So here's an actual image that shows my data on this:
For this variation, I have examined 14272 positions (all entered manually, no automation). For the positions after 10...Qxc5 I have only examined 159 positions.
So, if zullil can surpass my analysis by just examining
TWO positions, then I'm a fool and I've wasted my life, because I could just have let the engine running in the background analyzing by itself while I was doing other things, checked what it spit, and copied and pasted it like a parrot.
Otherwise, unassisted Stockfish builds a pitiful tree of variations, and to refute its depth 70 all you need to do is go to the tail of its variation and show it much superior analysis. It loses 1 ply per half-move, so whatever Depth 70 shows at move 11,
you can produce a better line by jumping to move 31 and analyzing at depth 31.
This is no secret, and I'm not doing anything outstanding, whoever is reading this, you can too refute unassisted Stockfish, and you can probably do it better than me, if you try.
At least, I'll claim I put more "time and energy" in these 14272 positions than zullil put in letting unassisted Stockfish reach high depth in 2 positions