What Kodak, Calculators and Telephone Operators Tell Us About AI and Development
Every generation believes its technology shift is unprecedented. History disagrees. Eight parallels that reveal what's coming for software developers.
A 10-Part Series on AI & Development
Ten data-driven articles on what AI is actually doing to software development — the productivity claims, the job market, the code quality, and the hype.
Every week brings a new headline: AI will replace developers, AI makes developers 10x faster, AI is destroying code quality. Which is it?
I started this series because I was tired of the noise. Vendor benchmarks claiming 55% productivity gains. Independent studies finding a 19% slowdown. Startups shipping 95% AI-generated codebases. Open-source maintainers drowning in AI slop. The signal-to-noise ratio had collapsed, and I wanted to find out what the data actually showed.
Over ten articles, I dug into the research, the employment numbers, the code quality metrics, and the economic history. I talked to developers who love these tools and developers who are terrified by them. I read every study I could find and tracked who funded it.
The answers aren't simple. They're messier and more interesting than either the doomers or the boosters want you to believe. AI is genuinely transforming software development — but not in the ways most people think, and not at the speed the hype cycle suggests.
Each article was researched and written using AI — mainly Gemini and Claude — under my direction and supervision. I prompted, reviewed, fact-checked, and approved every word. The editorial perspective and any mistakes are mine. AI was the tool; I take responsibility for the result.
— Pieter
Every generation believes its technology shift is unprecedented. History disagrees. Eight parallels that reveal what's coming for software developers.
GitHub says 55% faster. Google says 21%. An independent study says 19% slower. Here's what happens when you read past the headlines.
Junior dev hiring is down 67%. Companies are saving money today and building a crisis for 2030. Here's the data — and the history that explains why this is a terrible idea.
When amateurs got access to typography, design quality collapsed. They called it the 'ransom note effect.' We're watching the same thing happen to code.
In February 2026, AI agents wiped $2 trillion off software stocks. Salesforce cratered. Atlassian plunged 35%. Is this the end of SaaS — or a panic that'll look ridiculous in hindsight?
Sam Altman has a betting pool. Dario Amodei gives it 70-80% odds. Solo founders are 38% of all new startups. But has anyone actually done it yet?
AI writes code fast. It also writes code that's measurably worse by every metric we have. Here's what 211 million lines of data tell us.
Spreadsheets killed 400,000 bookkeeper jobs and created 600,000 accountant jobs. ATMs doubled the number of bank tellers. What happens when AI makes code nearly free to produce?
AI brings the cost of average work to zero. It also creates a massive ceiling of mediocrity. The new scarce resource isn't code — it's knowing what's worth building.
curl paused bug bounties. tldraw blocked all external PRs. Godot maintainers are demoralized. An AI bot published a blog post attacking a developer who rejected its code. Open source has an AI problem.
Disrupting AI is an independent editorial project examining what AI is actually doing to software development — the productivity claims, the job market shifts, the code quality data, and the economic forces at play. Ten long-form articles, published between December 2025 and February 2026.
This series was researched, written, designed, and published using AI — primarily Gemini and Claude — under the supervision of a human (Pieter). Every article was prompted, directed, reviewed, fact-checked, and approved by Pieter. AI did the heavy lifting; editorial judgement and final responsibility remained human. We believe in full transparency: this is a project about AI, made with AI.
The information presented in this series is based on publicly available research papers, industry reports, interviews with practitioners, and our own analysis. We cite primary sources wherever possible. Where we express opinions, we label them as such. Our goal is accuracy over speed — we'd rather be right than first.
This project is not funded by, affiliated with, or endorsed by any AI company, research lab, or technology platform. We have no financial interest in any AI product or service. Our only bias is toward clarity.