AI is taking my job
It feels like a decade of progress was made over the last few months in tech. We’ve moved past just hitting 'tab' to let an AI auto-complete a line of code. Now, we have sophisticated agents using frameworks to reason, act, and use tools, writing all of your code for you. The leap is so massive that even elite engineers like Linus Torvalds, creator of Linux, are converts. No one writes code anymore.
"It started out as my typical "google and do the monkey-see-monkey-do" kind of programming, but then I cut out the middle-man – me – and just used Google Antigravity to do the audio sample visualizer."
- Linus Torvalds
torvalds
Time for an awkward question... If it’s getting this easy to build things, do tech companies still need tech leadership and software engineers?
Look, I get it. Tools like Lovable, Claude Code and GitHub Copilot are helping non-technical people build genuinely impressive software. I’ve seen it. But there’s a dangerous truth hidden here.
There is a level of complexity you simply cannot manage if you don’t understand good engineering principles. A lot of my work recently comes from entrepreneurs who got pretty far using AI coding tools, hit a wall, and realised they just don't know what they don't know.
I have to constantly remind myself that building software is not universally simple. I have a set of hard-earned skills and experience that I take for granted.
When experienced software engineers build a product, even with AI, they start with a mental model. They consider system design, trade-offs, tech stacks, and edge cases. These early decisions dictate the security, reliability, and scalability of the entire system.
When you skip that step, bad things happen:
- Fragile codebases: AI tools often lack full-project context. They can easily make a sub-optimal change that causes unintended side effects somewhere else.
- Poor user experience: Unreliable, unscalable systems frustrate users. In a market flooded with new digital products, you rarely get a second chance.
- Critical security flaws: It’s incredibly easy to accidentally leave database tables open to the internet or bypass authentication. Think of it like buying a heavy-duty steel vault, but leaving the combination on a Post-it note on the door.
Just look at the numbers. Wiz recently reported that 1 in 5 applications created on Lovable are insecure, leaking data due to missing or incorrectly configured auth (https://www.wiz.io/blog/common-security-risks-in-vibe-coded-apps). That's a massive liability if you ever plan to handle GDPR compliance or aim for serious enterprise standards like SOC2 or ISO27001.
This isn’t a software engineer problem. It affects every industry. You can now buy amazing go-to-market tools for content and scheduling. They might make you highly productive, but they don't make you a GTM expert.
Asking Gemini to draft a legal contract doesn’t make you a lawyer, and owning a Formula 1 car doesn’t make you a race driver.
Domain knowledge is still the differentiator. These are just tools.
So, am I worried as a software engineer? Not really. My role is shifting, but honestly, I’m just excited. I can build things I never had the time to tackle before, using my prior experience to validate and guide the AI.
Writing the code was always only ever part of the job. It just takes less time now.
