Nadia Piet & Archival Images of AI + AIxDESIGN / https://betterimagesofai.org / https://creativecommons.org/licenses/by/4.0/
Reflecting on the first 10 things I’ve made with Claude
Since the start of the year, I’ve overcome my reluctance at being a vibe coder and have created a number of small things that I am actually very happy with. My resistance has been less to do with any moral objection to using AI and more to do with the stubborness I always have when it comes to much-hyped new things. It’s a character flaw perhaps but I like to think of it as a cognitive defense mechanism.
The stuff I have made falls broadly into three categories.
Serious tools
I started on solid ground by building some apps based on half-finished things I had lying around. Mostly spreadsheets. I had a bunch of these that I created as “ready reckoners” for books or ideas that interested me. The first thing I made was the Patterns of Strategy Explorer, based on the book by Patrick Hoverstadt and Lucy Loh. I had also toyed around with building and assessment tool based on Simon Wardley's doctrine. That became the Wardley Doctrine Assessment. Similar to the strategy patterns, I also had documented change patterns from the book Fearless Change and that became Patterns for Introducing Ideas.
For all of these, I had something that I had already made to measure the results against, which made starting feel less like a leap and more like the final version of what I had started. What surprised me was how quickly Claude translated structured knowledge into something interactive and found possibilities in the material I hadn't considered. This process extended to other sources besides those I already knew. After learning about a prescient academic paper from the 1980s about automation, this quickly became the Bainbridge framework. Each cycle of building got faster and faster, from days to hours as I leant what the tool could do and also what my preference were. Small, self-contained apps that I could carry around in my pocket.
I wrapped this phase of experimentation up with an interactive version of the Habit Card, a digital companion to a physical card I'd already designed.
Silly games
I often anchor ideas in wordplay, coming up with just the right title to hermetically seal the idea in as few words as possible. My notes on my phone are full of these. Some of them are super cringe when I go back to them and some still make me laugh when I rediscover them. Coffin Dodger is one of those. The term is mildly offensive British slang for an old person who is still alive despite the odds. I had written a note on my phone:
Idea for a game
Coffin Dodger
Vertical scroller in style of frogger
You play a character riding (illegally) on the road on your motorised scooter
Ahead of you are hearses that keep dropping coffins on the road you must avoid by moving L/R/F/B.
(Note: I didn’t realise at the time that there is actually a proper Coffin Dodgers console game. So much for original ideas.)
Kindness Kommando was next up. I’d written the phrase born of frustration with modern workplace culture where employees seemed to be increasingly smothered by wellbeing (I’m Gen X BTW if you need any more explanation). Given my new found ability to create 8-bit scrollers, it seemed to fit well with the gamplay of Commando.
Both of these were one-shot, throwaway games. Not serious or useful. But that became the point in a way. It didn’t have to be serious to be worth it. Disposable playability can be as worthwhile as disposable utility.
I wrapped this phase up with one final satirical idea. ClapFunder was a one screen mockup I’d made as a joke about people putting their money where their hands were during the whole Covid-era “clap for the NHS” period. Another throwaway idea. Or was it? It turned me on to the idea of making play more serious by exploring APIs. What if you could actually hear the clapping and donate the money? Turns out you can and pretty easily. Web technology has moved on so much since I was last hands on with building stuff that to me these capabilities seemed miraculous. Especially in the context of a throwaway idea. I took a lot of this thinking forward and got a little more serious about my play.
Design concepts
Having realised I could do stuff with audio, I came up with Expanse FM, the realisation of an idea I'd had for an interactive printed poster years earlier, where you could activate the sounds of the planets. What made this particularly satisfying was learning about how to extract and chop up the source audio and mix it with generated static to bring the radio dial mechanic to life.
Dreambot - conceived while half asleep, approriately - became the most technically ambitious idea I’d tried: a mobile dream interpreter where you submit a dream and receive readings from a cast of historical personalities, wired to the Claude API and hosted on Vercel.
What does it all mean?
Reflecting on what I been able to do so far, it seems less about building apps than following instincts. Starting with ideas I already had, discovering ways to realise them that, whilst knowing they were possible at the time, required too much time and effort to pursue.
Also, letting the capability shape the idea rather than just serve it. I went in thinking about disposable utility - purposefully making small one page mobile tools, that you can use once, for one purpose and move on (I still subscribe to that idea and found myself re-reading Scott Jenson’s polemic essay from 2011).
What I also found though was an experience of making things that was closer to disposable playability. Ideas worth returning to and that have a certain open-ended, never-quite-finished quality to them doesn't get quite so easily discarded.