Twenty Identical Cuts

Our staircase is open on one side. No wall, no railing — just a diagonal void between the living room and the steps. When we moved in, it felt like a design choice. After our daughter started crawling, it felt like an emergency.

The fix I wanted was a set of vertical oak slats running from a header at the ceiling down past the stairway, creating a divider. Tight enough spacing to be safe. Open enough to let light through. Clean vertical lines that make the staircase feel like an intentional architectural element rather than a hazard we patched.

I'd seen this done. I had the reference images. I knew what the finished thing should look like. What I did not know was how to actually build it.

The math problem

The header span is 67.5 inches. Each slat is 3/4 of an inch thick. I wanted 2.5 inches between slats, with 2.5 inches of open space at each end for symmetry.

This is arithmetic. It's also the kind of arithmetic where being off by a sixteenth of an inch on the first slat means being off by an inch on the last one, and every slat in between looks slightly wrong in a way you can feel but can't quite name.

I asked Claude to work out the spacing. It came back immediately: 20 slats at 3.25-inch repeating intervals (0.75 plus 2.5), with the first slat starting 2.5 inches from the edge. Perfect symmetry. The math landed cleanly — 20 slats exactly, no remainder, no awkward fractional gap to absorb.

Fine. But math is free. The hard part is transferring that math to wood.

The story stick

Claude recommended something I'd never heard of: a story stick. The concept is old — cabinetmakers have used them for centuries. You take a straight piece of scrap at least as long as your layout, mark the position of every element on it once with extreme care, then use that single stick to transfer all 20 positions to the header. One accurate measurement, twenty perfect copies. No tape measure, no cumulative error, no compounding mistakes.

The alternative — measuring each slat position individually from the end of the header — guarantees drift. Every time you read a tape and make a mark, you introduce error. Over 20 repetitions, those errors stack. By slat 15, you're hunting for fractions of an inch that should have been exact.

This is the kind of knowledge that lives in the hands of people who've built a lot of things and ruined a few. I haven't built a lot of things. The AI had never held a story stick. But it knew what one was, knew why it existed, and knew that it was the right answer to a problem I hadn't known how to articulate: "How do I make the same measurement twenty times without getting worse at it?"

The cut I couldn't picture

The top of each slat is a simple 90-degree cutout — it notches over the header and hangs. The bottom is where it got hard.

Each slat crosses a stringer that runs diagonally along the stair treads. Where the slat meets the stringer, it needs a compound notch: one surface sits flat on the top of the stringer, the other drops down along the outer face. The surface that sits on top of the stringer has to match the stringer's pitch angle — the slope of the staircase. Not 90 degrees. Not vertical. An angle I'd need to measure from the actual stringer and reproduce precisely on 20 boards.

I described this to Claude with a photo and a paragraph of halting prose. The response came back in two registers.

First, the technical version: "Set your table saw blade to the stringer pitch angle, position the slat so the angled cut removes material from the outer face while leaving the bearing surface parallel to stringer top. The sled fence keeps position consistent; the blade angle creates the pitch-matched surface."

I understood about half of that. Which is the problem with most woodworking instruction — it's written for people who already know what it means.

So I said: explain it as though I'm a noob with no jargon.

And it did. It described a guide system that holds each board in exactly the same spot so every cut comes out identical. It called the stop blocks "bookends." It said to take a small piece of wood, hold it against the stair stringer, trace the angle, cut it, and attach it to the guide — now when you tilt your saw blade to match that piece, it cuts the right angle automatically. It said the key trick is that after you get the first one right, you don't move anything, and every board after that will be identical.

Same information. Completely different accessibility. The first version was correct. The second version was usable.

What jigs are really about

The word "jig" sounds like a tool, but it's actually a decision. A jig is the decision to stop measuring and start indexing. To do the hard thinking once — where does the board go, what angle does the blade sit at, how deep is the cut — and then encode that thinking into a physical object that makes the answer repeatable without re-thinking.

This is what AI does in software: encode a solved problem so you can re-run it without re-solving it. The stair slat jig is the physical equivalent of a function. Inputs: a slat, placed against the stop blocks. Outputs: a slat with the correct notch. Side effects: sawdust.

The AI helped me design the jig I needed, but more than that, it helped me understand that a jig was what I needed. My instinct was to measure and cut each slat individually, test-fitting against the stringer, adjusting as I went. That approach feels careful. It's actually the riskiest path, because every adjustment is a new opportunity for error, and the visual consistency of 20 slats depends on them all being exactly the same, not individually "close enough."

Claude's response was blunt: stop hand-measuring each cutout position. Stop test-fitting individual slats before the batch is complete. Build the jig, validate the first piece, then run the batch. The metric is how many slats seat without shimming or re-cutting.

Fabrication is the hard part

There's a common assumption that design is the difficult phase of making things and fabrication is just execution. In software, that's often true — code is infinitely copyable, and deployment is (mostly) automated. In physical work, it's backwards. I can describe what I want a stair divider to look like in thirty seconds. The Pinterest image took three minutes to find. The spacing math took one conversation. The hard part was standing in my garage with 20 oak boards that were taller than me, figuring out how to feed them through a table saw without them tipping, binding, or kicking back.

The AI couldn't help me hold the boards. It couldn't feel whether the blade was bogging down in the cut. It couldn't hear the change in pitch that means the wood is pinching. But it could tell me to set up a sawhorse at the outfeed end for support. It could tell me to process all the top cuts first, then change the blade angle once and do all the bottom cuts, rather than switching back and forth twenty times. It could tell me to use the actual stringer as a template to set the bevel angle, rather than trying to measure the angle with a protractor and hoping I read it right.

These are sequencing decisions, not design decisions. They're the kind of thing a more experienced builder would tell you while leaning against the workbench, watching you set up your first cut. The advice doesn't require genius. It requires having done it before, or having access to someone who has.

What the slats do now

They hang in a clean vertical rhythm from ceiling to stair. Light comes through. The staircase is enclosed without feeling closed. Our daughter can't fit between them, which was the original requirement, but they also do something I didn't anticipate: they make the staircase visible from the living room as a pattern of shadow and light that changes through the day. In the morning, the sun comes through at a low angle and casts twenty parallel lines across the floor. By afternoon, the shadows merge. At night, the LED strips I routed into the stringer behind the slats glow through the gaps.

None of that was in the brief. The brief was "keep the baby off the stairs." Everything else came from doing the work, making the cuts, and discovering that twenty identical pieces of oak, spaced with precision, do something beautiful when the geometry is right.

The AI didn't make the stair divider beautiful. I didn't either. The math did — the same math that made the spacing safe made it look like it belonged. Twenty identical cuts, twenty identical gaps, and the light figured out the rest.