I've orchestrated a tool that squeezes every available token of #compute out of my account, and as a consequence I have a pretty good insight on compute burn (see the cached #token field).
And that's before installing a local token saver tool that can save up to 95% of burn by local indexation/cache.
