"When the cost of a powerful input falls, Jevons Paradox reigns supreme."

In 1865 an economist named William Stanley Jevons noticed something weird about coal. Steam engines kept getting more efficient. Every year they did more with less. So Britain should have been burning less coal.

Instead, Britain burned more.

Cheaper coal meant coal made sense in places it never had before. New industries picked it up. Demand climbed faster than efficiency did. Total consumption went up.

That's the Jevons paradox. Make something cheaper and you don't necessarily use less of it. Sometimes you use a lot more.

Software is going through the same thing right now.

A client example:

We built an agentic pipeline that processes 200+ page regulatory PDFs across all 50 states, with a chatbot on top so someone can just ask it a question.

Price that project the old way. Collecting the documents. Parsing them. Normalizing fifty jurisdictions that all disagree with each other. Structuring it, maintaining it, building an interface. A small team, most of a year, and a maintenance burden that never ends.

That’s the version that never gets approved because it's a year of engineering for a capability nobody can prove will pay for itself.

Here's the interesting half. Nobody at the client had to make that bet. The economics changed enough that the question stopped being whether it was worth a year and started being what else could get built with the year we didn't spend.

Cheaper is not the same as better.

Dropping the cost of producing code does nothing to the value of the code. You can now generate garbage at a speed that was previously unavailable to you. More tech debt. More review. More systems nobody asked for that somebody has to own at 2am.

So the constraint moves. It doesn't disappear.

For twenty years the hard question was can we build this. Engineering capacity was the wall. Everything queued up behind it and prioritization was mostly triage.

The wall moved. The new question is whether the thing should exist at all.

That judgment is broader than a simple ROI calculation. Teams have to weigh the severity and frequency of the problem, who benefits, how well the solution fits the product and strategy, what new complexity it introduces, how costly it will be to maintain, what happens when it fails, and whether automation improves the experience or merely makes it cheaper. Some ideas are worth pursuing because they create strategic options, improve resilience, reduce friction, or enable capabilities that were previously impossible to justify. Others should still be left alone.

Jevons is holding up.

It’s still early, but as Jevons holds, we're about to build a lot more software. Internal tools that never cleared the bar. Automation for workflows too weird to justify. Integrations too bespoke to touch. Small product bets that died in a spreadsheet… you get the point.

Most of it will get built. Only some of it should.

Go pull up whatever you shelved in 2023 for cost reasons. Run it again. IF it’s the right thing to build, reach out. Ravn will help you bring it to life.