Key Takeaways

  • Coding assistants now write up to 95 percent of routine code, but David Sacks points out that developer demand is climbing because that remaining 5 percent addresses a vastly larger market.
  • Ben Horowitz invokes Jevons paradox to show that plunging software unit prices will trigger a net explosion in total software creation and enterprise consumption.
  • Software development is shifting away from generic mass-market tools toward bespoke applications tailored to specific company workflows that previously were not cost-effective to build.
  • Raw code no longer serves as a defensible moat. Horowitz and Sacks argue startups must build defenses around proprietary data, vertical workflows, and continuous product iteration, citing ElevenLabs.

Cheaper code generation triggers higher aggregate demand

When software becomes almost free to write, panic sets in around developer jobs and enterprise margins. If automated tools can write software instantly, the initial reaction is that software prices collapse to zero and engineering teams get slashed.

Sacks and Horowitz view this panic as an economic misunderstanding. In the nineteenth century, English economist William Stanley Jevons noticed that when steam engines became more efficient, coal consumption rose rather than fell. Cheaper power made coal practical for dozens of new industrial tasks, which consumed far more fuel overall. Horowitz applies that exact principle to modern software engineering. Software unit prices will plunge, but the volume of software built will grow dramatically.

“Prices software prices will definitely go down,” Horowitz noted. “Um but the amount of software is going way way up. It's the it's the Jevans paradox of software.”

Historically, building software required millions of dollars in upfront engineering salaries. Because of that high floor, software companies could only justify building products aimed at mass horizontal markets. A company could justify an ERP or a global CRM system, but they could never justify building a dedicated billing application for a fifty-person logistics brokerage in Ohio.

Cheap code generation flips this calculation. Sacks explained that coding assistants enable teams to tackle use cases that were once economically irrational: “The value of these coding assistants is that now you can create software for highly bespoke use cases. Whereas software used to be created really just for the mass market.”

This shift explains why developer employment has not collapsed alongside rapid model adoption. “Now we're seeing that actually demand for coders is increasing even though 95% of the code's written by coding assistants,” Sacks observed. “And the reason for that again is well first of all you still have that 5% and that 5% is of a much bigger pie.”

Defensibility moves from raw code to proprietary data and workflows

If any developer can generate working code in minutes, writing code is no longer a durable competitive advantage. A founder cannot charge enterprise buyers a higher fee simply because their software solves a technical problem; competitors can replicate raw software interfaces almost overnight.

Startups must shift where they look for pricing power. Sacks pointed directly to proprietary information: “You got to think about what are the moes in a business data moes are a big one if you have data that other people don't have.”

When code is commoditized, value accrues to the data that guides the software and the custom operational workflows built around it. Horowitz pointed to voice AI company ElevenLabs as an example of a team that stays ahead by refusing to rely on an early technical lead. Instead of hoping raw models remain proprietary, they constantly ship product improvements, refine vertical integrations, and adapt their business to stay ahead of commodity models.

“So from a startup's point of view, you've got to evolve,” Horowitz said. “ElevenLabs does many many things to make their product better over the years since that's example that you brought up.”

Survival in an era of automated code generation comes down to two moats: owning unique data that third-party scrapers cannot touch, and embedding so deeply into an enterprise customer's bespoke operations that a generic replacement tool cannot dislodge you.

What to Do With This

Map the five most tedious manual steps in your customer's weekly workflow that you previously ignored because custom engineering was too expensive. Use your next sprint to ship a bespoke automated tool for that single workflow, and bind it directly to private data from their internal databases.