AI Won't Kill Jobs — It Redistributes Them

Every other week someone posts a chart predicting AI will erase the engineering profession. The math behind those charts assumes investors will quietly take the savings and go home. They won't.

The boring economics

Picture a company with one product that needs four engineers. Today AI cuts the per-product engineering cost roughly in half. The popular conclusion: fire two engineers.

The actual conclusion, in any company with a competent CFO: keep four engineers, ship four products.

Capital allocation does not shrink because per-unit cost dropped. The fund still has to deploy. The board still wants growth. The competitor across the street is also getting cheaper, so standing still is a loss. The rational move is more shots on goal, not fewer headcount.

This is Jevons paradox, applied to talent instead of coal.

We've already seen this movie: ecommerce

Rewind to 2005. A custom ecommerce site — cart, checkout, inventory, payments, admin panel — was a 3 to 6 month build and cost roughly $30,000 to $80,000. Only serious brands with serious budgets could play.

Then Shopify (2006) and WooCommerce (2011) commoditized the stack. Today the same brand can launch on Shopify for ~$30/month plus a theme, live in a weekend. Per-store cost dropped roughly 100×.

If the doom narrative held, ecommerce developers should have been wiped out. Instead:

• The number of live online stores went from hundreds of thousands to tens of millions globally.
• A new job category appeared from nothing: Shopify theme developers, app developers, agencies, headless-commerce specialists. The Shopify App Store alone supports thousands of full-time engineers.
• Total dollars spent on ecommerce engineering globally is higher today than in 2005, not lower — even though the per-store cost collapsed.

Cost per unit fell 100×. Number of units rose more than 100×. Net work expanded. That is the entire pattern, and it is the same pattern AI is running on engineering at large.

Why the layoff headlines feel real anyway

Two things are happening at once and they get conflated:

1. Reshuffling. Roles are moving from "ship one product slowly" to "ship four products in parallel, each smaller in scope." Same person, different shape of work. From the outside this looks like churn.
2. Macro tightening. 2024–2025 were rate-driven contractions. Lots of cuts got branded as "AI efficiency" because that's the cleaner story for an earnings call. Most of them were just the cost of capital catching up.

If you can't tell those two apart, every layoff looks like an AI layoff.

What actually changes for an engineer

• Surface area widens. You own more of the stack because the cost of touching unfamiliar code dropped. The "I only do backend" defense gets weaker every quarter.
• Taste matters more. When generation is cheap, the bottleneck moves to judgment — what to build, what to cut, what to keep simple. Generating code is no longer the scarce skill. Knowing which code shouldn't exist is.
• Throughput expectations rise. A team of four shipping one product becomes four people each owning a product. More autonomy, more ambiguity, more directly observable output.

None of that is the labor apocalypse. It's a different job, harder in some ways, easier in others.

The headline that won't go viral

Investors deployed the same capital this year. Per-product engineering cost fell. They funded more products.

That's the story. It doesn't fit on a doom thumbnail, which is partly why it loses to the alternative.

The engineers who do well in this cycle aren't the ones who out-type the model. They're the ones who can pick the right four products to build.