disagree with releasing a project every week. what makes a project worth using is continued maintenance, thorough tests, comprehensiveness, and battle-tested API; not just a few snippets of code you uploaded somewhere on a sunday.
IMO, the size of the project being small is the key.
Building bigger stuff may sound nice, but not everyone can spend time on something for too long. It's not time as in hours or days. But time as in motivation. For me long development times, without releasing prototypes would be de-motivating. I love to show it off to people and hear them call it "cool" or atleast "good idea, but not usable". So then when you have a prototype in a week, you can make your next project as working towards v1.0 of the previous week's project.
Disclaimer: The site is down, so I haven't read the post. I read the title and the comments were interesting, which pushed me to post my experience.
All the opensource stuff listed here http://akash.im/code since Jan-2011, were according to the 1-per-week rule and none of them commercial. The arduino ruby gem was a little bit popular. When there's only 1 project in the month, you can assume that I've been working as a remote intern or have been doing my last freelance work :)
You can see that I clearly did not keep up with the 1-per-week expected results. Like I said, it was 2 per month in the end, but I still practice it. Along with it, since last month, I also make sure I read one-interesting-paper-a-month (was the Amazon Dynamo paper last month).
Yet to work on something this month. This month, it'll mostly be a scheme interpreter (R7RS-small spec draft#1) and a light-weight key-value store for fun and then (possibly) get a job as a remote full-timer at some startup (remote jobs are tough to get I guess).
Besides, what opensource stuff do you write? (just interested in stuff and trying to make friends to collaborate).
For example, I'm working on recoding my mogade project. It'll take a couple months for the hour or two I spend on it a day. However, this week I managed to "release" the leaderboard component of it.