Shells

I’ve collected things for as long as I can remember: shells, paper, envelopes, fabric, wood, stones, postcards, pins, books. I can vividly remember my first scrapbook, a bright Orangutan on the cover, the smell of the ocean blue recycled paper pages – sticking down as much as I could on every single space available. This deep compulsion to catalogue, organise and categorise has been frustrating at times but the end is always rewarding – satisfying. I like making sense of things, even if it’s purely aesthetically, the way things are organised rather than the logic of the order itself. It makes me feel like I understand the objects better – and that somehow the objects now understand me.

When I first started building comprehensive design systems I began researching taxonomy and structural organisation. Carl Linnaeus was an 18^th century Swedish biologist who created the modern system for naming organisms. The absurdity of categorising everything with two basic phrases was impressively simple and bewilderingly complex at the same time. But this systemised way of thinking was the underpinning for my classification structures – both hobby and at work. Design tokens, version control, semantics – and more of the world around me – all started to fall into place.

Custodian

Brand design systems catalogue a brand’s visual components. They set up rules for those components and how they should behave. My thinking on how to set these systems up, and how to maintain them, stemmed from a simple idea – fluid taxonomy. The simple and easy to catalogue naming convention as the core starting point. The diversity and nuances that come with something more subjective such as design is the fluid layer on top. The room to grow into and the allowance for something different when the time is right: a one-off campaign, a sub-brand, an event identity.

Vira was born from this idea: applying and maintaining a system with a creative tolerance. Vira was on the cusp of speaking to members of the team and fluidly articulating the rationale behind design decisions – flagging issues before they shipped – suggesting changes in existing systems to make them more unified, consistent and easier to manage.

The system underneath, the storage in the middle, the applications on the outside – Vira mapped and fed into each. This is my shell collection, on a multi National Museum type scale, multiplied over several different locations, being managed, maintained and updated by a single entity.

I spent a month validating my idea before building. Going back to Linnaeus’ approach time and time again late into the night, how does this make sense with this – deep in Reddit posts, design forums and Substack theories. No real physical connections but the gut instinct was there the whole time. So after four months of building more or less every night after my day job, the day before I read the news, Vira had:

• A complete authentication and project system with multi-tenant support
• The Pulse compliance surface with scan history, Watch scheduling, and Fixes lifecycle
• The Memory surface with Tokens, Composition (AI-inferred relationships with confidence scoring), and Brand Profile sections
• The Counsel chat interface with project-aware context
• A Figma plugin that synced design tokens directly from Figma files
• GitHub OAuth integration with token import and pull request creation for fix application
• A sophisticated component primitive library with selection APIs, tokenized spacing, and decoupled semantic color systems
• A complete security audit foundation with SVG XSS protection, rate limiting, API key authentication, and access logging
• An architectural foundation for MCP integration including a persistent Violation collection with stable cross-scan identity, a normalized CSS value formula, an isolated reconciliation engine with dry-run safety guarantees, and a fully tested backfill migration path

Hours

The biggest threat to Vira was Figma releasing some kind of live-repo, MCP wire-up. But I dismissed it as a tertiary layer of an existing incumbent product that would be overlooked with enough room for plenty of other serious contenders to still play in. That said, early one morning in June, scrolling through news articles over a dawn-lit breakfast, I saw a post about Claude Design’s new feature release. As I read the second paragraph I zoomed in on the import from GitHub repos – bidirectional integration via /design-syncand /design commands – my heart sank.

The overlap with Vira’s MVP – still 2-3 months away – was incredibly substantial. The remaining differentiators were: live-site production scan, long term memory, multi source conflict mediation – real features but narrow by comparison and likely to be table-stakes for swallowing soon enough anyway.

The product was available to Pro subscribers at $20 a month. Anyone already using Claude Code would be pressed to justify moving to an enterprise level pricing tool with a few more specialised features than they were already getting access to. I knew it was over – I closed everything within an hour of reading the formal announcement from Anthropic. That night I slept for 3 hours in between the constant what-ifs, overwhelming doubts and relentless how-can-it-bes.

Contemplations

Vira's market-read was theoretically correct – the strategy, build, infrastructure were on point and the audience was arguably looking for this kind of solution. Closing the product before it reached daylight was due to the speed and distribution of Anthropic, but I do wonder if this could have been a blessing in disguise.

Did Vira have legs in the first place. Was the creative architecture that I had tried to instil in an agent good enough to compete with the taste of a seasoned designer. Was the lane so slim that the growth trajectory would have never moved away from purely design. The Claude-Design feature was just a feature. Small ideas do often find success. But it could have still been years of building something that looks beautiful, is built elegantly and functions at the highest technical level, only to find that the problem we’re solving isn’t actually deep enough.

Part of me was also chasing the glory, despite the real work, perhaps the potential payoff added weight to my blinkers. I wanted to play with the best – I could feel the drive for competition, chasing the likes of Cursor, Linear, ElevenLabs. Miles ahead of me on so many scales but I could see how to enter the arena.

It was validating to know in one sense I had seen around a corner. But the bigger player building a very similar thing with a wider market share made me refocus my attention on the craft. Which is where I’ve always been anyway. The shells, paper, envelopes and books are all still shaping my thinking. Not just the design as a craft, the technical application as craft too. So Vira’s scrapbook is on the shelf now – I’ll come back to see it next to the orang-utan cover from time to time – sift through the pages and will always remember why it was worth doing.