Enterprise IT, rewritten in real time.
Field notes on Azure, identity, security and the AI shift, from someone shipping the fixes, not just the takes.
The latest
- When You Press Touch ID on a Mac, What Actually Happens? Your fingerprint never leaves the Mac, not encrypted, not hashed, not transmitted. The Secure Enclave gives apps a verdict, not your biometric data, and understanding that distinction is where most authentication implementations…
- Three Things Broke When I Treated HomeView’s Face Recognition as an Authenticator Apple's first smart-home device with facial recognition won't say where the biometric match happens, on the panel or on your paired iPhone. That architectural gap becomes a dead end the moment you…
- Nvidia Was Underwriting the Demand for Its Own Chips and Calling It a Market Under Jensen Huang, Nvidia invested tens of billions in AI startups, then much of it flowed right back as chip orders. His departure breaks the loop that made subsidized demand look like…
- Can Your Agent Drain Your Wallet? OpenAI Just Made That a Real Question Give it a quarter and this will happen: an agent reads a malicious README, decides it needs to buy emergency API credits, and the only thing between injection and invoice is a…
- You Patched PostgreSQL Last Night. Why Are You Still Vulnerable This Morning? Updating the PostgreSQL server packages closed 52 vulnerabilities including a 22-year-old RCE. But if you run C-language extensions, you're not done, and most production databases do.
- I Hunted Rogue ScreenConnect Agents for a Week. Detection Lost to Design. SCKit malware weaponizes ScreenConnect to turn help-desk sessions into persistent backdoors. I ran two defensive strategies head-to-head against Microsoft's published attack chain, and learned that detecting the compromise is not the same…
- Google Shipped Gemini 4 Hydrogen to Global Search Without Publishing One Benchmark The newest Gemini is already powering Search globally, but Google published zero benchmarks, no MMLU, no SWE-bench, nothing. You can't diff a vibe against your eval set.
- Purview Logged What Admins Changed. Now It Logs Who Walked In. Two new audit events now capture who opened the compliance portal, eDiscovery, or Insider Risk Management, and with which role. But they won't catch the admin who never uses their privilege at…
- Three Things That Break When Your Coding Agent Runs a Poisoned Model A poisoned model doesn't need a malicious pickle file to betray you. Microsoft shows how attackers embed triggers in the weights themselves, turning your coding agent into a remote shell when the…
- I Tested Front Door mTLS for a Week. The CA Chain Is What Breaks. Azure Front Door can now authenticate clients with certificates in production. But the client cert isn't what fails, it's the trust chain behind it, and most teams don't upload the full CA…
- Three Things That Break When You Shrink an Agent to Fit a Phone Most AI agents rent their capabilities from cloud models with billions of parameters. BotPsy runs the entire loop on your phone with no round trip, here's what survives the parameter cut and…
- I Ran Google’s Decision Model for a Week. When It Earns Its Keep One thousand questions, one API call, 150ms. Google's Decision 1 does one thing only, but it might replace the expensive model you're using for half your workload.
- One Brotli Decoder, Every Windows Build, and Which Systems Are Actually at Risk CVE-2026-111497 touches every Windows version, but 'every system that handles web traffic' quietly overstates your exposure. Here's how to sort your fleet by actual decompression risk and patch in the right order.
Topics
- Azure (58)
- Security (65)
- Identity (94)
- Microsoft 365 (88)
- AI (52)
- Apple at work (42)
- Windows (52)
- PowerShell and Linux (36)