Thoughts...

Just 18 months ago, I pretty much gave up on AI-assisted coding. The LLM wrote some pretty bad code, and I could work faster and better without it "interrupting my chain of thought". I switched back to manually coding.

Then at the start of the year, I tried again, and this time the code generated was so much better than before. I started using it, and used it more. I created workflows, and I got some training from AWS (about their AIDLC). I changed my way of working — now I use the LLM basically for everything.

The LLMs have got a whole lot better at assisting us. Anthropic's 2026 Agentic Coding Trends Report describes a shift from single AI assistants to coordinated agent teams, with engineers moving from writing code to orchestrating the systems that write it. (Which is how we use it at work.) However, AIs getting better does not necessarily mean we'll be replaced. Not if we adapt.

Software engineers won't vanish — the role is splitting into new ones:

The first two get jobs in companies like Anthropic. The last two, well, I could fit into either of those roles.

I've come late to this new era of the cognitive assembly line (moving from hand-typing code to orchestrating digital factories) and the sovereign developer era (where developers rule over a team of AI coders).

Am I almost too late? Almost.

I am rapidly accelerating my thinking and adjusting very quickly. Otherwise, I will be a casualty, swimming in a sea of laid-off software developers who were too stubborn to take good counsel and adapt.

There are real risks with AI, but the main one is that some people (other than us) get power (through better use of a better AI), and control us.

My advice: spend as much time as you can in side hustles learning how to use AI. Convince your company to adopt "AI First". If you don't have that level of influence, jump ship ASAP to a company that does.