For years my default evening routine was pretty simple. Kids go to bed, I get a few hours back, I play games, watch TV, scroll social media. Not every night, but often enough that it became the default.
Somewhere along the way I started swapping a few of those evenings for messing about with small ideas instead. Not “starting a business” or “side hustling” in a big dramatic way. More like poking at things that annoyed me, or hacking together little tools I wished existed, or half-baked product thoughts I’d normally leave in Notes and forget about.
A designer’s perspective
I’m a service and UX designer by trade. Most of my time is spent in the messy middle of real services: understanding what’s actually happening, making sense of constraints, and working with teams to turn that into something people can use.
I’ve also always had frontend chops. Early career I did proper front end web work, and I’ve never really lost the habit. I’m comfortable in code, I can build prototypes, and I’m usually happiest when I can move between user needs, product decisions, and implementation detail without too much ceremony.
The bit I’d still tended to avoid was committing to the full “proper app” stack on my own – auth, payments, databases, deployment, all the glue around it – plus keeping up with modern frontend stacks and best practice. Not because it’s beyond me, but because the mental overhead of learning it properly and holding it all in my head never felt worth it for an evening side project.
I’m already good at the service/UX side of things, and I care about staying sharp, but between consulting work, kids, family, dogs, and trying to do a half-decent job of looking after myself, there’s only so much bandwidth left. I didn’t really want my spare time to turn into “catch up on the latest stack” homework.
AI tools have changed that equation for me.
No stupid questions (and no shame)
I’m no stranger to dev concepts. I work around tech all day, I’ve spent years solving problems alongside engineers, and I’ve got a decent feel for stacks, APIs, and how these systems tend to fit together.
It’s just that there’s a difference between knowing the shape of something and being willing to take it on properly in your spare time. Stuff like React patterns, Stripe, paywalls, auth flows, database wiring, deployment, performance basics, logging, error handling, even the whole “how do you evaluate LLM output sensibly” rabbit hole. I’d always assumed that if I was going to do those properly, I’d need to clear the decks, carve out a month, and keep it all in my head long enough to make it real.
Whereas now I can just start.
That sounds minor, but it’s not. The best part isn’t that it writes code faster. It’s that it removes the “ugh, I don’t want to learn that bit” friction. I can say “help me wire this up”, ask dumb questions, try an approach, throw it away, and try again. If I write something horrible, no one cares – the stakes are low. It’s a side project. I’m not shipping this into a critical production environment. There’s no PR review. There’s no judgement. The only goal is that it works, and that I learned something along the way.
To be clear, I’m not outsourcing thinking. It still needs judgement – I’m not autopiloting. It just makes the boring bits less sticky, so I can stay in the loop.
A year ago I wouldn’t have touched payments in a side project. Now I can get a basic Stripe flow working without it turning into a weekend-long ordeal.
I think that lack of judgement is massively underrated.
Momentum matters
In day job work there are always trade-offs. You’re balancing quality, standards, deadlines, team conventions, readability, maintainability, stakeholders, all of it. That’s all good and important. But it does mean experimenting can feel expensive. In a side project, especially one you’re doing for fun, you’re allowed to be messy. And with an assistant in the mix, you get un-stuck constantly. You don’t spend an hour spiralling on a tiny syntax issue or a config detail. You keep momentum.
It feels more like having a capable engineer on tap. Someone you can ask the stupid questions to without doing the usual “sorry, this might be a stupid one…” preamble.
I’ve always felt like I knew just enough to be dangerous. Enough to get myself into trouble, not always enough to get myself out of it quickly. Having an assistant there closes that gap in a really nice way.
Also, this is the first time in ages I’ve found myself learning in a way that feels easy. I’m not forcing it. I’m not trying to become a full-time engineer. I’m just following curiosity. “Can I get this idea working?” turns into “how does this actually work?” and suddenly I’ve picked up three new concepts without sitting through a course.
To be clear, I’m not claiming this is some virtuous life upgrade. I still waste my time in the evenings. This isn’t a productivity post. It’s more that I’ve found a hobby that happens to produce tiny, useful things, and it scratches the same itch as gaming in a different way. Progress is weirdly addictive. So is shipping something small and seeing it exist in the world, even if only a handful of people ever use it.
Anyway, that’s it. I’m enjoying it a lot, I’m learning loads, and I’m giving myself permission to build things without overthinking it.
Long may it continue.
The journey (as best I can remember)
-
Before 2023: VS Code + tab complete: low stakes prototyping, getting tiny things working.
-
2023: Copilot: late to the party, but tab complete really started to feel useful, so I started using it more and more.
-
Late 2023: Cursor: heard about it from a HN thread and gave it a go. It felt like it was actually following along, so I stuck with it for most of these evening projects.
-
2025: Codex & Codex CLI: the point it clicked for me. Couldn’t believe I could give it a repo and some instructions, and it makes coherent changes.
This is when my learning really sped up, mostly because I got better at using the tools and started properly grasping the concepts as I went.
Once that clicked, it stopped feeling like “learning to code” and started feeling like play. It scratches a similar itch to gaming: quick feedback, small wins, and that loop of “what if I try this?”
I’m mostly into management/simulation games anyway (city builders, tinkering, optimising systems), so it feels weirdly familiar: make a change, see what happens, tweak, repeat.