Navigating the Software Engineering AI Apocalypse

I had a shocking realisation about software engineering this week. Once I understood the implications, I stopped being afraid of the future, and now I'm genuinely excited about what's coming.

A warning first: this isn't for the faint-hearted. Some of it may sound inflammatory, and some of it may be upsetting. But that's the nature of change. I would rather anticipate what's going to happen and plan for it than ignore it and get hit with a massive shock in 18 months' time.

How I got here

Since February I've been obsessed with AI agents. I've used them to build AI-native apps that run my training business, starting with OpenClaw, then Claude Code, then Cursor Cloud Agents, and now Grok Bot.

Somewhere along the way, I lost all interest in software.

Software engineering as we know it is dead

The frustration, the challenge and the struggle are all gone. It's easy now. It's boring. It's no longer fulfilling, and that's very hard for me to admit after a career spent writing code.

I've started batching up work so that I can get a week's worth of work done in a single hour. That's when it hit me.

My profession now comes down to three skills

1. Understanding
2. Pattern recognition
3. Decisions

That's all I do now. Here's what each one looks like in practice.

1. Understanding

My day comes down to processing a lot of information quickly and building a mental model of what it means. If I skim, I make mistakes and have to backtrack. When I ask my agents to summarise the work we've completed each day, I'm aghast at how much I've processed.

2. Pattern recognition

This is the ability to recognise patterns and act on them quickly. A bad pattern is when an agent's language drifts from what I expected, for example vague wording where I expect a locked-in structure. A good pattern is sharing what worked before with the agents so they can reuse it. It's intuitive, and it's fast.

3. Decisions

Batching work means making a lot of very rapid decisions across different contexts, and this is the part that wears me down. Switching context causes decision fatigue. I can make about 12 rapid decisions before I have to stop and give the batch its GO.

What this means for the future

Agents will write 90% of all code. Engineering teams of the future will focus on three questions:

1. Do you understand?
2. What are the patterns, both good and bad?
3. Can you make decisions?

When I wrote that list down, I realised it sounded a lot like FX trading.

Years ago I traded FX. My strategy was simple: analyse as much information as I could get my hands on about what moves the market, look for patterns, and make decisions. I quit in the end. I'd wake up in cold sweats from dreams of falling graphs and margin calls.

Sound familiar? So what now?

Software engineers will become business owners

Software engineering will look very different in 18 months. Very few people will stay in the industry for the passion, the problem solving or the fulfilment. Most will leave. A lot of software engineers will become business owners, and they'll move a thousand times faster than slow, old-style corporates.

If you want to survive the AI "appocalypse", plan to become a business owner. Learn about value creation and market analysis. Learn how to create hooks, stories and offers. Most importantly, learn to recognise patterns and make decisions even when you're uncertain.

If you do, you win.

This post started life as a thread on X. If you found it useful, share it, or tell me where you think I'm wrong.