Over the last four months I’ve increased my use of AI for web dev work pretty dramatically. I use it every day at work, as do most of my coworkers. I don’t think it’s disputed anymore that AI-assisted development can save a lot of time and effort. The question is, what do we do with all that surplus?

The standard narrative in the industry is to use it to go faster. Ship three features a day instead of one. Get a half-dozen agents working in parallel. 10x your productivity. If you don’t move fast, your competitors will!

I get it. Speed is good and it has benefits beyond just finishing more stuff per unit time. Speed matters!

But we live in an era of extreme consumer choice, especially in the world of software. Even pre-AI, there were dozens of good email clients, to-do lists, fitness apps, and analytics tools, all chock-full of useful features at reasonable prices. What is “going faster” going to give us? Even more options, with even more features?

Apparently yes. In the past quarter, iOS app store submissions have risen by 84 percent, and Hackernews, “ShowHN” submissions have nearly tripled. We live in an era of software abundance.

But as software grows increasingly abundant, attention grows increasingly scarce. Attention was scarce even before AI but it’s going to get worse. Consumers have more choices than ever, and increasingly those choices are becoming undifferentiated as the industry uses AI to produce them.

What if we did the opposite?

Instead of using your AI surplus to go faster, what if you used it to go deeper?

What if, instead of building ten good features, you built two amazing ones?

What if, instead of blasting out a dozen landing pages, you poured your attention and care into a home page that knocked someones socks off—one that could not have been built by an AI, because nothing like it could have been found in the training data?

What if you doubled-down on performance? On accessibility? On user-experience? On creating a strong personal touch?

It would stand out! That matters, arguably, even more than speed.

Striking a balance

The more I thought about this, the more I recognized it as the same ol’ quality vs quantity debate, where it pays to strike a balance:

• If we pour all our surplus time into quality, we end up hurting quality by failing to iterate¹.
• If we pour all our surplus time into quantity, we end up hurting speed as tech debt accumulates².

But what’s the ideal balance? Is it 90% quality, 10% speed? 50% - 50%?

How would you spend your AI surplus?