but there's no way to like 'game' the game

There are ALWAYS ways to 'game' the game.

The article is clearly a reporter's awful rewrite of a marketer's interpretation of a product developed by people who actually understand the data.

I notice the quote in the article about "no way to game the game" didn't come from anyone inside the company, but came from a college student who found a job using their tool. That's a rather stupid way to get a quote about how the innards work. While you might ask a PowerPoint user about what the user experience was like, it makes much less sense to ask the same user about the nuance of how PowerPoint works internally. Similarly, the reporter talking to a past customer could ask what their experience was like, but questioning them on the data science behind the tool is bad reporting.

Looking over their team information they've got some people with names. No idea who they are but the names and details say Ms Yoo (PhD, numeric analysis of brain functions), Mr Cohen (PhD, computational analysis of emotional regulation in humans and animals), Mr Salvatore (MS, MIT, cognitive science), and a few more. So while the marketers may not know anything about statistics and data analysis, the geeks behind it probably have some idea what they are doing.

The theory behind seems fairly reasonable:

Study how lots of people play games involving logic reasoning, spatial reasoning, social reasoning. Use those results of known people in known jobs to find where a new user fits, and recommend jobs that are similar to those they match. Someone who suck as logic problems but excels in social reasoning may have lots of matches with people in marketing or sales. Someone who excels at logic problems but sucks at social skills may be recommended toward technical jobs.

As far as data goes, that seems fair enough. If they are careful about building "games" that actually equate to those areas of reasoning (which is hard to do but possible for low-exposure to the tasks), then they have enough data points to see how lots of people who succeed in lots of fields score (which takes a lot of data), then they could certainly spend a few minutes to find out where an individual's score tends to cluster, and returning the career types of others who clustered in that area and succeeded in their careers.

I actually think this could be a valid way to weed out good and bad candidates depending on the job they were recruiting for.

It sure beats the hell out of a startup I worked at a few years ago which used a candidates Klout Score https://klout.com/home to judge weather or not people were good candidates for programming and design roles.

I think this is an ok idea paired with sensationalist writing. Klout though... Oh man, that sounds horrifying/very easy to manipulate.

Also as someone with no social media profiles (Facebook/G+/Twitter/0 google results/instagram/linkein/whateverelseyoudamnkidsusethesedays), I wonder what my score would be. Would I be 0% trustworthy?