Why Cabinet Design Software Doesn't Understand Plywood
General-purpose CAD tools treat a cabinet like arbitrary 3D geometry. A CNC woodworker needs something that thinks in panels, kerf, and sheet yield instead.
Design a kitchen's worth of cabinets in a general-purpose 3D tool and you'll hit the same wall every time: the software doesn't know it's building furniture. It knows geometry. Somewhere between the 3D model and the CNC file, a human has to do the actual cabinetmaking — translate shapes into sheet-good panels, account for the width of the saw blade at every joint, figure out how to nest parts on a 4x8 sheet without wasting a third of it, and decide how the joinery actually gets cut.
That translation step is where most of the real skill in cabinetmaking lives, and it's completely invisible to the software. I built KerfOS to close that gap — design in terms a woodworker actually thinks in, and let the tool handle the parts that are mechanical, not creative.
What "thinking in panels" actually means
A cabinet isn't a solid — it's a set of flat panels, joined at specific edges, cut from specific sheet goods, at a specific thickness. A design tool that treats it as arbitrary geometry has to be told, after the fact, how to decompose it back into that reality. KerfOS starts from the panel model directly: you're placing sides, shelves, and dividers with real material and thickness attached from the first click, not sculpting a shape and hoping it maps cleanly onto plywood later.
Kerf itself — the material a saw blade actually removes as it cuts — sounds like a rounding error until you've built something that doesn't fit because it wasn't accounted for. A dado sized for a 3/4" panel that doesn't account for a 1/8" blade kerf is a joint that's loose by a margin you can see. General CAD tools don't know your blade width exists. A tool built for this problem has to.
The part that actually took the longest
Sheet yield — how efficiently your cut parts nest onto a physical sheet of plywood — turned out to be the hardest problem to get right, not the joinery logic. Plywood isn't cheap, and a design that looks identical on screen can differ by an entire sheet in material cost depending on how the parts are packed. Getting a nesting algorithm that produces layouts a human would actually choose — not just mathematically optimal, but sensible for grain direction and offcut reuse — took far more iteration than the 3D modeling did.
The lesson in building a narrow tool
KerfOS is deliberately not a general CAD tool. It doesn't do furniture that isn't panel-and-sheet construction, and it never will. That's the trade I made on purpose: a tool this specific can make assumptions a general tool can't — about material, about joinery, about what a sheet-good project actually looks like — and those assumptions are what make it faster than modeling in something generic and figuring the rest out by hand.
The broader lesson generalizes past woodworking: the software that wins in a specific trade isn't the one with the most features. It's the one that already knows the constraints its user has been working around manually, and just handles them.