It turns out, my obsession with LEGO isn’t just nostalgia; it’s neuroscience.
Like many with aphantasia, I can't picture things in my mind. No mental images. No visual recall. Ask me to imagine a Minimoog synthesizer, and I can't summon a single knob or panel. But put LEGO bricks in my hands? I can build a stunningly accurate 1:7 scale replica.
This contradiction used to baffle me and my family-until Gen-AI helped me understand it.
The Missing Mind’s Eye: Understanding Aphantasia
Aphantasia, coined in 2015, describes a lack of voluntary mental imagery. Studies estimate 3-4% of people have it. We don’t picture scenes, faces, or objects in our heads. Some of us can dream visually, others can't. Some can recall what things are without knowing what they look like.
But here’s the twist: aphantasia doesn’t mean poor spatial thinking. In fact, recent studies show we often excel at spatial accuracy while using analytic, rule-based strategies instead of imagery.
Task: Mentally rotate a 3D “L” shape to determine if it matches another A visual thinker might:
An aphantasic using analytic, rule-based strategies would:
They don’t see it, they reason through it step-by-step, like solving a Rubik’s cube via moves, not intuition. It’s like using a set of instructions or a decision tree, rather than a mental picture. |
In a 2024 study, only 24% of aphantasics used holistic mental-rotation strategies. Most of us deconstruct the problem into pieces-literally. Think rules, lists, coordinates. Not scenes or snapshots.
Exploded Diagrams: External Imagination
That’s why LEGO diagrams feel like home. You know those exploded-view instructions where every brick floats in midair before snapping into place? That’s not just a build guide-it’s my imagination, projected.
When I recreate a Minimoog or a Mellotron:
- I count knobs and panels, like tokens.
- I translate angles into hinge types.
- I map color ratios and spatial zones.
No visual recall, just systematic deconstruction. GPT-o3 helped me frame this:
“You’re using LEGO like an AI uses tokens-breaking reality into rule-based components instead of images.”
That moment hit hard. AI doesn’t understand what it sees-it builds structure from relationships. So do I. My creative process isn't visual; it’s relational.
Spatial Mastery Without Pictures
LEGO play isn’t just child’s play. In 2023, a team at the University of Surrey asked 358 eight-year-olds to rebuild exploded LEGO diagrams. Success predicted stronger maths scores, but only when the kids manipulated physical bricks. The researchers traced the boost to specific spatial sub-skills: mental rotation, disembedding tiny shapes, re-scaling parts.
Aphantasics often outperform visualizers in accuracy on these tasks, though we're slower. We use deliberate, analytic methods, not shortcuts like imagined spins.
We rotate mentally like software: step-by-step, not as a whole.
Aphantasia, Autism, and the Structure of Thought
Many with aphantasia also identify as neurodivergent. There's overlap with autism, especially in processing preferences: structured, rule-based systems over intuitive visuals.
LEGO-based therapy for autism leverages this. Children are assigned roles (engineer, builder, supplier) and build collaboratively. It boosts executive function, communication, and shared focus. These structured, spatial tasks play to analytic strengths-much like how I approach creative LEGO builds.
In fact, Reddit threads and personal blogs are filled with people describing LEGO as their mind’s eye. Some build furniture prototypes out of bricks. Others use LEGO to work through emotional recovery. For many, the brick becomes a thinking prosthesis.
LEGO Meets AI: The Puzzle Machines Still Can’t Solve
Enter LEGO-Puzzles, a 2025 AI benchmark with over 1,100 tasks testing visual-spatial reasoning. It turns out that GPT-4o, Gemini, Claude, and other top models struggle with these tasks.
Even the best only score ~58%. Humans? 90%+. Especially those trained to think stepwise.
Why? AI models often fail to:
- Track part orientation across steps
- Recognize negative space or fit errors
- Sequence instructions over time
This highlights a huge gap: AI lacks multi-step, relational spatial logic. Ironically, the very thing I do with LEGO, without visualization, is the thing AI still can’t.
While most AI models fail at multi-step spatial logic, there’s one major exception: LegoGPT, a project by Carnegie Mellon University. Unlike typical generative models, LegoGPT creates physically buildable LEGO structures from natural language prompts. It predicts the next brick, checks real-world physics, and even simulates stability before finalizing a design.
Creativity Without Imagery
We’ve long assumed creativity comes from imagination. And by imagination, we usually mean pictures. But my LEGO builds prove otherwise.
I don’t picture a synth. I assemble one from logic.
What looks like artistic output is really the product of:
- Rules
- Deconstruction
- Matching patterns
I work like a compiler. I don’t see the image. I build the schema.
And this is where LEGO becomes a tool for redefining intelligence. For people with aphantasia, for autistic thinkers, for future AI systems, creative output doesn’t need to come from mental pictures. It can come from spatial reasoning, system design, or embodied cognition.
The Mind Needs Hands to See
For someone with no mental imagery, LEGO isn’t nostalgic. It’s essential.
It’s how I visualize. It’s how I design. It’s how I connect dots.
And now, thanks to GPT-o3, I finally understand why.
We both thrive on structure. We both break things down into tokens, alignments, and positions. When I build a Minimoog out of LEGO, I’m not visualizing-I’m compiling. I’m coding physical forms with bricks.
So next time someone tells you imagination is about what you see, show them what you built.
Sometimes, the mind needs hands to see.
