An index of commodities would almost certainly have more stability than any single commodity and it's not really all that much more complex to peg to an index.
We could easily come up with 10-100 such commodities and peg to that. The only realistic difficulty would be people bikeshedding on what commodities should be included.
https://peakd.com/hive-111111/@demotruk/4-steps-to-repeg-hbd-as-a-commodity-currency
Finding a set of commodities could possibly be done by lineair fitting against historical data, pre-setting a weight range for individual commodities.
You can look to minimize the sum of square error for the recorded inflation over the last 50 years vs the yearly price development of 10 commodities over the same 50 years each with a weight factor between maybe 5 and 15 that needs to add up to a hundred.
It will take some brute forcing, but the computer could likely pick the historically best ten commodities plus weight factors.
Fascinating stuff. Get a neural net going afterwards with the reward function set to minimise that sum of square error you mention. Surely there's enough econometric data out there to train something like that.
Then you could sidestep the whole obsession with trivialities that @demotruk is talking about.