Ai

The first time it worked, it felt like magic. I typed half a sentence, and the AI filled in the rest, tests, edge cases, even doc comments. Like a dream you didn’t know you had until it showed up in your IDE. But two weeks in, the cracks appeared. The AI rewrote a query using a non-existent ORM method. It silently introduced a regression in a caching layer. It suggested mocking the wrong dependency, and no one caught it, because the PR still “looked fine.” What started as magic began to feel more like auto-pilot without a destination. ...

I opened the integration guide to fix a minor typo and realized the steps didn’t work anymore. Nothing dramatic, just drift. A renamed flag here, a new default there, a missing permission no one remembered to mention. The words were still pretty. The reality had moved on. That’s the quiet failure of static docs, they age in place while the system keeps walking. So we tried something obvious we had somehow avoided: make the docs run. ...

At some point last month, I realized I hadn’t opened a “real” app in days. No calendar app. No separate to-do manager. No Word, no Excel, no IDE even, just a browser, a keyboard, and Notion with GPT whispering in the sidebar. Sometimes Obsidian. Sometimes ChatGPT itself. Always something that blurred the line between note-taking and system-building. It started as an experiment: could I run my whole digital life from just notes and AI? ...