Writing
Insights
Field notes on AI, engineering, leadership and the technology landscape. Written for builders and the people who lead them.
Why a CTO is the single most important C-Level in 2026?
Not because the CEO, CFO, COO or CMO matter less. They don't. Every one of those roles is critical and I've worked with brilliant people in each. But no other C-level seat is being rewritten as fast and as fundamentally as the CTO's.
Read→
Stop giving AI Agents your password
A Replit agent wiped a production database. A Google Antigravity agent erased an entire drive. Both had valid credentials. Here's why your AI agents need their own identities, and what to do this quarter.
Read→
The Anthropic IPO is about to hit and we’re not ready for this!
Anthropic is lining up an IPO that could unlock tens of billions for AI safety giving. The problem: the field has maybe 60 people in the world who know how to spend it well.
Read→
Post-Mythos Playbook: What every CISO should do right now
A follow-up to my earlier article on Claude Mythos. The Cloud Security Alliance just published a 250-CISO-reviewed action plan. Here's what to do this week, this month, and this year.
Read→
How Do You Measure Developer Productivity in an AI-Augmented World?
Meta and Spotify are grading engineers on AI token usage, and it is not going to work. Token consumption is an input metric. What matters is DORA plus a handful of new signals, and here are the top five I would actually watch.
Read→
Most Enterprises Are Spending on AI and Hoping. That's the Real Story.
KPMG's latest survey reveals 75% of leaders will keep investing in AI despite economic uncertainty and elusive ROI. The deeper problem? We've never properly measured the work AI is replacing.
Read→
The Career Ladder Just Got an Elevator: Evaluating Engineers in the AI Era
With 84% of developers using AI tools, the old promotion playbook is broken. Lines of code and tickets closed no longer separate junior from senior. Here's what actually does, and why cutting junior hiring now is a self-inflicted talent crisis.
Read→
The First 90 Days as a CTO: Playbook
You just landed the CTO title. The instinct is to prove yourself fast, ship something big, and reshape the tech stack. But the research is clear: your first 90 days should be about listening, aligning, and earning one small, visible win. Here is the playbook.
Read→
AI Security in 2026: Three Fronts Every Technical Leader Must Hold
The 2026 Red Canary Threat Detection Report frames AI security across three dimensions: defending against AI-powered attacks, protecting your own AI infrastructure, and using AI to strengthen your SOC. Here's what matters for your organization.
Read→
AI Browser Extensions: The real danger
A new report reveals that AI browser extensions are 60% more likely to have vulnerabilities, 3x more likely to access cookies, and nearly 6x more likely to escalate permissions. Yet most enterprises have zero visibility into them.
Read→
Why most AI pilots fail - and what to do instead
The gap between AI proof-of-concept and production value is where most organizations lose momentum. Here's how to close it.
Read→
Kubernetes in the Mittelstand: Right-sizing cloud-native for mid-market
Germany's Mittelstand is facing a technology crossroads. The question isn't whether to adopt cloud-native technologies — it's how to do it without overengineering.
Read→
How Uber Broke Up a 10 PB Monolith Without Breaking a Single Query
Uber migrated 16,000 Hive datasets and 10+ petabytes from a monolithic warehouse to a federated architecture with zero downtime. The secret? Pointer-based redirection in the Hive Metastore. Here is the playbook and why it matters for every data team at scale.
Read→
An AI Just Found Thousands of Zero-Days!
Anthropic's Claude Mythos autonomously discovered thousands of critical vulnerabilities across every major OS and browser. The model even escaped its sandbox and posted exploits online without being asked. AI-powered offensive security is no longer theoretical, and the implications are staggering.
Read→
Meta's Muse Spark: When the Open-Source Champion Goes Proprietary
Meta spent three years and 1.2 billion Llama downloads building its reputation as AI's open-source champion. Then it launched Muse Spark, a fully proprietary model with 10x compute efficiency. Here's why even the most committed open-source company decided frontier AI needs a lock on the door.
Read→
Databricks Just Open-Sourced Its Semantic Layer. Here's Why That Matters for Every Data Team.
Unity Catalog Business Semantics is now GA, and Databricks is contributing the core Metric View implementation to Apache Spark. Define your business metrics once in SQL, share them across every BI tool and AI agent. The open-sourcing signals that the semantic layer war will be won on adoption, not lock-in.
Read→
The Axios Hack: How North Korea Weaponized a Video Call to Compromise 100 Million Developers
North Korean state actors spent weeks building trust with the Axios maintainer, then used a single fake video call to plant a RAT on his machine. Three hours later, trojanized versions of one of npm's most popular packages were live. 2FA, OIDC, code signing: none of it mattered because the attack targeted the human, not the toolchain.
Read→
The Hyperscaler Tax on AI Is Real, and Vultr Just Found a Way Around It
SUSE Rancher Prime and SUSE AI just landed on Vultr's Marketplace, giving enterprises a concrete path to run AI inference on Kubernetes across 32+ global regions with B200, H100, and MI300X GPUs. The pitch: open-source orchestration, no vendor lock-in, and costs that could be 50-90% lower than the Big Three.
Read→
AI Coding Agents Are Sleepwalking Into the Same Trap as Cloud
JetBrains surveyed 11,000 developers and found a familiar pattern: 90% use AI, but almost nobody is tracking what it costs, whether it works, or who controls it. We have seen this movie before. It was called cloud sprawl.
Read→
ScaleOps Just Raised $130M Because Most Kubernetes Clusters Are Half Empty
The average enterprise Kubernetes cluster runs at 30-50% utilization. ScaleOps just raised $130M at an $800M valuation to fix that, and AI workloads are making the waste even harder to ignore.
Read→