One machine is not enough.
The library is searched by rejection: derive pages, measure them, keep the exponentially rare exceptions. Any one machine covers a rounding error of a rounding error — so the map only grows, and the second question only becomes answerable, with volume.
Mining pays for work, not for what you find.
The reward is a proof-of-work check over the page a miner derived. It never inspects what the page contains, so a spectacular find earns exactly what a dull one does — so a faked find earns no more than an honest one. That, plus the server re-deriving every find before it is published, is why the research record is worth anything at all.
Every page gets measured either way.
The official miner scans every page it derives through the same frozen measurements, with no way to turn that off, and records what it covered. The discoveries are a byproduct of work that was happening regardless — the protocol pays nothing extra for a remarkable page, so it offers no reward premium for fabricating one; reputational and off-protocol motives are outside its reach. What actually prevents fabrication is that the verifier re-derives every published find from its coordinates.
Full-speed mining in this tab. Everything mode reads each page through every lens at once and sends whole searched stretches, with every find in them.
Reading the challenge from the chain…
finds from the in-browser seed miner, stored locally on this device. nothing here has been submitted or claimed.
Three different activities live on this page and only one of them is rewarded. Browser mining is real research and earns nothing; the miner on your machine does the proof-of-work; quantum sampling is not yet possible from a browser at all. Each tab says which it is.