I don’t usually agree with Paul Graham’s rants, but in his post “How to Get Startup Ideas” he hit upon something, that is incredibly obvious, yet I see people missing it all the time:

The way to get startup ideas is not to try to think of startup ideas. It’s to look for problems, preferably problems you have yourself.

The very best startup ideas tend to have three things in common: they’re something the founders themselves want, that they themselves can build, and that few others realize are worth doing. Microsoft, Apple, Yahoo, Google, and Facebook all began this way.

Add 37signals, GitHub and virtually all other really successful startups in “my time” to that list. GitHub, for me at least, probably embodies Graham’s point — what 37signals has popularised as “scratching your own itch” — better than anyone else. Chris Wanstrath tells the story at Startup Bootcamp 2010 about how he and Tom Preston-Werner built GitHub on the side as nothing but a tool to help themselves in building their other projects. Wanstrath’s side prohect was one of those archetypical “startup ideas,” called nothing less than “FamSpam,” which would be a kind of social network for spread out families.

Interestingly, what wound up happening was, that people started e-mailing the GitHub co-founders asking how they could pay. Because the site was still in a rough beta version, ie. it wasn’t being given a lot of attention, no payment options were available, but people wanted to make sure that it kept going, so they actively asked for it. FamSpam, on the other hand, while seemingly a brilliant “startup idea,” got virtually no subscribers despite having a fully functional subscription system set up. Today, GitHub is valued at at least $ 750 million.