A pattern I didn't see until I'd written three blogs

ChatGPT Image Jun 1, 2026 at 01_41_04 PM.png

Three blogs in. I sat down to write the fourth - and noticed something I hadn't seen while writing the first three.

They look like three different topics on the surface. Strategy. Engineering. Research. Different problems, different vocabularies, different lessons.

But underneath, they're all saying the same thing.

Here's the pattern I didn't see until I stepped back.

Blog 1 - Nobody cares how good you are at Figma

A vague problem statement landed on my desk: take a built career counselling bot on-ground across India. 4,000 students, 40 course streams. Build the pilot strategy.

I froze. Not because I lacked design or research depth - but because the ask was a validation strategy, not a design or discovery one. A different muscle entirely.

The blog walked through three pillars that strategy had to answer:

Who exactly are we validating with? Personas as the unit of validation, not decoration.
How broadly do we have to test? Diversity as a sampling constraint, not a sentiment.
What does "it works" mean? Success as a falsifiable claim, not a vibe.

The takeaway: the gap between an executor-level designer and a strategist isn't skill. It's synthesis.

Blog 2 - Did AI just make a UX Designer a UX Engineer?

A side experiment. I picked up something I'd tried before and failed at - designing a dashboard and building it. This time, with Claude and ChatGPT in a tab, I shipped it. Framework-adaptive components across React, Vue, Solid, Angular. Zag.js underneath. Vibe coding.

It worked. And then it didn't. Buttons broke in the fourth context. Tokens drifted. State machines fragmented across reactivity systems. The code compiled, but the system underneath was fragile in ways I couldn't always explain.

The takeaway: UX Design shapes the vision. UX Engineering makes it usable. The role is real. The shortcut isn't. AI gets you to the doorway, not through it.

Blog 3 - Personas are theories. Conversations are proof.

Once the pilot personas were defined, I had a quieter problem: how do I know they're real?

So I went back to actual student transcripts, adapted them per persona, and scripted conversations that stress-tested two things at once - the bot's responses and the persona model itself.

Some personas turned out to be too similar. The bot was uneven across emotional types. The abstract personas had to be rewritten into something testable.

The takeaway: if you can't write a conversation that distinctly belongs to a persona - and only to that persona - you don't have a persona yet. You have a label.

But also: this was only the first floor. The real test waits in the pilot, on the ground.

What ties the three together

Read in sequence, the blogs say the same thing from three angles:

Figma is not the thinking. AI is not the engineering. Personas are not the validation.

Every tool we lean on as designers - Figma files, AI prompts, persona templates - is a representation of the work, not the work itself. The work is the thinking underneath. The strategy. The synthesis. The judgement about what to build, who to test it with, and what would prove us wrong.

That's the muscle the series is really about.

Three blogs. One thesis: the craft isn't the job. The thinking is.

P.S. - More coming. The next one is about what happens when this pilot strategy actually meets the ground.