In August 2024, Buildspace shut down.
Buildspace wasn't a course. It wasn't an accelerator. It was a structure — a deadline, a cohort, someone watching. Farza called it "a place to build things." What he actually built was external accountability at scale.
When it closed, tens of thousands of developers lost the one thing that had been making them ship.
Then, six months later, every major AI coding tool got dramatically better.
And somehow, developers are still not finishing their side projects.
The explanation nobody wants to hear
In July 2025, METR published a study that should have been a bigger deal.
They measured the productivity of experienced, professional developers on real software tasks — with and without AI tools. The result: developers using AI tools were 19% slower on average. Not junior developers. Not people learning to code. Experienced engineers, on real work.
The immediate reaction was what you'd expect. Denial. "That's not my experience." "The study is flawed." "Wait for better models."
But a follow-up study in February 2026 didn't reverse the finding. It narrowed it — the effect is smaller than originally measured, but still directionally negative for complex tasks.
The uncomfortable explanation isn't that AI tools are bad.
It's that AI tools solved the wrong problem.
The planning problem vs. the execution problem
Every stuck developer I've talked to has the same story. They know what to build. They have a tech stack. They've probably sketched the architecture in a notebook, a Notion doc, or a Claude conversation.
The problem isn't planning. It was never planning.
The problem is the Tuesday at 7pm when you have 45 minutes, you open your IDE, and somehow you end up watching something else entirely. Or you do open the project — and you spend the time refactoring a file you've already refactored twice, because starting the thing you actually need to build is harder than it looks.
AI tools made the planning step faster, cheaper, and more detailed than ever. You can get a full architecture in 10 minutes. A database schema. A component tree. A deployment strategy.
None of that generates your discipline for the next 30 days.
Buildspace understood this. The product wasn't the curriculum. It was the cohort, the checkpoint, and the fact that someone would notice if you disappeared on Day 4.
When Buildspace closed, nobody replaced what it actually was.
What happens without external structure
The research on this goes deeper than the METR study.
Kahneman's Planning Fallacy (1979) shows that humans systematically underestimate how long tasks take — and overestimate their future motivation. This isn't a character flaw. It's how cognition works. You plan from a best-case scenario, then execute in reality.
Solo side projects are a perfect storm for this. No deadline anyone else cares about. No one watching. No consequence for slipping the timeline. The only accountability is self-generated — and self-generated accountability is the weakest kind.
AI tools made this worse, not better. The gap between "I have a plan" and "I shipped" is not a knowledge problem. AI closed the knowledge gap. The execution gap got wider.
What I'm testing
I'm running an experiment on this. It's called MVP Builder.
The idea is simple: a structured 30-day sprint for developers with a full-time job. You apply with your project. You get daily prompts tailored to your stack and where you are in the sprint. And there are milestone checkpoints where someone reviews your progress before you move forward.
Not an AI reviewing it. Me. Because right now, at Cohort #1, the human in the loop is the founder.
That doesn't scale. That's exactly why Cohort #1 is free.
I'm not selling a solution. I'm testing a hypothesis: that what developers with side projects actually need isn't a better plan — it's a system that holds them to the one they already have.
If that resonates: mvpbuilder.io/pipeline. Applications are open. 8 spots. No credit card.
If it doesn't: I'd still genuinely like to know what has worked for you. The comments are the interesting part.
Top comments (0)