It's an interesting speculation, but it might not be correct. In particular, the subhead about how it might have been highly efficient is crucial to its potential usefulness, but the highest efficiency number claimed is 5%, less than a quarter of current mainstream PV, and an eighth the efficiency of current lab PV. And the "thermo" part seems like it was a red herring.

How efficient is such a device? If Cove's device was 20% efficient it was an unheralded breakthrough, though a toxic one, unlike modern silicon solar panels (contrary to a false statement in the article), and made with relatively expensive materials (antimony isn't just poisonous, it also costs US$11/kg). If it was 2% efficient, it might have been viable in a few niche uses. If it was 0.2% or 0.02% efficient, it was a laboratory curiosity. The article says Cove's prototypes were 2.75% efficient and 5% efficient, producing 45 watts from 1.5m² and 60 watts from 1.125m² respectively, which I calculate as 3% and 5.3% efficient, respectively. But it also says the guy went to jail for a year for defrauding his investors.

But apparently nobody has been able to reproduce Cove's results in the last century.

If you do want to, as the article says, "build low-tech photovoltaic devices, which convert sunlight into electricity", you can get about 1% efficiency out of red copper oxide, another toxic semiconductor like the zinc antimonide intermetallic erroneously described as an alloy in the article; this is a common science fair project for kids, and Wilhelm Hallwachs demonstrated solar panels using this chemistry in 01904, the same year as Cove's first experiments. Toshiba managed to get thin-film red copper oxide PV panels past 8% efficiency last year.