" We shape our tools, and thereafter our tools shape us. "

Tropical Modernism

In the 1980s, Geoffrey Bawa's architectural studio in Sri Lanka did something unusual: they drew trees on architectural plans freehand with ink pens. This practice was central to birthing the architectural movement that dominates East Asia to this day: 'tropical modernism.

Every other modernist architect of the time used circles and triangles for flora, abstract symbols, location markers on a plot, efficient, universal and clinical. Bawa's draughtsman, Laki Senanayake, spent hours with traditional ink pens that produced lines of varying thickness. Their intention  was to  "render with botanical authenticity." plans that embodied the character of leaves and the way branches moved.

The tool choice mattered. A ruling pen typical for plan drawings produces uniform lines, perfect for walls and columns. But an ink pen with variable line weight could capture organic form. The thick-to-thin strokes followed the growth patterns of leaves. The tools here enforced observation.

You can't design a building's relationship with a landscape if your drawing method treats that landscape as empty space. By drawing vegetation with the same care as walls and columns, Bawa's office made it impossible to ignore the site.

Cross section and plan of Ena De Silva house 1984

The tools allowed for technique that enforced thinking.

Tools enable creation through constraints, and constraints shape what becomes thinkable. Tropical modernism was born out of a drawing technique that a particular tool enabled that made certain kinds of thinking possible. The buildings changed because the drawings changed because the tools changed.

Learning Through Material Constraints

In architecture school, I drew constantly. I developed preferences: pencil for early exploration when I needed to stay loose, pen when I was ready to commit. Each tool let me think differently about the space I was designing, though I had no vocabulary for why this mattered.



Two farmers inspecting a plant circa 2019

When I discovered Bawa's drawings years later, I understood my own tool choices differently. I realised I had been selecting what kinds of thinking would be possible. The pencil's erasability kept options open, allowing tentative exploration. The pen's permanence forced commitment, collapsing possibilities into decisions. Both were constraints, and both were productive.

You don't think "I'm using a pencil." You think "I'm drawing a line." This is material mediation: the resistance of paper, the friction of graphite, the weight in your hand. Physical constraints you can learn, predict, and embody until the tool becomes transparent, an extension of your body.

Recently I drew on an iPad with Procreate and felt this transparency break. The line appeared smoother than what my hand had done, the algorithm interpreted my trembling stroke as noise and optimized it into a clean spline. My hand learned to propose rather than define, to suggest knowing the system would refine. When I returned to pencil, the uninterpreted line felt strange. With the Ipad I was no longer using a tool, I was negotiating with one.

Bezier curve applied to vector paths

From rules to intelligence

Procreate's smoothing applies geometric principles, universal mathematical rules about what counts as a "clean" line. But AI-mediated tools represent something qualitatively different. When the system has been trained on vast corpora of images and gestures, it doesn't just optimize your line geometrically, it interprets what kind of mark you're trying to make based on learned patterns from millions of examples. The algorithm brings probabilistic judgment shaped by its training data.

The common concern is that this represents a loss of agency. I see it differently: it's a novel mode of creative constraint, one where you're no longer working against material resistance but negotiating with learned patterns. The question becomes not whether to accept this mediation, but what becomes possible when you work with it deliberately.

The Question Worth Exploring

What becomes possible with AI-mediated constraints that wasn't before?

When the tool doesn't only resist but interprets, when it brings learned patterns and probabilistic judgment into the creative loop, what forms of thinking emerge?

What can you draw, design, or build in partnership with computational intelligence that you couldn't conceive working against material resistance alone?

What if you drew but never saw your own strokes, only the AI's interpretation? What if you coded by voice while AI watched your screen, read your errors, saw what you see? What if your website generated software in real time, no predetermined functions, no prewritten code, only intelligence shaped by each interaction?

When any tool we use to think and create interprets our input through learned patterns, it shapes what we can imagine and build. The question is what becomes possible when we work with that constraint deliberately.

Constraints are generative

Bawa's office chose ink pens deliberately. They decided freehand botanical rendering would structure their thinking, and in doing so, they created an architectural movement. The constraint was accepted, explored, pushed and it opened possibilities that ruling pens and abstract circles had foreclosed.

We're at a similar moment with computational mediation. These tools offer genuinely new constraints: interpretation, learned patterns, probabilistic judgment in the creative loop. The only way to know what they enable is to work within them, to develop fluency with them, to see what emerges when you stop fighting the algorithm and start thinking with it.

Flora AI an AI-powered 'infinite canvas' for creative professionals

Not every constraint is worth accepting. But dismissing computational mediation because it's unfamiliar, because it mediates differently than material tools, means foreclosing possibilities before we've explored them. Bawa didn't know tropical modernism was possible until his office committed to the constraint of botanical authenticity.

We won't know what's possible with computationally-mediated tools, from algorithmic smoothing to AI interpretation, until we commit to exploring their constraints seriously. I'm not sure yet what language I'll need for what these experiments reveal. That's the point. We're developing genuinely new forms of thinking-through-making, not better than previous modes, not worse, but different in kind. The only way to understand them is to work within them.

Websim an AI-powered browser that hallucinates your query.

The tools we use shape what we create. This need not be read as a warning, it can be an invitation. The question is what we'll make with these new tools.

References

1. Abbas, S. M. A., Aslam, M., Baleanu, D., Alzabut, J., & Agarwal, P. (2024). An optimal B-spline approach for vectorizing raster image outlines. PLoS One, 19(1). https://doi.org/10.1371/journal.pone.0296496

2. Bergson, H. (1913/2001). Time and Free Will: An Essay on the Immediate Data of Consciousness. Dover Publications.

3. Dominici, E. A., Schertler, N., Griffin, J., Hahmann, S., Sheffer, A., & Sharf, A. (2020). PolyFit: perception-aligned vectorization of raster clip-art via intermediate polygonal fitting. ACM Transactions on Graphics, 39(1), 1-16. https://doi.org/10.1145/3144456

4. Guerlac, S. (2006). Thinking in Time: An Introduction to Henri Bergson. Cornell University Press.

5. Hespanhol, L. (2023). Human-computer intra-action: A relational approach to digital media and technologies. Frontiers in Computer Science, 5, Article 1083800. https://doi.org/10.3389/fcomp.2023.1083800

6. Ihde, D. (2002). Bodies in Technology. University of Minnesota Press.

7. Levien, R. L. (2009). From Spiral to Spline: Optimal Techniques in Interactive Curve Design [Doctoral dissertation, University of California, Berkeley]. https://www2.eecs.berkeley.edu/Pubs/TechRpts/2009/EECS-2009-28.pdf

8. Verbeek, P.-P. (2015). Beyond interaction: A short introduction to mediation theory. Interactions, 22(3), 26-31. https://doi.org/10.1145/2751314

9. https://x.com/pieratt/status/1833701511895154828