My Website
The site you're on - built research-first, fully static, and mine end to end.
I wanted a place to put everything - the photographs, the projects, the random stuff I've built or at least tried to. Less portfolio, more library. This is the site you're reading it on.
It was the complete opposite of how I built Stride. That one grew on vibes; this one I planned. I had a clear vision of how things should look and was already hosting an earlier version on 22slides.com - so the job was to rebuild it properly and own every part of it.
The first thing committed wasn't code - it was deep research: what stack made sense in 2026, what a photography-plus-project portfolio actually needs, and what it deliberately shouldn't have. Out of that came a roadmap of five phases, ordered by hard dependencies, with the landing page nearly last because it's a composition of everything else.
deveshksd.com ↗Built around the photographs
Every layout decision deferred to the images. Full-bleed heroes, a nav left floating and transparent over them, and one warm paper background on every page so a photograph never has to fight the interface for attention. No dark mode, by design - a photo deserves to sit on the same surface every time you look at it.
Snowdonia, Wales - the kind of frame the whole site is built around.
Beer, on the Jurassic Coast - from the Uncharted Canvas collection.
Room for the unplanned
The Museum of Curiosities was never on the roadmap. It appeared one exhibit at a time - the home for everything that didn't fit a collection or a case study. The site is a bit all over the place on purpose, and this is where that shows.
Built by someone who just keeps optimising stuffs
By day the work is data; the photographs and projects are the creative stuff on the side. This is the first place both live together - the tinkering shows up in the random projects, the pictures in nearly every full-bleed frame. A personal home more than a portfolio.
Bristol harbourside - from the Bristol collection.
Five phases, ordered by dependency
The font saga and other fights
The plan said Cormorant Garamond. The site shipped briefly with Playfair Display and Raleway (classic AI bias). It runs today on Spectral and Jost. Each swap happened because the type looked right in a spec and wrong on a full-bleed photograph. There's even a commit titled "correct stale font comments" because the CSS couldn't keep up.
The nav was the longest-running fight: a transparent bar floating over photography looks great over a dark hero and vanishes over a white page. It took five attempts - backdrop blur, a CSS filter hack, mix-blend-mode, currentColor SVG, and finally a gradient vignette with route-aware dark mode. All of that for what the spec called, in one line, "floating transparent nav".
The landing hero went through its own evolution - static image, then scroll animation, then a slow random slideshow with a blur crossfade. The randomness introduced the site's best bug: shuffling with Math.random() during render meant the server and the browser disagreed about which photo came first. A hydration error, a visible photo swap, and massive flickering on the most important page of the site.
The audit
Two months in, with the site live, I ran a full Fable 5.0 review across four tracks - design, performance, security, and code quality. It produced a score and, more usefully, a ranked list of every sharp edge.
The fixes went out as eight focused batches over two days: visible bugs first, then script hardening, then a dependency purge that removed an entire animation library (GSAP came along for one experiment and never left), and finally SEO and cleanup. The very last commit of the build deleted the privacy policy page - I'd built the full consent machinery for analytics, then accepted that a personal photo site doesn't need to act as a SaaS product.
Key Technical Decisions
- Next.js 16, fully static: App Router, React 19, strict TypeScript. Every route is pre-rendered; unknown paths 404 at the edge.
- Typed files as the CMS: collections are TypeScript arrays, case studies are MDX. The whole "backend" is basically a folder of content files.
- ImageKit + a 30-line custom loader: one master upload, with thumbnails, heroes, and OG images as URL parameters and blur-ups server-rendered.
- Motion in thin client wrappers so pages stay server components - the site animates everywhere but ships very little JS.
- Anti-features, decided in writing: no dark mode, no contact form (it's a mailto: link), no loading spinners, no carousels. A photograph deserves to sit still while you look at it.
What I'd tell you if you're building one
- 01Weighted decisions before building anything.
You won't build a skyscraper on unstable ground.
- 02Static gets you far on a content site.
There are no servers, no cold starts, and nothing to patch at 2am.
- 03Your content model is your CMS.
Typed files plus small scripts go a very long way for a single author.
- 04Decide what you won't build, in writing.
The anti-feature list killed more scope creep than anything I actually built.
- 05Always audit your own site.
The scored review found real bugs that went under the radar.
- 06Leave room for the Museum.
The unplanned section turned out to be the most personal part of the site.
The site is a bit all over the place, which is fine for me. And honestly, so was the process.
deveshksd.com · built April – June 2026 · Next.js · MDX · ImageKit · Motion