The 19% slowdown nobody wants to talk about
Experienced developers are 19% slower with AI tools — and they don't even know it. The data says the productivity revolution isn't about faster code. It's about fixing the system around the code.
Duration: 2:38 | Size: 3.0 MB
Here’s a number that should make every engineering leader uncomfortable: experienced developers using AI tools are 19% slower than when they work without them.
That’s not a blog post hot take. That’s a randomized controlled trial from METR, published last July, with real open-source developers working on codebases they’d maintained for an average of five years. The developers themselves? They thought AI sped them up by 20%. The gap between what they felt and what happened is almost poetic.
And it gets worse at scale. Faros AI’s 2026 report — 10,000 developers, 1,255 teams — found that high AI adoption teams completed 21% more tasks and merged 98% more pull requests. Sounds great until you notice that PR review time ballooned 91%, bugs per developer rose 9%, and average PR size jumped 154%. When they looked at DORA metrics — the industry standard for delivery performance — there was no significant correlation between AI adoption and better outcomes at the company level.
So we’ve built faster typewriters and are wondering why the novels aren’t better.
The uncomfortable truth is that most teams are using AI to accelerate the part of software development that was never the bottleneck. Writing code was already the easy part. The hard parts — understanding the problem, reviewing changes, testing edge cases, coordinating across teams, sharing institutional knowledge — those are exactly where AI-as-autocomplete adds nothing. In some cases, it’s actively making them worse by flooding the review pipeline with larger, noisier diffs.
But there’s a crack of light. Atlassian’s research found that the 4% of companies actually seeing transformative results from AI aren’t optimizing individual developer speed. They’re using AI to fix the system: how knowledge moves through the organization, how decisions get made, how teams align. These companies are nearly twice as likely to report significant efficiency gains.
The pattern is clear enough to make a bet on: the next era of AI productivity won’t come from better code generation. It’ll come from AI that makes the humans around it more coherent. Less “write this function for me” and more “here’s what you need to know before you review this PR” or “this decision was already made three months ago — here’s the context.”
Individual speed without systemic change is just faster production of things that take longer to review, test, and ship. That’s not productivity. That’s a speedier treadmill.
Why customer tools are organized wrong
This article reveals a fundamental flaw in how customer support tools are designed—organizing by interaction type instead of by customer—and explains why this fragmentation wastes time and obscures the full picture you need to help users effectively.
Infrastructure shapes thought
The tools you build determine what kinds of thinking become possible. On infrastructure, friction, and building deliberately for thought rather than just throughput.
Server-side dashboard architecture: Why moving data fetching off the browser changes everything
How choosing server-side rendering solved security, CORS, and credential management problems I didn't know I had.
The work of being available now
A book on AI, judgment, and staying human at work.
The practice of work in progress
Practical essays on how work actually gets done.
The headcount lie
The assumption that work scales with people is so embedded in how organizations think that questioning it feels like questioning gravity. But one operator just ran ten parallel operations in a single day. The unit of capacity isn't the person. It's the decision-maker.
AI and the Götterdämmerung of Work
Work is dead. And we have killed it. AI didn't defeat the myth that human value comes from reliable output — we built the systems that exposed it. What comes next isn't replacement. It's revaluation.
Everything pointed at ghosts
Most organizations are measuring work they stopped doing years ago. The dashboard is green. The reports are filed. Nobody realizes the entire apparatus is pointed at ghosts.
Manual fluency is the prerequisite for agent supervision
You cannot responsibly automate what you cannot do manually. AI agents speed up work for people who already know how to do it. They do not replace the need to learn the work in the first place.
Your process was built for a different speed
When work changes velocity, governance systems don't just fall behind. They become theater. And theater is worse than nothing—it gives you the feeling of control without any of the substance.
The work that remains
When AI handles implementation, the human job shifts from doing the work to understanding the work. Speed without understanding is just technical debt with better commit messages.