[Me & X] Happiness or production?
X: When you code with AI, do you tend to use small or large changes?
Me: It depends upon what I'm optimising for - happiness or production.
X: Explain.
Me: If I'm building a prototype, I'll go big. I'll use many agents, a high level specification, get it to build functional tests, create artefacts if I feel I need it like architectural diagrams and then just enjoy the journey of it building something "good enough" with a bit of wrangling along the way and rollbacks when you need them (which is pretty much a given but no big deal). It really is quite exhilarating. It makes me quite happy.
X: Can you show me an example?
Me: Prototypes never go to into production. However, that said, about a year go, I vibed this game - https://lnkd.in/e4bTydNb - it's more of a demonstrator, something for very little work and things have only improved since then but it's good enough to show you what's possible.
X: Production?
Me: I'd go small. Tiny changes to code, which I'll read and then throw my hands up in horror at the sloppy or incorrect assumptions, faulty decisions, made up stuff, questionable workflow and generally things which make me want to hurl expletives. It's still impressive but also like wading through garbage some of the time and that doesn't make me happy.
X: But, why does the big stuff make you happy? Doesn't it also have garbage?
Me: Because I never read the code, it's a prototype. I just admire the scale of it, the adrenaline rush of it having done this, to have broadly built the right functionality, to have passed whatever function tests are needed .... 20k lines of code for which I barely did anything, that's just wow. It's not like it is something that will have to be maintained or I have to be concerned about structure. If I actually read the code and looked at the structure then I would probably despair or more likely there would be tirades of expletives. I've done this before, for example when it built me a testing engine which turned out to be a simulation. The architectural diagram was an outstanding work of fiction. I'm not doing that again.
For Happiness: go big, don't read the code, just wrangle the system, it's a prototype, rollbacks are your anchors, ignorance is bliss and don't worry about structure. Just go with the flow.
For Production: well, you need to understand things to maintain them, go small, expect a lot of swearing ... a lot of swearing.
Originally published on LinkedIn.