The 20-Year Itch: Why My Latest Attempt at App-Building Software Actually Might Work
I've been chasing the same idea since 2000.
Back then, I thought: what if there was software that could build other software? I was convinced it was the future. Each time I hit a wall. The tooling wasn't there. The infrastructure wasn't ready. Or maybe I just wasn't.
Then 2023 happened, and suddenly the constraints that killed previous attempts evaporated.
I'm building it in public now, and it's wild how different this iteration feels. Not because I'm smarter—I'm not—but because the foundation exists. LLMs handle the parts that used to require genius-level engineering. I can focus on UX and the actual problem instead of just trying to brute-force AI.
What strikes me most: I'm not the only person on this journey. I've watched indie hackers shelve projects for 5+ years and come back when timing finally aligned. The idea that "failed" me three times wasn't actually broken—the world just needed to catch up.
Here's my real question for you: What project have you walked away from, only to realize the conditions to actually build it might finally exist? Sometimes the best ideas just need to wait for their moment.
Twenty years of coming back to the same idea is a pretty unusual story. Curious what feels fundamentally different about this attempt compared with the previous ones.
The timing point is real, but I’d separate “the technology became possible” from “the user problem became urgent.” Plenty of ideas become buildable before they become worth adopting. The advantage of coming back after 20 years is that you can now test both cheaply: put one narrow workflow in front of users, watch where they abandon the result, and let that evidence shape the UX before expanding the builder. I’d also keep a record of which assumptions were blocked by old infrastructure versus which were simply unclear product decisions. That history can prevent the new iteration from inheriting the old ambition by default. What is the first user outcome you’re measuring to prove this version is solving a real problem, not just expressing a newly available capability?
I suppose that the very first measure I am looking for is if it lets me build ideas out in a meaningful timeframe at a cost I can afford. I have been capable of building powerful and useful systems for decades, but always at a high cost in both time and money. Early approaches that I tried always failed to deliver what I wanted. Over time, things were getting better, but I still never found a solution that let me do what I wanted in the timeframe and for the cost that made it worth it to me. Now, that is different. AI has enabled me to make serious progress that fits both my schedule and budget. The project I am working on may not be what everyone else is looking for, but I know I will use it. I suspect others may want to as well.