This is also the story behind my current profile name.

Earlier in my career I was stuck maintaining another developer's nightmare Excel/VBA code. If this had existed I would not be a developer now.

Floss picks are the biggest waste of money. They're everywhere. All you have to do is sterilize them.

Today I feel like I'm flying. I'm doing complex work, but I only need to set direction and make certain choices. Those choices are important and valuable. They are why I get good results. I'm earning my paycheck, but I'm thinking and answering, not typing code. It's amazing.

What an amazing time to be a developer. I'm starting things I'm not sure I can finish, and as I work the technology advances, extending my runway before I hit the end.

Everything is not political. If it was, we wouldn't need a word for "political." We would just call it "everything."

What's the story behind LLMs doing crazy things like deleting databases? What are people telling it? Is it something like "PLEASE I need to this to work, make it happen, do anything, I don't care, just MAKE IT WORK!!!"

I'm trying to imagine telling someone, "We're not going to consider you because you brought iced coffee to the interview." It sounds silly. But if that's the reason, don't you have to stand by it and say the words?

There are no "bad ideas," only things we shouldn't do because they don't make sense.

I drew the line with our outgoing Product Owner: If AI wrote the ticket, I won't discuss it. Delete it. It's okay if someone uses a tool to organize their own thoughts. But he was letting Claude create work for us. He didn't even realize that the decisions were technical.

I reflected more on this and realized that my concern is - at the risk of being harsh and simplistic - what will happen to stupid people. If you're inclined to learn, AI delivers more of it. It's not the same. Maybe you won't bother learning some things because AI knows them so well. And you're going to get AI's version. But we can interrogate that. What is the mechanism by which varying viewpoints and information feed back into models? How will they decide what is accurate, or how to position it? If we depend on them more, how great is the risk that they are trained by output from earlier versions of themselves? If you're (again, being simplistic) stupid, you were probably going to be stupid anyway. Is it better that AI thinks more? Does that make some humans particularly subservient? That's too simplistic. I don't want to be mean or think too much of myself. Most of us are smart about some things and stupid about others.

What will happen to the generation that grows up with AI? Imagine a parent who answers every question, solves every problem, and tells you what to do, but never asks you to figure it out or expects you to learn. And the robots aren't even here yet.

I land at DAL airport and call an @Uber from the baggage claim. The pickup is in a place to which no signs point. Uber shows me a picture to help. It is of a place you cannot see without walking minutes from where you are. Every airport sends you to some mystery location instead of the spot 50 feet away where you can see cars picking up passengers, just like what an Uber does. TPA sends you to Blue Express which only accessible two floors up from baggage claim. Just like at DAL, there are no signs near baggage claim. You only see signs if you randomly guess that you should take the escalator to ticketing. I blame Uber for making me select those pickup locations. But at the very least, give me clear directions. Don't show me a dotted line which could be in any direction so I have to start walking to see if I get closer or further.

I have "new C# questions" on Stack Overflow bookmarked. I'd check it once in a while and scroll through the questions looking for anything interesting. I've answered over 1,000. There are now about two questions per day.

My relationship with my previous Product Owner felt like a role-reversal. I'm the developer. I'm supposed to chase cool tech and frameworks with total disregard for the business. The Product Owner is supposed to rein me in. It was the opposite. My PO wrote tickets describing APIs, database schemas, Kafka messages, frameworks, and testing, none of which he understood at all. Over the course of two years I tried to pin down the reasons and the desired outcomes. His responses were 100% word salad as his voice descended into a nervous, irritated vocal fry. He simply wasn't capable of grasping that work needed to produce outcomes, or how to accomplish that. This culminated in our manager completely going around him to find valuable work for the team. Even then he derailed us for a month by writing tickets to create a testing framework he didn't understand for an application he didn't understand. He didn't even write the tickets. He pointed Claude at the repo and told it to make stuff up. When someone adds such negative value, I begin to question how they collect a paycheck at all, or why they're granted a badge to access the building. We would be further ahead if we fired him, withdrew his salary in cash, and burned it every two weeks. He was promoted, but he still shows up in our meetings asking stupid questions and finding ways to interfere.

Thoughts and feelings are not facts. That doesn't diminish them. They're real, and they're a huge part of what we are. They're important. Be glad they're not facts. Facts we don't like don't change. Thoughts and feelings can.

This gave me a crazy epiphany. It's REST for AI. Not fake REST, but real REST, where you call an endpoint and it returns information about a thing and instructions for what else you can do with that thing. If you get a list of orders, you can ask for more information about one, and that information also tells you what else you can do, like cancel it, change shipping, request status, etc. We hear about REST, but I don't think most of us have ever actually implemented or used it. I haven't. We think anything with HTTP verbs is REST. One of the premises of REST is that given no knowledge of what operations are available, you can generate a UI in real time based on the result of a single request. I'm sure someone, somewhere does that. I've never seen it. It's not how we write code. But maybe it makes sense as a way for agents to interact with systems. I have to process this.

AI agents seem to play well with hexagonal/ports-and-adapters architecture. An agent is a type of user. It interacts with other systems but remains separate from them. It may execute the very same use cases as a user interacting through a UI, or some other external system.