Projects shipped, lessons compounded, creative experiments attempted, including one that involved AI making a short film. An actual account, not a highlight reel.
2025–26 was the year the toolkit stopped feeling like a set of tools and started feeling like an extension of how I think. The work got quieter. The decisions got faster. Something clicked.
It was also the year AI stopped being a conversation topic and became an actual collaborator, sometimes useful, sometimes humbling, occasionally surprising enough to make me rethink the whole process.
This isn't a highlight reel. The messy projects are here too. So are the lessons I didn't go looking for but found anyway.
"The year the tools became invisible, and I finally stopped thinking about design and just started doing it."
Personal note, early 2026Some were wins. Some were complicated. All of them moved the needle somewhere.
A digital collectibles platform, the design challenge wasn't the UI, it was building a visual language for ownership. Defining what "yours" feels like on a screen took longer than the entire component library. Returned to this in 2025 for a V2 pass; cleaner system, better handoff.
Civic data mapping for real people. This one set the tone for the whole period, shipped in two weeks, taught me more about information hierarchy in a fortnight than months of prior work. The kind of project that recalibrates what "fast and right" actually means.
Gamified learning, a product that could easily become visually obnoxious. The restraint was the whole job. Keeping the motivational energy without the noise took three rounds of scope-cutting. The client loved it. I see the seams. That's probably fine.
A branding and interface project where the brief kept expanding. Learned to push back earlier. The final output was tighter than it would have been if I'd said yes to everything, and the client relationship was better for the honesty.
Healthcare coordination for distributed teams. The UX had to be invisible, no friction, no ambiguity, nothing clever. The most restrained work of the period and probably the most consequential. Stakes change how you design. They should.
None of these came from a course. Most came from getting something wrong, then paying attention.
ArjiMap proved that moving fast with clarity produces better outcomes than moving slowly with uncertainty. The two-week ship forced every decision to be defensible. That discipline stayed.
The designers who struggled with AI in 2025 treated it as a time-saver. The ones who got interesting results treated it as a thinking partner with an unusual perspective. Those are different relationships with different outputs.
Scope that expands without conversation doesn't expand, it degrades. The clearest thing I learned from Limson is that a well-placed "not in this scope" is a better deliverable than a compromised yes.
CollabCare reinforced something I've been circling for years: the best UX work leaves no trace. Users don't notice good design. They only notice bad design. Designing to be unnoticed requires more confidence than designing to impress.
Some worked. Some didn't. One of them involved AI making a film, which deserves its own section.
Spent six weeks learning how to direct an AI-generated short film from scratch. Not a proof-of-concept demo or a prompt screenshot, an actual short with narrative structure, visual continuity, and a point of view. The tools were Runway, Midjourney for visual reference, and a lot of manual sequencing in After Effects to make the output feel authored rather than generated. The result was rough around the edges. It was also the most genuinely surprising thing I made all year, not because the AI was impressive, but because the process of directing it forced me to articulate visual decisions I'd been making intuitively for years. Explaining composition to a model that doesn't understand it is a strange way to understand it yourself. Recommended.
Finally built a token-based system that actually held up across a multi-client engagement. The discipline of maintaining it was the hard part. The tool finally made it achievable.
The novel. Two architects who keep narrowly missing each other across heritage restoration sites in India. Writing is slow, intentional, and completely mine. Still going nowhere fast. Still fine.
The brand has a spine now. The portfolio site is live. The scattered work finally has a coherent identity to sit inside. More to build, but the foundation is solid.
Did this for about six weeks. Felt like a bit from a tech satire sketch. Back to one. Some experiments exist to confirm what you already suspected.
The work got clearer. The tools got quieter. The experiments got stranger. That feels like the right trajectory.
Here's to the next set of problems worth solving.