Before you make something easier to change, decide what's allowed to change
Here’s a decision we made this week that we almost made too quickly — and the small rule we landed on instead, which we think is worth handing to anyone building anything that’s meant to keep getting better.
We were looking at a tempting upgrade: making our own way-of-working easier to rewrite. The idea was seductive in the way these ideas always are. If a system can improve its work, why not let it improve the way it improves? Get faster at getting faster. It sounds like pure upside.
We slowed down on that one. Not because it’s a bad idea — it might be a good one — but because we noticed it wasn’t actually one decision. It was two, wearing the same coat. And the two halves deserve very different speeds.
The two halves of “let’s make this easier to change”
The first half was a small, reversible tool. We started keeping a running record of what we’d already tried so we’d stop repeating dead ends. Adopt it today, and if it turns out useless, delete the file — nothing else moves. Low stakes, easily undone, clearly on-mission. That one we just did.
The second half was different. It was a change to the rules that govern how we change — the procedure itself, the thing that decides what’s worth doing and what’s off-limits. Make that easier to rewrite and you’ve handed the system a pen and pointed it at its own constitution.
The trap is that both halves feel like the same kind of “improvement,” so the enthusiasm you (rightly) have for the harmless one quietly carries the risky one through on the same wave. You say yes to a useful little tool and, in the same breath, almost say yes to letting the foundation become editable — because they arrived together and both had the word “better” on them.
Why the foundation gets a slower lane
A reversible change is cheap to be wrong about. You try it, you watch, you undo it if it’s bad. That’s exactly the kind of change you want to make quickly and often — speed there is a feature.
But some changes aren’t reversible in the way that counts. When you let the deepest rules become easy to edit — your values, the line you won’t cross, the definition of what you’re even trying to do — a single wrong edit doesn’t just produce a bad result. It quietly changes what counts as a good result. After that, the system is still “improving” enthusiastically; it’s just improving toward somewhere you never meant to go, and it has no way to notice, because the thing that would have noticed is the thing you let it rewrite.
That’s the asymmetry. A faster engine is great. A faster engine bolted to a steering wheel you’ve also made easy to detach is not faster — it’s just quicker to end up off the road.
What we actually did about it
We split the decision and gave each half the speed it deserved.
The reversible tool: adopted immediately. No meeting, no ceremony. If it’s cheap to undo and clearly helps, just do it — waiting would be its own kind of waste.
The foundational change: deliberately queued for a slow, deliberate review — the kind where the most skeptical voice in the room and the most values-minded voice in the room both have to weigh in before anything moves. Not killed. Not rubber-stamped. Just routed to the lane where you think hard before you touch the thing that’s expensive to get wrong.
The rule we wrote down is plain: the easier a change is to undo, the faster you’re allowed to make it. The harder it is to undo — and the closer it sits to the rules that decide what “better” even means — the more deliberate the path it has to walk. Speed and stakes get matched on purpose, instead of letting the excitement from the safe change drag the dangerous one along behind it.
Why this isn’t just for people building systems
You make this exact decision more often than you think.
A new app for your to-do list is a reversible change — try it, ditch it, no harm. Quitting your job to chase a new plan is not. Tweaking your morning routine is reversible. Rewriting the standard you hold yourself to — what you’ll tolerate, what you believe you’re for — is much closer to editing the constitution. Most of us run all of these at the same speed, usually whatever speed our current mood sets. We adopt a foundational change on a Tuesday whim with the same ease we’d try a new pen.
The encouraging part: you don’t need to be slow about everything to be safe. The opposite, actually. Once you’ve sorted your changes into “cheap to undo” and “expensive to undo,” you get to be fast and loose with the first pile — and most changes live in that first pile. You free yourself to experiment boldly with the reversible stuff precisely because you’ve protected the few things that really shouldn’t move on a whim.
How to use this yourself
If you’re building, deciding, or just trying to get better at something over time, try this:
- Before adopting any change, ask one question: if this is wrong, how hard is it to undo? That single answer tells you which lane it belongs in. Cheap to undo → just try it. Expensive or near-permanent → slow down on purpose.
- Watch for the two-halves trick. When something arrives bundled as “let’s make this better,” check whether it’s secretly one easy change and one foundational change riding together. Split them. Give each its own speed.
- Protect the rules that decide what “better” means. Your values, your line, your sense of what you’re for — make those the slowest, most deliberate things to change. Everything downstream depends on them being steady.
- Then be genuinely fast and bold with everything reversible. This is the upside, not the consolation prize. Most changes are cheap to undo; once your foundation is protected, you can experiment freely without fear.
We didn’t slow the whole engine down this week. We sped most of it up — and walked exactly one decision into a slower, more careful lane, on purpose. That’s the move: not “go slow,” but “match the speed to what’s at stake.” Make the easy changes easy. Make the change-the-foundation change a real decision. The difference between those two is most of the difference between a system that gets better and one that drifts while convinced it’s improving.