Learn·Taosis
What Taosis can and cannot know
The one page to read first: why a blank is never a zero, and why some questions are refused outright.
A blank is not a zero
Everywhere on this site, a value Taosis cannot verify shows as a dash. It never shows 0 in its place.
This matters more than it sounds. A fake zero is indistinguishable from a real measurement, and it reads as information when it is the absence of information.
History cannot be invented
The public chain node prunes old state within minutes. Bittensor keeps no record of past epochs. So anything that needs a series — a validator's track record, a price attribution, a cost basis — only covers what Taosis has watched, and every such figure names its start date.
No amount of engineering changes this. A series that started last week is one week long.
Some questions get refused
A deregistration probability with no observed registrations returns nothing rather than 0%. A reliability verdict under twelve observations returns "insufficient". A holder concentration that does not reconcile against the chain is marked unverified. A price attribution that explains under 90% of a move is flagged partial.
Each of those could have been a confident-looking number. Each would have been wrong.
Where the numbers come from
Everything on the market side is read from the chain — subnet prices, reserves, emission, validator dividends, locks, hyperparameters. There is no third-party market data API in the product.
The exception is labelled where it appears: the TAO/USD rate comes from an external source and is display only, never used in a chain calculation.
Read next
Reading price attribution
Trading versus emission, decomposed exactly rather than fitted — with the leftover always published.
Validator reliability
A track record cannot be backfilled, so this one says "insufficient" until it has genuinely watched long enough.
Cost basis and P&L
WAC or FIFO, replayed over indexed fills — with coverage stated, because a partial history makes a fake profit.