DesksNear.Me

DesksNear.Me was dreamed up in a second by me, refined by Keith Pitt, Warren Seen, Alex Eckerman, and myself over 4 weeks, and then built and delivered in less than 48 hours by all of us during the RailsRumble competition.

I wanted to briefly go over a few of the things that were good and bad about our adventure thus far.

What went right:

• Having a fantastic team is key. I was fortunate to work with some very dedicated, talented, and passionate people.
• We got a reasonable amount of sleep. I did RailsRumble in 2008 and stayed awake for 57 hours straight. More time to code does not mean you get more code done. After about 18 hours of coding straight, you stop thinking clearly, make mistakes, and ultimately freeze up. Our project deserved more than this, so we got our sleep and focussed in our respective days.
• For the periods where our differing timezones allowed, we kept a 4-way Skype chat open the entire time. This kept us efficient and focussed because we could hear each other working, didn’t get bored/distracted/lonely, and if we got stuck, help was only a question away.
• We worked in two 24-hours iterations and aimed for (and almost achieved) our minimum viable product in the first iteration.
• We fleshed out our backlog and planned our weekend strategically.
• The first thing we did was deploy and we deployed very regularly so we wouldn’t be scrambling at the end.
• We covered all functionality in integration tests. We’ve had no recorded 500s and less than 1% of requests are 404s.
• We set up a CI server (in that same 48-hour period) so we could move agilely and have shared knowledge and responsibility for keeping our test suite thorough and green.

What went wrong:

• We did not anticipate how quickly we would get attention. Because of this, our naïve implementation of our geocoding (all done server side) hit the Google Maps API limit in about two hours. We provisioned a new IP, and hit the limit again an hour later. Fortunately, @ThorMitchell got in contact with us and white-listed our IPs. A big thanks to the Google Team for helping us out here.
• We had to rewrite our search functionality three times because our assumptions about how intuitive the search would be were wrong. Luckily, thanks to our CI server and quick thinking, this didn’t waste more than a few hours.
• We lost about 12 hours across the entire team trying to test OAuth logins. We ended up switching plugins and reworking our test suite too many times than we can count and I think this really hurt us. We didn’t get to about 3-4 features we were hoping to get into our final product. In the end, I’m glad we persevered, after seeing how many final RailsRumble entries simply 500ed when I tried to log in to them.

What went REALLY right:

• @collis from Envato got wind of our application in the first hours of it’s life and tweeted it. We are pretty sure this is the catalyst for a whole series of successes, ultimately leading to (in these first 4 days) 11,981 visits, 41,309 page views, a HackerNews front page article, 354 users, 119 real self-listed workplaces, and 63 real bookings. We are incredibly excited with these results. We’ve received consistently positive feedback and several reports of freelancers who made bookings this week turning up at locations such as The Frontier Group in Perth and EngineYard in San Francisco.
• @ThorMitchell tweeting to me and coming to our rescue about 5 minutes before we were going to pull out of the competition in order to fix our geocoding issues. I’m not kidding, we were on the phone to one of the competition organisers as it happened.

I’ve gone on long enough now. Thank you all so much for helping us come this far, and please continue to do so. I think with your help we can reach the critical mass needed to make this site a thriving, self-sustaining, and wonderfully useful community of coworkers!