In the early days of the commercial internet, starting a software company required serious capital—not just for people, but for infrastructure. Servers. Racks. Data centers. Hardware alone could consume funding rounds before a product even shipped. If you couldn’t raise significant money upfront, you couldn’t compete.

Virtualization and the cloud changed that. Hypervisors abstracted away the hardware. AWS and its successors turned capital expense into operating expense. You could start with a credit card instead of a purchase order. Infrastructure stopped being the gating constraint. More of your resources could go into product and people.

Now the cost structure is shifting again—this time at the level of human output.

AI-assisted development is compressing the amount of labor needed to build software. A developer working with a strong model can prototype, debug, and iterate far faster than before. The gains are not theoretical; they show up in daily workflow. Small teams can now ship what once required much larger ones.

At the startup level, this favors tight, focused groups that move quickly with AI as leverage. At the corporate level, the pattern looks different. Many organizations are treating AI primarily as a cost-reduction tool—using it to maintain output with fewer people rather than to expand what their existing teams can produce. That may be a transitional phase, or it may signal a deeper shift in how firms think about knowledge work.

Either way, the dynamic is familiar. When the cost of production drops sharply, power redistributes. Incumbents lose some advantages. New entrants gain ground. The transition is uneven and often painful.

Schumpeter called this creative destruction—the continuous replacement of old structures with new ones. It’s not a smooth process. Workers with valuable skills can find those skills suddenly cheaper to replicate. Entire roles can narrow or disappear. The disruption isn’t hypothetical, and minimizing it doesn’t help anyone prepare for it.

But the same forces that unsettle existing positions also open doors. When production costs fall, experimentation rises. More people can try, build, and ship. Access expands, even if outcomes remain unequal. The field doesn’t become level—it rarely does—but it does change.

We’ve seen this pattern before: the printing press, industrial machinery, electrification, the internet, the cloud. Each lowered the cost of creating and distributing something valuable. Each triggered resistance, overreaction, and real harm alongside genuine progress. The institutions that emerged afterward were not the ones that came before.

We’re in another one of those intervals now—when the old rules are weakening and the new ones aren’t fully formed. That uncertainty is uncomfortable but not unprecedented.

The pressure to adapt isn’t new. It’s structural. It doesn’t pause while we debate it.

It never has.