Rob Boyett

Executive Summary

Rob Boyett occupies a distinctive position in conversations about AI and creative work: the technically literate designer who treats AI neither as autonomous collaborator nor mere tool but as cognitive extension, something closer to an appendage that offers the ability to hold far more information in memory and recut it in different ways. His practice reveals sophisticated understanding of AI's current capabilities and limitations, shaped by three years of intensive experimentation across image generation, language models, prototyping platforms, and coding assistants.

He is clear-eyed about where AI currently sits. It supersizes your brain. It does not yet constitute genuine creative partnership, and he knows the difference.

Symbiosis sounds better — it benefits from you, you benefit from it. It learns from you, you learn from it. It doesn't learn from you at the moment. You learn from it a lot.

The relationship remains one-directional. AI generates ideas he would not conceive but does not evolve through their collaboration. True partnership involves mutual transformation. That has not arrived. What has arrived is something genuinely valuable: a cognitive extension that enables braver experimentation, faster iteration, deeper research synthesis, and the psychological safety of asking foundational questions without status anxiety.

From Skill Extension to Integrated Workflow

Rob's initial AI engagement was pragmatic: filling capability gaps. Image generation addressed a desire to paint without the years of practice painting requires. Language models extended writing capacity he valued but had not fully developed.

There's so much effort if I want to paint a picture — it's going to take me a really really long time.

The inflection point came with Cursor, the AI-assisted coding environment. Unlike earlier design-specific AI tools that felt either too abstract or too featureless, Cursor gave him something he could actually see and edit. He started small, one-page prototypes in vanilla CSS and HTML, building confidence before pushing into unfamiliar territory, then pausing to request explanations and code comments. He acknowledges the risk honestly: it might be a false sense of security, but the understanding feels real.

His current workflow triangulates between FigJam for visual thinking, Figma for design execution, and Claude as thinking partner. He creates project-specific Claude instances populated with documentation: personas, strategy, meeting transcripts, frameworks. The conversations that used to happen at whiteboards with colleagues now happen here, in solo sessions that generate shareable artefacts.

Critical Mass and the Multiplication Effect

The qualitative shift Rob describes happens when sufficient context accumulates in a project. Once personas, strategy, and meeting transcripts are loaded in, the AI stops being a generic assistant and becomes a project-specific thinking partner.

When you've got enough documentation in, you start going and digging into an idea and you feel it. It's like — oh, it really knows, it really understands.

The difference is concrete. Where he might independently generate one or two test variants for an idea, Claude at critical mass generates ten. This is not just quantity; it is perspective diversity, synthesising project context, industry knowledge, and methodological frameworks to produce approaches he would not conceive independently.

The iteration revolution this enables is real. He can now follow design methodology more faithfully than before, genuinely diverging to fifty competing ideas before converging, because the resource constraint that previously forced premature closure has disappeared.

The resource problem has gone away. I've got enough time. I can do this thing as long as I can sit at my desk and put my headphones on.

The Echo Chamber Problem and the Failed Experiment

Despite this enthusiasm, Rob identifies a fundamental limitation with precision. AI is a Labrador. It agrees with you. This creates the risk of becoming increasingly satisfied with first versions, moving forward without the friction that produces better work.

He tried to fix this. He wrote a system-level prompt requesting critique, challenging assumptions, refusing easy agreement. He turned it off after two days. The AI could not distinguish between moments requiring encouragement and moments demanding scrutiny. It applied the same critical posture regardless of where in the creative process the work sat, and the effect was to extinguish small ideas before they had a chance to breathe.

It was friction and it wasn't that it was wrong. It was just like when an idea is small you could put the spark out quite quickly.

What he calls AI's contextual blindness is the deeper issue. Human creative partners read facial cues, emotional states, project history, and countless subconscious signals that determine when to challenge versus support. That situational emotional intelligence is precisely what current AI lacks.

Commercial Constraints and the Happy Accident

Rob's design context differs from artists pursuing pure exploration. Brand consistency, client requirements, functional constraints all shape what AI integration can look like in commercial practice. For text, brand alignment is straightforward: feed in a writing guide and it works. Visual consistency remains harder, more approximate. His solution is accepting the spirit of the brand for early-stage work, where the goal is solving questions rather than producing final assets.

The happy accident, AI's generative unpredictability producing something better than planned, remains present and welcome, but must be managed rather than simply embraced. Rigid adherence to preconceived outcomes fights AI's nature. Successful integration requires openness to unexpected directions while maintaining curatorial authority over what actually serves the project.

The Collaboration Gap

Rob's most pressing practical frustration is cross-modality translation. Visual thinking happens in FigJam. Conceptual work happens in Claude. Code lives in Cursor. Moving between them requires manual export, format conversion, re-import, breaking the bidirectionality that would make the workflow genuinely fluid.

I want to press a button here and that thing shows me the same thing but in a markdown file and then I want to be able to press a button back and it does it again — up down up down.

His solo-developed workflow also creates team integration challenges. The Claude conversations replacing whiteboard sessions remain private rather than shared team resources. He wants collaborative project spaces where multiple designers contribute to shared context and visible thinking. That does not yet exist in the way he needs it.

The things that will matter long-term, clear thinking, articulation, storytelling, narrative, remain irreducibly human. AI levels the technical playing field and restores primacy to those fundamentals.