3. Thinking Big -- The Flipside. Like we mentioned above, it was critical for us to think big and to try to advance every aspect of the game in order to stay competitive and come out with a AAA game in 2007.

This was the only viable strategy -- but it had a flipside. We had minimal time to spend on preproduction and feature testing in the early stages. We had to be working on assets at the same time as the technology to support those assets came online. Obviously, we were stumbling upon unforeseen issues all the time.

One of the biggest technical challenges we had was with our ShaderCache system (see Wrong #4) -- but that's not the only example. ShadowMap technology only became functional at the time we were at Alpha, so Olga Cheremisova and our other lighting artists had a very limited time to go through all the levels and set up proper lighting.

Sometimes we did not have the time to properly upgrade the tools to accommodate the volume of new assets. Our lightmapping tool was developed two years ago, but it was never tuned to handle the very large scenes that we were coming up with.

In the worst cases, it was taking up to 45 minutes to calculate lighting for some of the larger scenes, which is obviously totally unacceptable and made the life of our lighting artists even tougher. The only viable option we had was to continually upgrade hardware and give multiple computers to the key lighting artists.

Another issue we faced was production speed due to constantly evolving technology. Although our engine was highly optimized in time for ship, we were not able optimize the engine fully during development, which affected the work of artists and scripters who unfortunately had to struggle with the game running at sub-optimal framerates.

It was only at the end of production that we had time to get the engine running fast. Had we been able to do this earlier, the design department would have had an easier and more efficient time scripting our levels.

Finally, because the final dev schedule was so compressed, most assets were coming online late -- even though everybody involved was pushing as hard as possible, the sheer volume of new assets was mind-boggling. This included delivery of mocap, VOs, localization data, FMVs, and other components.

It was our first attempt to use mocap, and probably because of that, it took us over a month to get the first mocapped animation in the game and nail down the pipeline (of course, when we finally got that first animation in we lost Ruslan Vizgalin, our lead animation programmer, for three weeks -- he was in the hospital fighting pneumonia).

This whole process was wreaking havoc on our lead AI programmer Sergey Mironov, the scripting team and the producers -- most animations and animation transitions were not working for about a month, the game was not playable, and it was impossible to easily tell what the source of a bug was -- was it one of the many animation glitches, or a scripting issue, or an AI navigation problem?

It was not much easier with VOs and localization assets. VOs were recorded and localized at the same time, even before integration of English-language VOs in the game was complete. Anna Naleushkina, Saber's AP responsible for those production tracks, was frantically coordinating from our St. Petersburg office the movement of those tens of thousands of assets between the recording studio in LA, the processing house in Vancouver, and localization team in Dublin. Surprisingly, nothing was "lost in translation".

To be honest, we don't know if there's a good solution to handle this -- other than add more time and make the project more sane. That clearly wasn't an option for our team, though. When you have such a massive number of assets coming in and you need to do the integration very quickly, it simply creates a project-wide havoc, and it feels as if the skies are falling.

There was never enough time to fully update the tools or pipelines, so all we could do was to clench our teeth and plow forward. A good enough number of Saturday night beers helped us succeed.