The System Produces Bad Outcomes, Not Bad PeopleIf you've worked in security long enough, you've sat in this The Incentive Mismatch — They're Not Your EnemyEvery security leader eventually decides the infrastructure team is th The Execution GapSomewhere in your organization there's a list. It might be a Jira The Same Bet, TwiceA while back I helped build a company called Drawbridge Networks. The Rollback Has to Be Designed InA rollback you haven't exercised is not a rollback. It's a hope with a UI. SOAR, ITSM, and Consultants: Nobody Completes the RoadmapPick any well-run security program and look at the tools it runs on. T Fit Over Correctness: The Security Tool GraveyardEvery security team I've run had a graveyard. Not of ideas that w Resilience Is the Right Frame"Security" is a word that quietly tells you what to build. I The rollback nobody builtEvery infrastructure and security tool has some notion of rollback. No The Rollback GuaranteeI've watched a lot of good changes not happen because nobody coul Chesterton's FenceThere's an old idea from G.K. Chesterton that I think about more The Asset Inventory ProblemEvery security program depends on knowing what you have. And in most o The AI Safety Harness, Part 1: The IncidentsWhile I was building Nexplane, I asked Claude Code to fix a VPN proble The AI Safety Harness, Part 2: The MechanismIn Part 1 I argued that you can't prompt your way to a safe infra The Nexplane ManifestoInfrastructure changes are inevitable. Outages are optional. Every mea Security PhilosophySecurity is the ability to safely improve a system, not merely protect Infrastructure Changes Are Inevitable. Outages Are Optional.Every meaningful organization eventually has to change production infra Safe Execution Is the ProductA lot of infrastructure products sell visibility. Visibility matters, The Change Request Is the ProductMost products treat a change request as paperwork. Nexplane treats it Blast Radius Should Be Computed, Not GuessedEvery production change has a blast radius. In many organizations, tha Simulation Before AutomationThe industry is obsessed with autonomous execution. I think simulation Infrastructure Is a Versioned SystemOne of the biggest mistakes in enterprise infrastructure is treating ch AI as Planner, Not DictatorAI is strongest when it helps create plans that humans and systems can Human-in-the-Loop Is Not a CheckboxHuman-in-the-loop is often used as a comforting phrase. In practice, Infrastructure Change IntelligenceThe long-term moat is not the model. The moat is accumulated operatio Why I'm Open Sourcing NexplaneThis project started because I repeatedly encountered the same problem Open Source First, Commercial LaterThe open-source core of Nexplane should be genuinely useful. This matt The Databricks AnalogyDatabricks did not become valuable because Spark existed. It became va What Belongs in the Open CoreThe open core should include enough value that practitioners can solve

Back to Basics: Network — MicrosegmentationAsk any security team whether they should microsegment the network and Back to Basics: OS — Kernel Upgrades and the App Nobody Will TouchNobody argues about whether you should patch the kernel. Keep the OS c Back to Basics: Application — Seccomp and the Syscall You Don't Know AboutSeccomp is one of the better controls Linux gives you, and almost nobo Back to Basics: Identity — Least Privilege Requires a Map Nobody HasAsk a room full of security people what good identity hygiene looks li Back to Basics: Secrets — Rotation Requires an Inventory You Don't HaveAsk any security team whether they rotate their secrets on a schedule Back to Basics: Observability — The Log Agent That Never Got DeployedNobody argues about this one. You cannot detect what you cannot see, a Back to Basics: Cloud — The Security Group Nobody Will TightenEvery cloud environment I've ever worked in has one. A security g

How to Rotate Keys Without PanicKey rotation is one of those tasks everyone agrees is important and ma How to Patch a Production Service SafelyPatching sounds simple until the service matters. The actual patch sho How to Move DNS SafelyDNS changes are small until they are not. A safe move starts long befo How to Microsegment One Application FirstMicrosegmentation fails when organizations try to boil the ocean. Sta How to Containerize a Legacy Service Without Breaking ProductionContainerizing a legacy service is rarely just a Dockerfile. The hard

Why we built NexplaneI've had a version of the same conversation more times than I can We gave an AI agent infrastructure access. Here's what we learned.There's a version of "AI for infrastructure" that sound Why security teams keep creating work for engineeringIf you've worked in security long enough, you've been in thi What actually happens when you try to recover an AD forest after ransomwareIn May 2017, WannaCry hit the NHS. Within hours, hospitals were divert Built on Real InfrastructureThere is a meaningful difference between software that demos well and The Cost of Fear in InfrastructureOrganizations pay a hidden tax when they are afraid to change infrastr The Day I Stopped Trusting AutomationAutomation isn't dangerous because it is fast. It's dangerous Why I Didn't Start With the UIThe honest answer is that the UI is not good enough yet. That is not

Fear vs. RiskOrganizations often confuse fear with risk. Risk is an estimate. Fear AI Doesn't Need Root. It Needs a Better Engineering Process.The debate about giving AI production access may be asking the wrong Reasoning Is the New Source CodeFor decades, software engineering has treated source code as the prima Your Architecture Is a Record of Your AssumptionsAsk an engineer why a system looks the way it does and you'll oft Stop Measuring Lines of Code. Start Measuring Engineering Judgment.Software engineering has always struggled with metrics. Postmortems Should Measure the Quality of the Reasoning, Not the OutcomeMost engineering postmortems begin with the same question: Architecture Decision Records Are Missing the Most Important SectionArchitecture Decision Records have become a staple of modern software Every Production Change Is a Scientific ExperimentMost engineering organizations don't think of production changes The Most Important Engineering Metric Doesn't Exist YetSoftware engineering has no shortage of metrics. Your Engineering Organization Has a Memory LeakIf I told you one of your production services had a memory leak, you&# Your Best Engineers Aren't Bottlenecks. They're Uncompiled Knowledge.Every engineering organization has them. Why Every Production Change Should Begin With a PredictionOne habit separates engineering from trial and error: prediction. The Real Productivity Killer Is Decision ReconstructionAsk most engineering leaders what slows teams down and you hear famili The Best Architecture Review Starts With One QuestionMost architecture reviews begin with diagrams. AI Doesn't Need Chain of Thought. It Needs a Chain of Evidence.One of the most interesting AI debates centers on reasoning. How do we The Engineering Org Chart Is Missing a Role: The Organizational Memory EngineerEngineering org charts have familiar roles: software engineers, platfo The Future of DevOps Is DecisionOpsDevOps changed software engineering by collapsing the distance between The Most Dangerous Words in Engineering: We've Always Done It This WayEvery engineering organization has heard the phrase: The Best Engineering Teams Build Time MachinesSoftware preserves implementation. It rarely preserves intent. Complexity Is Interest on Yesterday's DecisionsEngineering leaders often explain complexity with growth: more custome The Highest-Leverage Engineering Work Is Often InvisibleThe highest-leverage engineering work often produces no visible output Your Change Management Process Is Probably Optimizing the Wrong ThingAsk why a change management process exists and you hear predictable an Why Every Infrastructure Team Eventually Reinvents a Control PlaneEvery sufficiently large infrastructure organization eventually builds Every Outage Is a Broken Mental ModelIncidents are usually described as failures of software. Documentation Is Dead. Long Live Context.Documentation has a maintenance problem. Engineering Organizations Have a Context BudgetEvery engineer can understand only a finite amount of complexity. Infrastructure Should Remember More Than ConfigurationInfrastructure remembers desired state. The Best Engineering Organizations Optimize for RegretEngineering decisions are made under uncertainty. Why Most Platform Engineering Teams Eventually Become Internal Product CompaniesPlatform teams often begin as infrastructure teams. Engineering Debt Isn't Technical. It's Cognitive.We talk about technical debt constantly. The Next Competitive Advantage Is Organizational IntelligenceThe strongest engineering organizations will not merely hire smarter e Every Architecture Diagram Is Missing TimeArchitecture diagrams show topology. Engineering Is Becoming an Information ScienceThe bottleneck in engineering is shifting. What Is an Engineering Control Plane?Cloud platforms have control planes. Kubernetes has a control plane. N Why Infrastructure Needs Version Control for DecisionsGit versions code. Terraform versions infrastructure. Kubernetes recor The Future Infrastructure StackToday's infrastructure stack has compute, storage, networking, ob Why AI Should Review Infrastructure Before It Changes InfrastructureThe highest ROI for AI in infrastructure may come before execution. The Missing Feedback Loop in InfrastructureMost infrastructure workflows have plan, deploy, and monitor. Infrastructure Should Explain ItselfImagine clicking on a firewall rule and seeing not just its configurat The Next Decade of EngineeringTwenty years from now, we may look back at today's engineering pr