Insights on cybersecurity, architecture, and engineering
Every system eventually fails. The question is whether systems are designed to fail in ways you can detect, contain, and recover from. Closes the series arc on resilience as an intentional architectural property.
Read more →Tokens are the unit your AI bill is measured in and the unit a model's memory is limited by, yet they're rarely explained. A plain-English breakdown of what a token actually is, how big one is, and why it quietly governs cost, context, and a few of AI's stranger failures.
Read more →Security maturity models are useful measurement tools but misleading as goals. Optimizing for a score is subtly different from building genuine resilience. What mature security behavior actually looks like in practice.
Read more →Ransomware is the final act in an attack chain — entry, lateral movement, escalation, encryption. Remove the malware and the underlying architectural conditions remain. Root cause analysis of why ransomware works.
Read more →Every architecture choice is a security choice — whether treated that way or not. The orgs with the worst posture aren't those that made bad security choices; they're the ones that didn't recognize they were making security choices at all.
Read more →Most organizations have an IR plan. Most have never practiced it under realistic conditions. The gap between documented procedure and operational reality is where manageable events become disasters.
Read more →Log4Shell, XZ Utils, SolarWinds — all through code organizations trusted without verifying. Supply chain risk is an architecture problem, not a patching problem.
Read more →Keepacy is a secure vault and delivery system for the things families actually need when someone is gone. Here's how we built it, and why.
Read more →SOC 2, ISO 27001, HIPAA, CMMC — passing them is not the same as being secure. Compliance describes a floor, not a ceiling.
Read more →Blast radius — how damage propagates from a single point of failure — is a design choice. Most architectures are brittle on containment.
Read more →Zero trust was a response to architectures that assumed safety based on location. AI agents are being deployed with the same assumption.
Read more →When an AI agent acts on behalf of a user, the identity model is more complicated than 'the user authorized it.' That gap has consequences.
Read more →How we built uRoute's freight optimization software — multi-module Java TMS with an automated load-matching engine that delivers 15-30% freight cost reduction for broker operations.
Read more →I've been running an AI system with real access to real infrastructure. The architecture, the decisions, and what I'd do differently.
Read more →Prompt injection is being treated as a prompt problem. It isn't. It's an architecture problem — and it's the same one we've already described.
Read more →AI is being described as a fundamental transformation of the security landscape. Some of that is true. Most of the important parts aren't.
Read more →Humans make mistakes at a predictable rate. Security is one of the last domains still designing systems that require perfect behavior.
Read more →MFA was supposed to end credential theft. It didn't. Understanding why reveals the architectural mistake we keep making.
Read more →Why treating authentication as a one-time gate instead of a continuous signal is the root cause of credential theft incidents.
Read more →How we built a federated lung cancer research platform for IASLC — hub-and-spoke architecture, portable Docker algorithm containers, and 80% compute reduction across 10+ international sites.
Read more →How we designed a linear broadcast scheduling interface for NBCUniversal — 20+ network feeds, time-zone math, zero-downtime deployments, and the architecture that holds it all together.
Read more →How we built lung cancer screening software on the bones of VA-PALS — modernizing MUMPS to TypeScript, integrating AI-assisted CT reading, and shipping it to three continents.
Read more →