If you earn your living building or selling software, every demo of AI writing code in minutes can feel like watching your price tag shrink. If you run a small business or a small nonprofit, the same demo may look like relief. Both reactions are reasonable.
I'm optimistic that AI will make useful software much cheaper. I'm also building toward more affordable software with Wrk.Dog, so I have a stake in that outcome. Weigh my view accordingly.
We already have a precedent for what shared, low-cost software can unlock. Harvard Business School research estimated it would cost firms roughly $8.8 trillion to rebuild the shared open source software they rely on if they had to write it themselves. That is a replacement-value estimate, not yearly savings or GDP. And free to download has never meant free to run. Someone still has to keep it secure and answer the support calls.
Now picture a hypothetical two-truck plumbing business that pays a local developer for scheduling and invoicing software. Suppose AI-assisted development eventually lets the owner obtain a dependable alternative for substantially less. A weekend prototype alone would not establish that it is ready to run the business. The plumber could put the savings toward other needs. If several customers make the same switch, the developer could face a sharp loss of income.
Cheaper tools don't pay someone's rent. The plumber's savings are spread out and quiet. The developer's loss is concentrated, and it lands on one kitchen table.
Here is the strongest pushback, and I share part of it. Lower prices tend to expand demand, new kinds of work show up, and shielding every existing seller would slow progress for everyone else. I lean free-market, and I don't want to hold the plumber back to protect the vendor. But "new work will appear" is a claim about averages over years. It says little about whether this developer covers rent next spring.
So here is a question I think we should ask out loud: when the gains are broad and the losses are concentrated, what should a society decide to do about the gap? I don't have a finished policy answer, and I'd be wary of anyone who claims one this early. I do think it's bigger than any single company can settle.
If you buy software, budget for the parts that aren't code, especially support and security. If you sell it, notice that those are the parts your customers will keep needing. The developer in this example might sell fewer builds and more ongoing care to clients they already know well. That's not guaranteed income. It is a direction worth testing now.
AI-assisted editing. Published with Rob Price's approval.