Data Custodian

I asked an assistant to run the same workflow it ran for me last week. It failed the exact same way. Same missing edge case, same wrong assumption, same fix I'd already explained once, out loud, in plain English. I hadn't told it to remember the fix. I assumed it would, because that's the story everyone's been sold. AI that gets smarter the more you use it. AI that learns your preferences. AI that improves with every conversation, like a junior employee who eventually stops asking where the stapler is. "Here's the part that's going to make some people mad: none of that is true, and it was never close to true." Anthropic calls them Skills. OpenAI has custom instructions and GPTs. Other platforms call them workflows, playbooks, automations. The label changes. The architecture underneath does not. Every one of them is a text file. Instructions written down once, read fresh every single time, with zero memory of the last time they failed you. The file does...

My coworker Bri posted a prompt in one of our Slack channels back in August. About fifty lines in a code block, a sunshine emoji, and the line that did the real work: feel free to borrow the instructions to create your own. I borrowed it that afternoon. The design instincts in it are good ones, and they are the reason I did not start over. It ranks work against her actual goals rather than by due date. It links the doc you need so the list is one click from the thing it is about. It runs on a schedule, so it shows up whether or not you remember to ask. Those three calls are the hard part, and she got all three right. She had also gone to the trouble of having Claude rewrite her version to be general so anyone could pick it up, which is a generous thing to do with something you built for yourself. Then I pointed it at my Tuesday. The problem was my week, not her prompt A recipe assumes a working kitchen. My kitchen is a mess. Four back-to-back meetings, an approval nobody has...

Saturday morning, the CNC is running a pocket for an inlay and I am eight feet away with a chisel, cleaning up the corners of a mitered box the machine will never touch. Two tools. Same project. Nobody in the shop thinks this is a contradiction. But post a picture of it online and someone will show up in the comments within an hour to tell you it does not count. That a real woodworker cuts everything by hand. That the machine did the work. I used to have opinions about this. Now I mostly just think about who is going to sand it. The question people are actually asking The comment is never really about the machine. It is about whether you did anything. And it is the exact same comment I hear at work about AI, delivered by people in much nicer clothing. Did you write that or did the model write it. Is this your thinking or is it borrowed. Does it count. Both times, the question is aimed at the wrong end of the process. The machine did not decide the box should be ...

 I was standing in the kitchen yelling "DYLAN" up the stairs for the third time when it hit me. Somewhere in my house there is a Raspberry Pi, a touchscreen, and a Home Assistant instance with more sensors than most small businesses run in production. And I am still using my voice as the primary notification system. That is not engineering. That is a man shouting into the void and hoping for the best. So I fixed it. I built four buttons on the kitchen wall dashboard, one for each person in the house. Tap a name, the phone alarm screams until someone finds it. No more stair-yelling. No more "has anyone seen my phone" turning into a twenty-minute search of the couch cushions. Here's the part that actually matters, though. I didn't write the YAML myself. I described what I wanted, AI wrote the config, and the whole thing went from idea to working button faster than it would have taken me to remember the correct Home Assistant action syntax from memory. Mushr...

The invoice landed and nobody wanted to open it. Not because the number was surprising. Because it confirmed what everyone in the room had suspected for months: the platform that enterprise IT had built its entire operational model around had just repriced itself out of the relationship. No negotiation. No grandfathering. New terms, new math, and a renewal window that didn't leave much room for alternatives. That moment — repeated across thousands of enterprise environments between 2023 and 2024 — was not a licensing dispute. It was a stress test. And a lot of stacks failed it. 59% of VMware customers saw prices rise by more than 25% after the Broadcom acquisition, with the median increase landing between 25% and 49%. Extreme cases hit 350% to 700%. Gartner's Peer Community survey found 74% of IT leaders actively exploring VMware alternatives. And yet most of them haven't moved. 86% of organizations are actively reducing their VMware footprints, but just over half plan phas...

The meeting was supposed to be about the next three-year refresh cycle. Standard stuff. Servers, storage, networking, the usual debate about what stays on-prem versus what moves to cloud. The facilities manager walked in halfway through, sat down, and put one number on the table: 18 months of power headroom left at current growth rates. Nobody was ready for that conversation. Not the architects. Not procurement. Not the infrastructure lead who had been running that environment for nine years. Everyone had been planning for compute and capacity. The building itself was the constraint, and nobody had thought to ask. That's where enterprise IT is right now. The problems stopped being theoretical. Multiple forcing functions are converging simultaneously. Power. Cooling. AI workloads. Virtualization economics. Cloud cost reality. Skills attrition. Each one is a serious problem on its own. Together, they're a stress test most infrastructure teams are not built to pass. Average rack d...