AI Fluency · Daily practice · Side projects

AI is a tool.
The ideas are mine.

How AI shapes the way I work day to day, from brainstorming layouts with Claude, to vibe-coding small apps, to knowing exactly when not to trust the output.

AI-assisted design Vibe coding Brainstorming partner Security mindset Daily practice

AI sits next to me, not in front of me

I open Claude most mornings. It is the place where I think out loud before I touch Figma or VS Code. When I'm stuck on a layout, I describe the goal, the audience, the constraints, and ask for three or four directions. Then I argue with it. Most of what comes back is wrong, but the conversation is what helps. By the time I close the tab I usually have a clear picture of what I actually want to build.

That's what AI is for me: a brainstorming partner that never gets tired and never gets bored. The decisions stay with me. The taste stays with me. The accountability stays with me. AI just speeds up the part where I'm circling around an idea.

AI is good for the layout. It is not good at owning a complex workflow. The closer the work gets to real users and real edge cases, the more I take back the keyboard.

A small travel planner for Indonesia, built in an evening

This is what vibe coding looks like for me. I had an idea, opened Claude, and worked through it until something existed. The point wasn't to ship a product, it was to keep the muscle warm: prompt, review, adjust, prompt again, until the thing on screen matches the thing in my head.

Travel planner prototype for Indonesia, built end-to-end in one evening with AI as pair programmer. Layout decisions, flow, and final polish are mine.

A notation tool to help authors create interactive theory

Authors I work with kept hitting the same wall: they could write theory in a doc, but turning it into something interactive meant handing it to a developer and waiting. So I built a small notation tool that lets them mark up plain text with simple inline tags and get clean, interactive HTML on the other side, ready to drop into a learning module.

It's a tool with a clear job: shrink the gap between "I have the content in my head" and "learners can click through it." AI helped me with parser logic, but the interaction model, the tag syntax, and the author-side UX are decisions I owned.

Where AI helps, and where I take over

AI is genuinely good at layout: spacing, hierarchy, a first pass at a component, a quick variation when I'm not sure between two options. It's good at boilerplate, syntax, and the kind of code I'd otherwise look up on Stack Overflow.

It is not good at the work I actually get paid for. My day job lives in complex workflows, B2B learning systems with role permissions, content variants, scoring rules, accessibility constraints, and a hundred small business decisions baked into every screen. AI cannot hold that context the way a real product designer does. If I let it drive there, I end up rewriting more than I saved.

So the rule I work by: AI handles the mechanical, I handle the meaningful. Layout and first drafts, yes. Workflow logic, edge cases, anything a user will trust me with, I do that part with my own hands.

Knowing what I'm actually feeding the model

Working in B2B means client data is never far from the surface. So I treat AI tools the same way I treat any third-party service: I'm aware of what goes in, what comes out, and what I'm allowed to share. Real user data, internal screens, anything proprietary, those don't go into a prompt. Sample data, dummy users, abstracted versions of the problem, those do.

The same goes for code AI writes. I read every line before it ships. I check what packages it pulls in. I look for the small dangerous things: a token logged where it shouldn't be, a permissive CORS rule that "just works," a dependency I don't recognise. The speed AI gives me only stays a benefit if I stay the one who's responsible for what gets shipped.

It changed the shape of my day, not the source of my ideas

The biggest shift isn't speed. It's that I rarely sit on an idea anymore. The cost of trying something dropped enough that experimenting is cheap. That's freed me to be more curious, to explore more directions before committing, to test small ideas without budget conversations.

But the source of the work is unchanged. The ideas are mine. The judgment is mine. The taste is mine. AI is the tool that helps me move from thinking to building faster, the same way Figma did a decade ago. It shapes how I work. It doesn't decide what the work is.

Back to the beginning

Teacher portal, the flagship project →