Early 2025 AI was a toy. ChatGPT could write a function, sometimes correctly, often not. It was impressive in demos and frustrating in practice. A year later the picture is unrecognisable. AI is now genuinely useful — useful enough that I can build almost anything I want in a single day, regardless of how complex it is.
The funny thing is that AI itself still pushes back. Ask Claude to build something non-trivial and it will tell you, with a straight face, that it’s a multi-week effort. It isn’t. Almost every task I take on now finishes in one or two days, with higher confidence in correctness than I ever had before. Backtesting, regression tests, schedulers, an Anthropic API key wired into a cloud monitor watching the thing run — none of that existed at this level of accessibility before. Today it’s a side-project default.
What this actually looks like
I built litreprice in a handful of evenings. Maybe two or three weeks of stolen post-6pm time after the kids are down. The result is a tool that works properly, that I’d happily ship and use. Before AI? It would have taken me one or two years, and that’s only if I survived the learning curve. Realistically, it wouldn’t have happened at all. I would never have had the time, the knowledge, or the feedback loop to figure out the security model, the device-compatibility tradeoffs, the dynamic scaling based on host CPU cores — the kind of things that only a senior developer with a decade of context would think about unprompted.
That’s the part that gets undersold. AI doesn’t just write code faster. It hands you the questions a senior developer would ask. How will this hold up against an attacker? Will it work on an old Android? Should we adapt to the host’s available cores? Even with infinite time, a non-specialist would never get to those questions. AI surfaces them for free.
Builder is a new kind of role — it absorbs the developer, but also the engineer, the analyst, even the business-only colleague, as long as they are curious enough. The technical knowledge gap that separated them has been filled in.
The doom narrative is wrong
The standard line is that AI will open security holes, kill jobs, push people into the streets, trigger a recession. I think this is exactly backwards. It opens more doors than it closes. If you lose your job, you can create five for yourself. You can build apps, websites, tools that save other people money or time, in evenings, for the cost of a Max subscription. A hundred and twenty dollars a month and a few free evenings is the entire risk profile.
Before, building something meant being a senior developer, having time, or having money to hire one — and most projects failed anyway, after months or years. Now the failure mode is “I spent two evenings on an MVP that didn’t work out.” That’s not a risk. That’s a Tuesday.
Finding the gaps
There’s a niche on Twitch right now where streamers run auctions through chat. Bidders type prices, a moderator manually counts down, the whole thing is rudimentary and confusing. Streamers are already busy presenting the auction and engaging viewers — they have no tooling. No minimum prices, no step prices, no anti-sniping, no scheduling. None of it.
That’s a gap. I spent one evening on it and I have full documentation of what I want to build plus a working MVP. A week or two of evenings and I’ll have a product I can sell to streamers. That’s it. One example, but the world is full of these gaps. The constraint isn’t team size or skill. It’s whether you noticed the gap and had the curiosity to start.
Creativity is the new bottleneck
If I were hiring for my team today, I would not screen for technical skill. AI handles the code part. I’d screen for creativity, and specifically for a can-do mindset. The hard part of any interview question now is no longer “can you implement this,” it’s “how would you approach this, how would you decompose it, what would you build first, what assumptions are you making.”
How do you measure creativity? You can’t with a tidy rubric. But you can ask:
- How would you approach this complex problem?
- How would you handle a non-creative teammate?
- How would you define a creative person?
- What’s your reaction when someone tells you something isn’t possible?
That last one is the tell. In my world right now, almost everything is possible. If a candidate accepts “it can’t be done” as a valid answer, that’s a mindset I don’t want on the team. The people I want are the ones who treat that sentence as a personal challenge.
Creativity isn’t a skill you learn in a course. It grows. You’re curious, you try things, you discover something that makes you more curious, you try more things. I’ve started building on my Raspberry Pi — ePaper displays, GPIO sensors, humidity, proximity, temperature. A year ago I got a basic IP-address readout working on an ePaper screen with ChatGPT. Now I can do a hundred times more, just by talking to Claude Code. And every sensor I plug in suggests three more things I want to try. The creativity compounds.
The backlog problem solved itself
Two months ago my team’s ticket backlog had items in it from a year ago. Today I have five or six tickets, all created in the last few days. The reason is embarrassing in hindsight: I sat down, looked at the old backlog, and just closed it. Twenty to twenty-five tickets in two or three days. Tickets that had been “blocked on developer time” for over a year disappeared in an afternoon.
That gives me an unreasonable amount of peace of mind. With no backlog hanging over me, I can spend energy on what people actually need rather than triaging what we already have.
Dashboards: numbers vs stories
I went through our BI dashboards recently and deleted most of them. Before AI, we crammed numbers into dashboards because creating one was expensive and we wanted to justify the cost by including everything. There was no story behind the numbers. No connection between them. No clear action a stakeholder could take from looking at one.
So I deleted them. What’s left are story-based dashboards: a top-level KPI, a path you can follow when something looks wrong, drill-downs that lead you from “this number is bad” to “this specific thing went wrong, this person needs to be informed.” Each dashboard tells a story end-to-end. That kind of thoughtful design wasn’t possible before — not because we couldn’t think of it, but because we didn’t have the time to actually build it. Now we do.
Anyone can be a builder
This is the change in mindset. From the perspective of a data engineer (or industrial engineer, which I also am) inside a software company, the role has fundamentally shifted. And it isn’t only my role. Sales, marketing, HR — anyone — can now ask AI for developer time and ship something real. The only requirements are knowing what you want and having the patience to iterate. You’re a product owner, and your developer never sleeps.
There’s no excuse anymore. If you’ve had an idea for a tool, a side project, a small business — the cost of trying it is one evening. The senior developer barrier is gone. The “I don’t have time after 6pm” barrier is mostly gone too, because you don’t need 200 hours, you need 10.
What’s left is whether you’re curious enough to start. That’s the whole game now.
