across messaging platforms, file systems, and enterprise APIs—reveals a critical blind spot in enterprise technology governance. For an entire week, I heard nothing but discussions of moltbots, OpenClaw, Clawd, and various implementations being created with these tools. But when I realized the permissions being exposed and how enterprises were approaching agent governance, I felt compelled to document this moment in AI history. ...

What is the Model Context Protocol? The Model Context Protocol (MCP) is an open standard for connecting AI agents to external systems. Think of it as a universal adapter that lets any AI agent talk to any tool or data source without custom integration code. Anthropic announced MCP in November 2024 and donated it to the Linux Foundation’s Agentic AI Foundation about a year later, in December 2025. The adoption has been swift: OpenAI integrated it into ChatGPT, Google DeepMind uses it for Gemini agents, AWS built AgentCore around it, and development tools like Zed, Sourcegraph, Replit, and Codeium all support it. In just a few months, the community has built thousands of MCP servers. The protocol has become the de-facto standard for agent-to-tool communication. ...

Here’s what should make every security leader uncomfortable: organizations routinely deploy vulnerable code to production to meet delivery deadlines. Not because they don’t care about security. Because security can’t keep up. Over 60% of organizations update their web applications weekly or more frequently. Nearly 75% test those applications for security monthly or less. The math doesn’t work. The gap between development velocity and security validation grows wider every sprint. At re:Invent 2025, AWS CEO Matt Garman announced AWS Security Agent—not as another security scanning tool to add to the pile, but as a fundamentally different approach to the problem. ...

At an AWS Road Show this fall, Darko Mesaros demoed a URL shortener he’d built in Rust called krtk.rs. Something about watching a clean, fast URL shortener just work stuck with me. I’ve built a few of these for demos since then, but I wanted to try something different this time: build one in Python with a retro 90s vibe, and let Kiro handle most of the heavy lifting. Kiro is one of AWS’s three frontier agents announced at re:Invent 2025—autonomous AI systems that maintain context and work independently for hours. While DevOps Agent handles incident response and Security Agent conducts penetration testing, Kiro is your AI developer that takes specifications and generates production-ready code. ...

At re:Invent 2025, AWS CEO Matt Garman announced something that made me stop and actually pay attention during a keynote—which doesn’t happen often. He introduced frontier agents: AI systems that don’t just help you write code or answer questions. They work autonomously for hours or days, maintaining context, investigating problems, and making decisions without you holding their hand. Three agents got announced: Kiro - your AI developer AWS Security Agent - your AI security engineer AWS DevOps Agent - your AI operations engineer This isn’t another coding assistant that autocompletes your Lambda functions. This is AWS betting that AI agents can handle the kind of multi-hour incident investigations that currently wake up humans at 2 AM. ...