Talking to a friend about our AI use, I came to a realisation about what role Claude has assumed in my life: It’s an idle clicker game for me to feel productive.
If you don’t know what an “idle clicker game” is, Wikipedia does a good job of summerizing the genre:
An incremental game is a video game subgenre characterized by the incremental accumulation of in-game resources, and gradual, often exponential progression through repetitive actions or automation. […] players typically start off by performing simple actions – usually clicking a button or object – to earn a form of in-game currency. This currency can be spent on upgrades, items or abilities that enhance income generation, often automating the process and reducing the need for direct interaction.
Wikipedia: Incremental game
In the most basic format, imagine that you tap a button to get 1 gold. Once you get 10 gold you can buy a robot that will click that button for you every second, and once you have 100 gold you can upgrade your robots to do it faster. And so on. Stuff happens on the screen and you are making “progress.” It can be extremely addictive and for certain types of people it scratches an itch of “min/maxing” in order to optimize your “progress” – i.e. “should I upgrade the robots when I have 100 gold or wait for a better upgrade at 500 gold?” I’ve actually setup Automator scripts to “play” such games for me, in order to try out the best strategies and maximise “progress.”
And I realise that I’ve started using Claude Code in a similar way. I used it to “vibe code” a couple of homepages (monocultured.com, matepo.se and the relationships evaluation form) and I’ve dipped my toes into cataloguing notes and organising them in Obsidian, as well as automated my job search by having Claude digest my resume and experiences to tailor job applications. But once those low hanging fruit were done, I started to feel that every idea needed to be run past the LLM and, more importantly, if I have any unused tokens left it feels wasteful not to use them. Which is an insane take!
(When you pay for AI you can use a certain amount of tokens in any given period of time – once those run out, you have to wait for the timer to reset. But the tokens left don’t carry over into the new timer, so you’ve “lost” those)
I know that most of my AI use is a terrible waste of resources (water and power) and yet my impulse is to treat it as a button that I ought to push to LARP productivity.
(The recent Gas Town post and the resulting discussion about being “more productive” merges with the hustle culture of Linkedin bros and suddenly every un-used cycle is a potentially missed opportunity.)
I don’t think everyone has these same instincts as I do. I have a peculiar scarcity-driven mind which gives me a terrible case of FOMO in all aspects of life, but I’m a good enough canarie in the AI mine to serve as a warning of how AI can twist someones incitaments and drives in regards to what it means to create value and to be productive: Do more, refactor everything, optimise your outputs and everything can be an input.
I’m rereading Peter Watts Blindsight, where there’s a post-human whose job is to synthetisize input and arrive at an output, without any self to get in the way.
Through it all I tried to do my job. I compiled and collated, massaged data I would never understand. I watched the systems around me as best I could, factored each tic and trait into the mix. One part of my mind produced synopses and syntheses while another watched, incredulous and uncomprehending. Neither part could trace where those insights had come from.
Peter Watts: Blindsight
And the way the character (funnily enough named Siri) approaches information reminded me of how I use Claude – a conviction that if I just add enough inputs, the outputs have to have some value. And because humans are lazy by design, this simulation of insight-creation is convincing enough that it’s easy to gloss over the difficult job of thinking things through – which might have helped you realise that what you’ve spent the last day doing is just make-work, or even make-pretend-work.
In the end, this is neither here nor there as far as usability of AI is concerned. It’s just an observation of how my approach to it is making me a disservice. Just like social media, tobacco and procrastination, it’s a behaviour that I need to monitor in order to find a balance between what is adaptive behaviour and what’s merely me giving up initiative in favour of a meaning-shaped activity – an idle clicker for my productivity impulse.