In 2026, AI (LLM) sandboxing is an exhaustive and exhausting ecosystem. The delicate balance of restricting your agents permissions whilst paradoxicially letting them do a lot of privilieged action...

Don’t worry! This isn’t actually a 75 minute read! It includes a bunch of LLM output. The other week, I posted the article Where’s the moat?. The intention was to capture my feelings around th...

This post documents my observations and experiences regarding the current state of the tech industry, specifically the quality of LLMs and their adoption. I know a few of my peers feel the same, an...

I had an idea for a PII anonymization service, with the intention of maintaining integrity with the data it’s anonymized. The original context was for anonymizing Microsoft Sentinel database dumps...

Summary This blog post explores three AI guardrail products, and how to integrate them with the Strands Agents SDK. I opened a pull request into the Strands Agents Samples repo two months ago, and ...

Summary Modern Microsoft Azure hypervisors support nested virtualization; meaning you can run virtual machines within virtual machines. With a bit of magic, you can create an effective “Virtual ESX...

Unlocking Next-Gen Kubernetes Security with eBPF tooling and Microsoft Sentinel This post explores the setup and configuration of Cilium and Tetragon in Azure Kubernetes Service and integrating &am...

Google Cloud - Managing Google’s Unverified Authentication Library In Google Cloud, Application Default Credentials (ADC) allows your code/applications to automatically find and use credentials. F...

TL;DR Observable systems can support a security function but are distinctly different. Buzzword cloud “observability” tooling typically only covers either VMs or Linux-based systems such as Ku...

Preamble I keep hearing the term observability thrown around a lot, often used in the wrong context. Working in Cybersecurity, I think people often confuse the idea of observability with blue team...