The "Software Industrial Revolution" is a thesis that AI-driven automation of software production will mirror the original Industrial Revolution — dramatically reducing costs, exploding supply and demand, and reshaping industries far beyond tech. Just as the Spinning Jenny parallelized thread production and crashed textile prices by 90%, AI coding agents are making software orders of magnitude cheaper to build, threatening existing monopolies and enabling an era of software abundance.
The analogy
Before the Industrial Revolution, clothing was handmade, expensive, and scarce — most people owned two outfits. The Spinning Jenny (1764) parallelized spinning, crashing cotton cloth prices by 90% over 50 years and triggering cascading changes across manufacturing, energy, and markets.
Software engineering today is similarly labor-intensive and expensive. The cost of building software has driven the VC-funded "grow at all costs" playbook, enshittification of products, and tech monopolies. AI coding agents represent the Spinning Jenny moment for software.
Core claims
- Software production is being rapidly automated — AI coding agents reached an inflection point in late 2025, suddenly "just working" for complex tasks
- Software will become dramatically cheaper and faster to produce — enabling radically different business models beyond the VC-funded monopoly playbook
- Non-engineers can now produce software — researchers, doctors, small business owners can conjure custom tools on demand
LLMs are becoming an operating-system layer
Karpathy's "Software 3.0" framing treats LLMs less like chatbots and more like an operating-system layer for software work. In that model, the prompt behaves more like a program, the context window acts like working memory, and tools plus loops become the control plane. That is a different mental model from "ask a question, get an answer," and it is the one that better explains why agents and copilots are changing how software is built.
Real agents are still early
The strongest current demos are still not the fully autonomous agents people sometimes imagine. The nearer-term reality is bounded automation wrapped around ordinary workflows, with AGI-like capability diffusing into daily life gradually rather than arriving as a clean threshold event. That does not weaken the industrial-revolution analogy; it suggests the transition will be broad, messy, and easy to miss while it is happening.
Implications
- More software, not less — just as cheap textiles created thousands of specialized clothing companies, cheap software will create an explosion of bespoke software across every industry
- End of the VC era — founders won't need to trade massive equity for engineering costs, shifting capital toward physical-world applications
- Threat to monopolies — the high cost of complex software (like Epic's healthcare monopoly, with 42% of US acute care hospitals) has functionally blocked competition; cheaper software production changes this calculus
- Software engineering evolves — hand-coding becomes less relevant, but understanding domains, managing complexity over time, and modeling the interplay between software and the real world become more important
The Epic example
Epic holds 42% of the US acute care hospital market and manages ~55% of all hospital beds. Mass General Brigham spent $1.2B, Mayo Clinic $1.5B, and Kaiser Permanente $4B on Epic rollouts. The astronomical cost of building competing software at that complexity level has blocked alternatives — a dynamic that cheap AI-built software could disrupt.