I built my first webpage in middle school back in 1998, hand-typing raw HTML into Notepad before CSS was standard. Over the next two and a half decades I watched the web move from table-sliced Photoshop layouts and early PHP scripts to modern component frameworks and distributed services.

Through every shift I kept coming back to the same habit: break things into smaller pieces. Capped files early, watch out when lines go beyond 500. Keep to single-purpose composables, simple two-stage boolean logic. It wasn't a rule I picked up from a book. It's just how I kept my own head straight. Once a file got big, I lost track of what depended on what, and I'd get slower and more careful instead of faster. Smaller pieces meant I could hold the whole thing in my head. For years some engineering directors and colleagues told me this was overkill for a human team.

Then the vibe coding shift happened, and I started handing real work to Claude, Gemini, GPT, Grok, and Cursor.

What I noticed felt familiar. Agents working against big files lost the thread the same way I used to. Missing imports, duplicated hooks, edits that touched more than they should have. When I fed the same kind of work into small, single-purpose files instead, the agents seemed to stay on track more reliably, and for the kind of narrow, targeted edits I do most often, it usually meant less back-and-forth to get there. I don't have a lab and a control group behind that. It's a pattern I noticed in my own day-to-day work, not a published study. But it held up enough times that I started writing it down as a set of rules for myself.

That's what Chemical X is. Not a discovery that AI needs this. Twenty-five years of a habit that turned out to carry over once the reader of my code stopped being just me.

My 25-year obsession wasn't overkill. It's the missing ingredient for a super powered codebase: Chemical X.