Journal · The AI Architect · 2026-07-25
Save a good version before you change it. Label the moment — "works, before the redesign" — so you can step back.
Save a good version before you change it. Label the moment — “works, before the redesign” — so you can step back.
The tension isn’t whether to save your work. Everyone knows to do that. The tension is that saving a version feels like admitting you might fail at the next one — and so we skip it, half-consciously, because starting the redesign with total confidence feels better than starting it with a safety net. The net implies you might fall.
But the net isn’t there because you’re weak. It’s there because good building requires a kind of experimentation that only works if going backward is cheap. When going backward is expensive, you stop experimenting. You get cautious in ways that don’t look like caution — you get rigid, defensive, unwilling to touch things that actually need touching. The save isn’t a hedge against failure. It’s what makes bold changes possible in the first place.
Naming the moment is the real work
Anyone can hit save. The harder habit is naming what you’re saving and why. “Works, before the redesign” does something a filename like “draft_v3_final” doesn’t: it tells your future self what was true at this point and what’s about to change. It’s a note passed forward in time.
Without that note, old versions pile up as noise. You end up with a folder of ghosts — v1, v2, v2_edit, v2_edit_real — and no memory of which one actually worked and which one you were in the middle of breaking. The labeling is what turns a backup into a landmark. It’s the difference between having a map and having a pile of old maps.
Try this: next time you’re about to make a change you’re not fully sure about, stop and write one sentence describing the current state — not the file name, the actual state. “Onboarding flow works, takes four steps, users complain about step two.” Save that sentence next to the version. You’re not just protecting the code or the document; you’re protecting your own understanding of it, which decays faster than the thing itself.
Stepping back requires a place to step back to
There’s a reason “step back” is the phrase that matters here, more than “save.” Stepping back is a physical metaphor for a mental move: getting enough distance to judge whether the new direction is actually better, or just different, or just familiar because you’ve been staring at it for three hours.
That judgment is nearly impossible to make from inside the change. You’re too close, too invested, too aware of everything you just did to fix the last problem. The saved version gives you an outside point to compare against — a fixed thing that isn’t shifting under the pressure of your current effort. Without it, “better” becomes a feeling instead of a comparison, and feelings about our own recent work are notoriously unreliable, usually too harsh or too generous depending on the day.
Try this: when you’re deep in a redesign and start to feel unsure whether it’s working, don’t just look at the new version harder. Go back and open the saved one. Sit with it for a minute. Ask plainly: is the new one actually solving something this one didn’t? If you can’t answer clearly, that’s useful information too — it might mean the redesign is driven by restlessness rather than a real problem.
The discipline is emotional, not technical
Version control tools make the mechanics trivial. Commits, branches, snapshots — the infrastructure has existed for decades. What’s missing is rarely the tool. It’s the pause before the change, the moment where you decide this version is worth marking before you touch it.
That pause is emotional labor disguised as a technical step. It asks you to admit, mid-momentum, that the thing you built might be worth preserving exactly as it is — and that the thing you’re about to do might not be an improvement. Most of us would rather charge ahead than sit with that uncertainty for the ten seconds it takes to save and label.
But the builders who last are the ones who’ve made peace with that pause. They don’t treat it as doubt. They treat it as ordinary maintenance, like locking a door — not because they expect a break-in, but because it costs nothing and saves everything if one happens.
I write about this more in the book, in the context of a few habits that don’t look like much on their own but change how sustainable your work becomes over time. This is just one piece of that. The habit is small. Its absence is what gets expensive.
Go deeper. The full method is in The AI Architect. New here? Start with the free companion pack, or explore the series.