On Friday I was made redundant. A year of work, wrapped up in a meeting.
I gave myself an hour to feel weird, then I got to work.
This is what came out the other side.
The rule
One rule, set on Saturday morning: ship something every day. Not polish, not pitch, not plan. Ship. Public URL, shareable link, real users if possible. AI doing the heavy lifting, me doing the taste.
I wanted to find out, honestly and without a salary funding the experiment, what an agentic design-and-build practice can actually output when the only constraint is your own throughput.
Five days later, here's the receipt.
01 / A new portfolio, from scratch
pjcooper.design, rebuilt end to end. New IA, new visual system, new blog, new catalogue, new playground. Built in Next.js with Claude Code as the primary collaborator.
Not a refresh. The old site got binned. This one positions me how I actually work now: design, engineering, content and AI as a single craft, not four departments to coordinate between.
02 / A journal that writes itself, sort of
/blog, a working publication, not a "coming soon."
Multiple posts on AI and design went up across the five days. The pipeline is the same OpenClaw-shaped setup I've been refining for months: agents handle structure and drafting, I handle voice and call the shots on what's worth saying. The bottleneck is taste, not typing.
03 / A typeface
CooperPixel, a two-weight pixel typeface, designed in code, shipped as .otf, .ttf, .woff and .woff2. Release notes here.
I'd been doodling glyphs for months and never finished it. Turns out "finishing" is what an agent is for. Generated, iterated, exported, packaged, published.
04 / A mindful game
Cairn, a browser stone-stacking toy. Real physics, gallery-white scene, no fail states, no score. Stack until it falls, or until you stop.
I'd wanted this since seeing someone build a cairn on a Korean travel show. It existed in my head for two years. It existed on the internet four hours after I started.
05 / A mint, finally
Phanackapan #14, a long-forgotten NFT collection of mine, dragged out of the drawer and given a proper mint site on Ethereum mainnet. Claim is live, capped at 460.
Designed and built in an afternoon. The collection has been sitting on a hard drive since the last cycle. It deserved a home.
06 / A word game, and a community
Tessera, a daily 4x4 word puzzle. Same puzzle for everyone, resets at UTC midnight. Now lives at its own domain, tesserapuzzle.com, with its own subreddit at r/TesseraPuzzle.
Designed, built, shipped, posted. Over 150 plays in under eight hours. Real strangers, real solves, real arguments about whether ESES is a word. (It isn't, in my dictionary. That's a whole other post.)
What this proves
- Day
- Output
- 01
- Portfolio rebuild + first posts
- 02
- CooperPixel released
- 03
- Cairn shipped
- 04
- Phanackapan mint live
- 05
- Tessera launched
I want to be careful not to oversell the lesson here. Five days of solo making is not a business. A typeface and a stone-stacking toy don't pay a mortgage. I'm aware.
But the speed is the point, and the speed is real. Six shipped artefacts across product, brand, type, web3, and game design, with one operator, in five days. A few years ago that would have been a studio quarter. Now it's a long weekend with good tools and a clear head.
Not "AI-assisted." Not "AI-augmented." Operated. The AI is the output layer; I'm the strategy, taste and direction.
This is what I mean when I say agent-operated design.
What's next
I'm open to work. UK, EU, remote. Senior IC, design lead, FDE, or short-form contract. If your team is trying to figure out how to ship at this kind of cadence, that's the conversation I most want to have.
Onwards.
From the Catalogue