Building with room to grow

Our company is named after a term from topology. An open set is a region where every point has a little space around it that still belongs to the set — no point is pinned to a hard edge, and the boundary is never quite included. It's a precise way of describing something very human: room to move.

We think the best software has that same quality. Not slack as in waste, but slack as in capacity to absorb change without breaking. The systems that age well are rarely the ones optimized to the millimetre on day one. They're the ones with a little air in the joints.

Where we deliberately leave room

"Room to grow" isn't a vibe; it shows up in concrete decisions. A few we make on almost every engagement:

• Seams where the system is likely to change — a model provider, a data source, a pricing rule — kept behind a boundary you can swap without surgery.
• Names and data shapes chosen for what the thing is, not just what today's feature needs, so next quarter's feature isn't fighting the schema.
• Evaluation and observability built in early, so you can change things later with evidence instead of nerves.

The opposite failure mode

The temptation, especially with AI, is to wire everything tightly to whatever works in the demo. It ships faster and feels great for about six weeks. Then the model changes, or the use case shifts, and the thing that was perfectly fitted to one moment has nowhere to go.

Tight systems are cheap to build and expensive to change. Open ones cost a little more up front and keep paying you back.

None of this means over-engineering. An open set isn't infinite — it has a boundary, just a soft one. The craft is knowing which edges to leave open and which to close. That judgment is most of what we're hired for.

If you're staring at a system that's outgrown the shape it was built in, that's exactly the kind of conversation we like to have.