RSI in Architecture

I remember sitting at my desk sketching a 100-story tower, my pen racing ahead of the code book. In those moments, the building felt effortless—pure geometry rising into the sky. But then came the call with the AHJ, the question of sprinkler systems, the reminder that two egress stairs and a core aren’t placeholders—they’re anchors. Without them, the tower collapses into drafts and revisions, and my client wonders why the dream keeps shifting shape.

That’s the first truth of architecture: where you place your stairs and elevators can decide the life of a building. Put them wrong, and every corridor, every square foot of rentable space, every dollar of value unravels. Developers see it as numbers on a spreadsheet—columns inside a unit reduce sellable square footage; a column pushed out increases value. But for architects, it’s more than math. It’s the heartbeat of the plan.

And yet, context pulls us further. A survey might reveal an easement, a traffic controller’s box, or an old tree the community refuses to lose. The AHJ may whisper “R-3 residential” when the investor was dreaming of commercial. Every site tells its truth, and ignoring it means wasted time, wasted trust.

Still, even as we wrestle with codes, columns, and corridors, another story runs in the background—a story hidden in the cloud. Every drawing we upload, every TikTok clip we watch, every model saved in BIM 360 lives not in thin air but in vast data centers, cooled by millions of gallons of water, powered by electricity that still leans too often on coal and gas. Posting an Instagram photo or generating an AI draft feels harmless, but each keystroke ripples into rivers, grids, and aquifers.

And here’s the paradox: the very tools straining our planet are the same tools reshaping architecture itself.

I once asked AI to generate construction documents. They were flawed, laughable even. But I was stunned—not because they worked, but because they almost worked. For the first time, I saw the tremor of what’s coming: the neat handoffs between Schematic Design, Design Development, and Construction Documents rattling like tectonic plates before a new continent forms.

This is Recursive Self-Improvement at the edges of our field. In AI theory, RSI is when a system improves itself, each upgrade accelerating the next. Architecture now mirrors that curve. Our scripts, plugins, and generative engines learn from our experiments, feeding discoveries back into the loop. Each test makes the next one sharper. We’re not just drawing buildings anymore—we’re building the tools that will one day build without us.

Some resist, clinging to the standards they trust. Others taste the future like a new drink they didn’t know they craved. But either way, the acceleration has begun.

And here I sit, a Martian at heart but an architect by training, caught between the tangible weight of stairs and sprinklers and the invisible weight of water in a desert data center. I sketch, I code, I post, knowing each action shapes not only the skyline but also the world that sustains it.

The question isn’t if architecture will change—it already is. The question is whether we can steer it, balancing the developer’s dollar, the occupant’s safety, and the planet’s fragile resources, while standing inside a recursive loop that is shaping us as much as we shape it.

Because once recursion begins, the story doesn’t just continue—it accelerates.

It accelerates to AGI, Artificial General Intelligence, and with it accelerates the consumption of water, electricity, and minerals—the hidden fuel behind the sleek glass towers and glowing screens. Every recursive loop of intelligence comes with a recursive loop of resource demand. The irony is almost unbearable: we architects spend our days sketching visions of sustainable futures—green roofs, net-zero towers, self-sufficient communities—yet the very tools that help us imagine them are quietly burning through rivers and watts in distant data centers.

It makes me wonder: how can we talk about sustainability if our methods are themselves unsustainable? How can we demand efficiency from the corridor, trimming inches to maximize daylight, while ignoring the gallons of water evaporated just to render the model? The footprint is everywhere—the light in our stairwells, the hum of cooling fans, the endless posts and uploads that stitch together the fabric of our digital lives.

And now AI, stitched deeper into the DNA of infrastructure, promises to accelerate design itself. It will give us endless variations, infinite experiments, recursive visions of towers and cities. But every new iteration consumes more power, more cooling, more hidden water. The very acceleration that excites us may undercut the ground we stand on.

So what is the “best” design, after all of this? Is it the high-rise optimized for rentable square footage, with its efficient core and maximized units? Is it the parametric wonder that dances with wind, or the digital twin that predicts every move? Or is it something much simpler—a treehouse?

A treehouse needs no data center. It breathes. It shades. Its structure is grown, not mined. It teaches us about scale, humility, and balance. Perhaps the lesson is not that we abandon technology, but that we remember the principles embodied in something as ancient as a tree: reciprocity, self-sufficiency, and lightness of touch.

Because maybe the future of architecture isn’t only about recursive self-improvement in silicon—it’s about recursive self-reflection in us.