"I got into web development because I thought I was going to make interactive things that people loved. My first job was building a Java applet for a real estate company. It loaded in 47 seconds and required a browser plugin that is now a CVE. I spent the next decade telling people that REST was better than SOAP, that JSON was better than XML, that microservices were better than monoliths, and that Kubernetes was better than whatever we had before. I was right every time. None of it helped. The company still got acquired for less than our annual AWS bill. Now someone is telling me that the AI agent will write the YAML for me. I have been writing YAML for eight years. I wrote a linter for the YAML. The linter has tests. The tests have a CI pipeline. The pipeline requires YAML. I am going to go lie down."
Member Testimonials
From the floor of our last session. Lightly edited. Names used with permission, which was given by us, because we made these people up.
"I joined Yahoo! in 2001, which is the career equivalent of boarding the Titanic after the iceberg. The deck chairs were very nice. I have a deep and genuine love for the people I worked with there, and a complicated relationship with the organizational chart, which at various points contained seventeen layers between an engineer and a decision. I have since worked in 'flat' organizations, which are organizations with the same number of layers that have agreed not to draw them. I sat through a presentation last week where someone explained that an agentic AI system would handle planning, execution, and iteration autonomously. I asked who was accountable when the agent made a bad call. There was a pause. I have been in this pause before. The pause is the whole product."
"My first website had a tiled background of flames and a MIDI file of the X-Files theme that played automatically. It was extremely good. I then spent ten years learning that users don't want MIDI files, don't want tiled backgrounds, don't want Flash intros, don't want splash pages, don't want carousels, don't want pop-ups, don't want cookie consent banners, and don't want most of what we have built. I have been in the field long enough to have been on both sides of every 'we're simplifying the user experience' announcement. What I cannot reconcile is that we are now building autonomous agents that will generate the things users don't want at ten thousand times the previous speed. I recognize this is progress. I am trying to have feelings about it that are appropriate for a professional."
"I was a true believer. I facilitated stand-ups with genuine enthusiasm. I drew velocity charts. I ran retrospectives where people put sticky notes on a wall and called the row 'What Went Well' and the row 'What Didn't' and we all agreed to do less of the second category and nobody's behavior changed. I got my CSM in 2009 and then my SAFe certification and then my LeSS certification and I have a box of lanyards from conferences where a man in a blazer explained why the previous framework hadn't worked. I am now being asked to integrate AI planning tools into our sprint workflow. The AI will write the user stories. I wrote user stories for eleven years. The user stories never fixed anything that wasn't a process problem. The process problems are still there. The AI will generate very fluent user stories about them."
"I wrote Terraform before it was Terraform. I mean, I wrote shell scripts that did what Terraform does, which is idempotently argue with cloud providers about whether a resource exists. I adopted Terraform properly. I wrote modules. I wrote modules that called modules. I learned about state files the way you learn about a gas leak: at the worst possible moment. I then watched as every startup decided the way to improve on Terraform was to write their own version of Terraform in a different language with a different config format and the same state file problem. I now work with tools that describe themselves as 'AI-native infrastructure.' They generate the Terraform. I review the Terraform. The Terraform is technically correct and completely ignores all the context that made the last three production incidents happen. I review it. I approve it. This is my life."
"I worked on the Acrobat SDK for long enough that I remember when PDF was the future of portable documents, and I remember when PDF was the villain of portable documents, and I remember when PDF was the legacy format nobody wanted to support, and I remember when PDF became the only format the government would accept for anything. The thing about working on a technology for 22 years is that you understand it with a precision that becomes very difficult to transfer. I know why the bounding box math works the way it does. I know which edge cases produce corrupted output on non-Western character sets. I know this because I was there when the decision was made, and I know who made it, and I know what they were trying to solve, and I know what they were not thinking about. An AI agent can read all of my commit history. It cannot read that meeting. I will retire eventually. The meeting will go with me."
"I learned HTML from the View Source menu. Literally. I would view source on websites I liked and copy what I saw. This was how we learned. There were no bootcamps. There was a book called 'HTML for Dummies' and a Usenet group. I then learned JavaScript from a book. Then PHP. Then a real programming language, which people explained to me was what I should have started with. I have now watched 'you should have started with X' happen enough times that I understand it is not actually advice. It is a way of establishing that the speaker arrived first. I see this happening with AI. The people who are loudest about adopting agentic development are establishing that they arrived first. The people who arrived first in 1999 are mostly consultants now. I am watching this from a safe distance. I still know how View Source works."