Basically there were three iterations of the technology. The first one used microfluidics, which are basically the repurposing of the micro-fabrication techniques that the computer chip industry pioneered to move tiny volumes of liquid. Theranos tried to work on that for several years before Holmes lost patience in late fall 2007 and abandoned it. At that point she pivoted to what was essentially a converted glue-dispensing robot. One of the engineers ordered a glue-dispensing robot from a company called Fisnar in New Jersey, studied its components and rebuilt a smaller version of it. It was a robotic arm on a gantry with three degrees of motion and a pipette stuck to its end, and it replicated what the scientist does at the bench. That ended up being called the Edison and it could do only one class of test, known as immunoassays. It was also error-prone. The third iteration of the technology was the miniLab, which Holmes wanted to be able to do more than just immunoassays. The miniLab didn’t invent any new techniques to test blood with. It miniaturized existing laboratory instruments and packed them into one box. By the time Theranos went live with its tests, the miniLab was a malfunctioning prototype.
The interview also explains how Holmes was able to attract so many influential people on her board. The short answer is her Jobs-like "reality distortion field" -- she was exceedingly good at convincing non-scientists that their science worked.
[1] http://m.nautil.us/issue/60/searches/does-theranos-mark-the-...