You've been thinking a lot about where AI is actually landing in enterprise right now. What's shifted for you recently?
Where Stu had believed the machine could never get to the quality, control, and correctness we need in the enterprise, he now believes software engineering is the first domain where it will. That's a significant statement from someone with his level of mastery. It's been top of mind ever since.
What specifically changed his view?
We used to assume that AI slop would appear everywhere and we'd be cleaning it up, that humans would be the quality control layer sitting above a machine that could produce volume but not discern. Stu came back from working with Opus 4.6 and said: “The machines are going to clean up the human slop in the codebase.” Not the other way around.
Where for 3 years we believed the machines would never reach the complexity, correctness and quality at scale demanded in enterprise, it was now clear they had. And that's a fundamental inversion of the assumption most transformation programmes are still being built on.
How did we get here so fast?
What we've watched is a transition from pattern recognition to semantic understanding, and it happened almost overnight at an enterprise level.
Developers have had machine-assisted coding for years. Auto-complete, bracket closing, snippet tools. In law, there were discovery tools and OCR that could chew through millions of documents and keyword search across them. These were useful tools used every day. They were also fundamentally limited.
Then the machines stopped matching patterns and started understanding deeper connections, this seems like meaning. The implications of that are still playing out and the organisations struggling the most are the ones that underestimate how complete that shift is.
What does that mean for the talent pipeline?
This is the question everyone is asking and is fearful of answering honestly. If you look at the data on software engineering roles over the last two years, there's been a significant collapse in programmer roles, base coders, entry-level, particularly in the US. What hasn't changed? Systems architects. Designers. AI-fluent engineers. And the signals now point to 75,000 more roles of that profile coming into the market.
So it's not the decimation of the discipline. It's a transition in what's required.
But that transition creates a real problem: how do you build the next generation of people like Stu? He can look at AI-generated code with deep discernment, he knows when something is correct, when it's nearly there, when it's plausible but saying nothing. That judgment is hard-won over years spent in the depths of the work.
The traditional path for juniors was painful but formative: bashing your head against the wall until you got a piece of code working. In law it was reviewing hundreds of contracts that would never get signed. We don't yet know what replaces those pathways but something will, it always does.
So where does the answer lie?
Mentorship. That's probably the most complete answer I've found to the whole talent pipeline question, and I've been thinking about it a lot.
Specifically: mentors who aren't just open to these new tools, but at the very edge of them. Who've already grappled with the limits, understood the edges, and can guide the next generation through a learning journey that is different from the one that made them.
Stu is exactly that person. He founded the London React community when Facebook first introduced it. Same with Node.js and with Rust - championing the technology horizon, applying the lens of is this enterprise grade? before anyone has caught up. The fact that he did the 180 on agentic coding is a high level of craft informing and honest new assessment. And that's precisely what good mentorship in this new world looks like. What's beginning to show up is a clear distinction between those who are open to these tools and those who are resisting them. The ones who are open are becoming the natural leaders for the next generation. The resistors, however experienced, risk becoming obstacles.
And for the organisations trying to lead AI and transformation programmes right now?
Most of them fail not because the technology doesn't work, but because the humans around it haven't changed how they think. Consultants get brought in with a playbook. The playbook made sense when the world looked a certain way. But the world doesn't look that way anymore.
The organisations that will come through this well are the ones asking not just can the machine do this? but what does it mean that it can? Not just how do we implement AI? but how do we rebuild our talent model, our mentoring culture, and our definition of quality around what's actually true now?
Listen to the full episode on the AEPS podcast. To talk about how Red Badger can help your organisation navigate this: https://red-badger.com/