When the site is running smoothly (we had a rough week) there are a couple
of basic functions.

Users who log in get a history of their recent semantic tags. We have had varying luck with automated tagging, we use Zemanta and Orchestr8's AlchemyAPI and have a long way to go in improving accuracy. Then a user can view this tag cloud and fire off a real time search by clicking to see who else is talking about the same concepts/ideas. In addition, and this could be a very strong value point, users can view their lists as tag clouds to quickly scan for topics of interest. One of the issues we have now is using category or high level tags versus specific tags. Both could be useful (one for more specific search, one for better clustering of like messages).

On the ad side anyone can install a plugin that users can opt into with their Twitter account using oauth. The ads are them dynamically matched to keywords or tags associate with the history of that account.

When the site is running smoothly (we had a rough week) there are a couple
of basic functions.

Users who log in get a history of their recent semantic tags. We have had varying luck with automated tagging, we use Zemanta and Orchestr8's AlchemyAPI and have a long way to go in improving accuracy. Then a user can view this tag cloud and fire off a real time search by clicking to see who else is talking about the same concepts/ideas. In addition, and this could be a very strong value point, users can view their lists as tag clouds to quickly scan for topics of interest. One of the issues we have now is using category or high level tags versus specific tags. Both could be useful (one for more specific search, one for better clustering of like messages).

On the ad side anyone can install a plugin that users can opt into with their Twitter account using oauth. The ads are them dynamically matched to keywords or tags associate with the history of that account.

Basic functional elements transfer over, but where the user was me and 3-4 other engineers before, now the potential user base has become unlimited. From the nuances of web software (so many languages, tools, and such a different environment), to the practicalities of implementing concepts in a scalable way (data, memory, where apps run either client or server), I find that I much to learn each day I work in this space.

Now to find where our project is kept on the server (my first time getting down and dirty since it was hanging up, and the tech lead is sleeping).

Your overall point is correct – usage and actual usability should drive future design and development decisions. But at the starting point, the rationale of an initial design should be based on the experience of a designer or developer. An experienced interface or usability designer doesn't operate on hunches. Instincts are formed through study of best practices and user research. With experience your starting point is not a shot in the dark or a hunch.

You can make up for inexperience by taking advantage of beta testing BEFORE you launch. All teams should test, no matter what, and pay attention to the user's experience. But don't just do it after you launch. Testing is a constant. The work is never done – especially on the web.