This is how DeepMind will train its pacman AI. Brilliant. :)

Actually what datasets will you guys be working with? Anything fun?

Propose your favourite datasets!

Guys this is amazing.

Guys, really great stuff! Only missing piece for me was some comparison on the scaling and performance differences between CDB and GFT. I saw a PostgreSQL slide at FOSS4G-NA showing 350k QPS with 100 concurrent requests in v9.2. Maybe a totally different post though. :)

Minor nit, but what stuck me most about the apology was that Allan didn't include a link to http://37signals.com.