I sat down at my desk in the second week of January, opened my laptop, and started catching up.
Within twenty minutes I knew I had a problem.
Not that things had moved on without me. I expected that. The problem was the scale. Tools I’d been using in November had been superseded. Frameworks people were excited about in October were already yesterday’s news. The setup I’d spent weeks building, the one that finally felt stable, belonged to a different year.
Six weeks. That’s how long I’d been away.
A year earlier, in February 2025, I’d picked up Claude Code. It was still a research preview back then, a terminal tool for chatting with Claude about your codebase. Basic stuff: file editing, bash commands, git commits. I added a few MCP servers, some slash command plugins, a tool called GSD for spec driven development. It wasn’t everything, but it was enough. I’d landed. For the first time in the AI tooling explosion, I had a daily driver and a workflow that actually worked.
I was still exploring, still testing new things, but the exploration was optional. The foundation held.
Then two surgeries knocked me off the board.
The first one, in April and May, was annoying but manageable. I was back at my desk within a couple of weeks and picked up roughly where I’d left off. Things had moved a bit, but not enough to matter. I still knew what was what.
The second one, in December, was different. I came back after six weeks and the AI tooling world had been rebuilt from the ground up. Claude Code had gone through four major release eras while I was recovering. Microsoft had launched an agent framework. Google had released one too. The MCP ecosystem alone went from around 5,000 public tools to over 177,000 in fifteen months, according to a study by the UK’s AI Security Institute. That’s roughly one new tool every three minutes.
My mental map of what was current and what was stable wasn’t just outdated. It was from a different geological era.
Since then, I’ve been sprinting. Not building things, not shipping features, just trying to close the gap. And every time I think I’ve caught up, another wave hits. Sometimes several in a single day. A new agent framework drops. An existing tool gets a major update that changes how it works. Someone launches something that supposedly makes everything else obsolete.
There’s a term for this now. Harvard Business Review published a piece in March called “When Using AI Leads to ‘Brain Fry’”, based on a survey of nearly 1,500 workers. The finding: AI tools don’t simplify work, they intensify it. You go faster, but the track got longer. The promise was “AI as amplifier.” The reality is closer to “AI as accelerator.”
That matches my experience exactly, except I’d drop the academic framing and put it more simply: I’ve spent more time choosing AI tools than actually using them.
Here’s where it gets personal, and where I have to be honest about my own patterns.
I jump between projects. I always have. New thing catches my eye, I dive in, build something, learn something, then the next shiny thing appears and off I go. My wife could tell you stories. It’s not a secret. And a tooling space that releases something revolutionary every week feeds that tendency like nothing else.
The problem isn’t just that there are too many tools. It’s that the idea of “settling” on a stack has become a moving target. Settling in 2025 meant picking a tool and using it for six to twelve months. Settling now means picking a tool and knowing you’ll probably need to re-evaluate it in six weeks. The half-life of a stable development stack is shrinking faster than the tools can mature.
JavaScript developers went through something similar around 2015. New framework every month, new build tool every quarter, the same exhausting cycle of picking and abandoning. But that played out over years. This is playing out over weeks.
For now, I’ve settled. Hermes Agent for my multi-agent setup, GSD-pi for task management, oh my pi for quick terminal stuff. A Pi, a workflow, and the uneasy knowledge that it might not last.
Ask me again in six weeks. I might have a different answer. Or I might have the same feeling I had in January: that the conversation moved on without me, and all I can do is run to catch up.
Part of The Glitch Diary.