Pace layers as a coping mechanism

I've been working with software for twenty years now, and I can't remember a time where the pace of change comes anywhere close to what's going on in the industry .

Tech and software has always changed, of course, but never as fast, as deeply or as widely as things have moved in the last 12 months. It used to be that you could reckon on your current skillset having a half life of five years or so; and for some technologies there was always the option of riding out the unfashionable legacy period until your skills became so niche they commanded a premium again.

If it all feels overwhelming, you're not alone.

There's a mental model that can help to keep pace with this, though. It's called "pace layers", and it was originally devised to think about how the use of buildings and structures change over time.

Some things can and do change quickly, like the arrangement of furniture in rooms. Some things adapt more slowly - it's a lot harder to change the layout of rooms and spaces inside a building once the walls have gone up. And some things are more or less fixed - adding additional floors, or changing the roof are major undertakings where it may actually be cheaper to pull the building down and start again.

That same model applies well to the tech industry.

The fastest layer is AI technology, and this is the the one that's changing frantically. Agent harnesses pop up daily. You could spend the entire day wading through your inbox if you subscribed to only one in ten AI-related newsletters, and podcast minutes are being created faster than you can listen to them even at 2x speed.

The next pace layer is the platforms, systems and processes involved in building and deploying software into production use. There's a large tech component in this, which is the component that gets the focus - but the human aspect is just as, if not more, important, and often gets overlooked

Think of the processes that you need to keep the development effort from spiralling into chaos. There's tooling, but if the team isn't aligned on how the tooling gets used, it's worse than useless.

The third is everything around what the software is FOR - the interfaces between the customer, the product and the tech operation. That's the slowest part, because you're trying to serve customer needs in a way that enables long-term usage of your product.

You probably can't keep up with everything that's happening in the fastest layer, and it could be actively problematic if you try to chase the latest new tool before they've had a chance to be stress-tested by serious adoption.

There's a huge amount of change needed in the middle layer, because getting the best out of agentic development means changing aspects of how we build software.

But I'd argue that the lowest layer is where we still need the human approaches that we've always used, and AI doesn't change that. Most software is still used by humans, despite the agent hype - so focussing on just the fast, sexy, exciting and bewildering agentic hype risks ignoring your customer base at your peril.