Estimating historical DOSC value

Or why 1 MEL is likely to hold a stable value

Eric Tung
Mel Blog
2 min readJan 28, 2022

--

One of the most exciting features of Themelio is its base currency — mel (or as a ticker symbol, MEL). Mel is the world’s first trustless, non-fiat stablecoin: a low-volatility cryptocurrency that has completely autonomous, trustless issuance like Bitcoin, without any sort of oracles or pegs to fiat currencies like the US dollar.

What guarantees the stability of mel is Melmint, a mechanism that measures and pegs 1 MEL to a unit called a DOSC. The DOSC is defined as the value of the amount of sequential computation that the fastest processor at that time could do. For example, a DOSC in the year 2000 would be the value produced by running the fastest processor available in 2000 for 24 hours. Similarly, a DOSC in the year 2022 would be the value of running a top-of-the-line 2022 processor for 24 hours.

But the central assumption is that 1 DOSC is, in fact, a relatively stable measure of value. Anecdotally, that seems to be the case — renting a 1-CPU server has always cost about the same ever since I could remember — but hard evidence wasn’t the easiest to come by. After all, people don’t really track the value of a day of computation, since the value of a certain amount of computation is much more interesting in pretty much any context other than Melmint.

So we did the dirty work of scraping old processor prices, benchmark results, and electricity prices to roughly estimate how much a DOSC was worth for the past 10 years or so.

(Note the small range of the Y axis)

And it turns out that the DOSC does keep a relatively stable value — about $0.3. Note that this is just a rough estimate that makes lots of simplifying assumptions about facts like how fast CPUs depreciate. It also doesn’t take into account anything other than CPU and electricity prices, while in the real world, a mel-minting operation would have all sorts of other costs (like renting space to put the servers in, etc).

All the detail of our methodology can be found in this GitHub repository!

--

--