Don't Buy the New Buzzsaw

A little while back I wrote about the Sawdust Principle. Quick recap in case you missed it: every process spits out a main product and a pile of byproducts, and the smart move is to squeeze money out of those byproducts before you run off to start some shiny new business. A sawmill sells planks, sure, but it can also sell the sawdust it was going to sweep into the trash anyway.

Today I want to push that idea somewhere I didn't cover: it doesn't only apply to what your process produces. It applies to the tools you use to produce it.

The crash that started it

I was building some things on our CRM when one of our deploys crashed. Annoying, but fine, I hit deploy again. Crashed again. When I dug in, I found the culprit: we'd hit a hard limit on the number of "routes" our software would allow.

Quick translation for the non-engineers. A "route" is basically an address inside your app. Every page a user can visit, and every little behind-the-scenes request the app makes, has its own route, its own address. Think of it like rooms in a building. Our building had gotten so big it bumped into the maximum number of rooms our current hosting plan allowed. That cap was 2,048.

The shiny button

The obvious fix was sitting right there, glowing at me. Upgrade to the next plan. My current one runs about twenty bucks a month, and the next tier was somewhere around fifty. The exact numbers don't matter. The point is there was a big friendly "Upgrade" button promising to make the problem vanish in a single click.

And here's the thing: not long ago, I'd have clicked it without blinking. Because pre-AI, the alternative was brutal. Fixing this by hand meant a developer, me or someone I pay, digging through the whole codebase, carefully rewriting it to use fewer routes, then hunting down every new bug that rewrite introduced. Days of work. In terms of the time-value of money, the math was easy: just pay the fifty bucks and move on. Hand-coding your way out of a twenty-dollar problem is insane.

But it's not pre-AI anymore

So instead of clicking upgrade, I pointed Claude Code at our actual CRM build and said, in plain English: we're hitting the route limit, go through the code and condense the number of routes wherever you can, without breaking anything.

What it did next is called refactoring. Another quick translation: refactoring is cleaning up and reorganizing the guts of your code so it does the exact same thing, just more efficiently. Same building, same rooms your visitors actually care about, but you knock down the pointless interior walls and combine the redundant closets so the whole thing takes up less space. Nobody using it notices a thing. Under the hood, it's leaner.

In one quick pass, we cut our route count by 149. No upgrade, not an extra dime spent.

Let me put real numbers on the difference. Pre-AI, that refactor is easily two to three days of focused developer work, call it fifteen to twenty hours once you include tracking down the bugs it inevitably introduces. At real developer rates, that's a few thousand dollars of labor to dodge a thirty-dollar-a-month upgrade. No sane person does that trade. This time? I typed a few sentences to Claude Code, went to the gym, and by the time I got home it was done. That is the entire shift, in one sentence.

Sweat your tools

Now, I'll be honest with myself. This was a small pass, I'll probably have to do it again down the road, and eventually we really will outgrow this tool and upgrade for real. That's fine. The point isn't to duct-tape something together forever. The point is that we just bought ourselves a big chunk more runway on a tool we already have, almost for free.

That's the Sawdust Principle again, aimed at a new target. It's not only about monetizing byproducts. It's about wringing every last drop of value out of the equipment you already own before you go buy the shiny new buzzsaw. The mill owner who keeps his saw sharp and running three extra years beats the one who runs out to buy a brand-new one the second the blade gets a little dull.

Sam Altman said it well: the best startup founders are frugal, they get the absolute most out of every resource they have. This is exactly what he means. AI changed the math on this in a big way. A whole category of "just pay to upgrade" problems are now "spend twenty minutes and optimize" problems instead. So before you reach for the shiny new tool, ask whether you've actually squeezed the one already in your hand. More often than you'd think, there's plenty left in it.

Getting the most out of what you've got?

We help businesses wring real value out of the tools they already own. That's what we do at HyppoAI.

Build it right