AI writes 180% more code. So why aren't we shipping 180% more software?
A recent MIT study, covered by Forbes, found that AI coding agents increased the amount of code developers write by around 180%, yet the amount of software actually shipped increased by only about 30%.
That statistic made us rethink where the real bottleneck is.
As founders and builders, we spend a lot of time choosing models, learning prompt engineering, and trying to generate code faster. But building a successful product has never been just about writing code. It starts with understanding the problem, defining clear requirements, challenging assumptions, aligning everyone around the same vision, and making the right product decisions before development begins.
That's exactly the mindset we've been building MySpec around from day one. AI can dramatically accelerate execution, but it still depends on the quality of the thinking that comes before it.
Curious to hear what other founders and builders think. If code is no longer the biggest bottleneck, what is?