Flores book criticizes a move and suggests a much better one. Engines laugh out loud.

So I was researching pawn structures so I can improve the detection feature in LiChess Tools when I stumbled upon this position: [FEN "r1bq1rk1/1p2bppp/p1n1pn2/8/3P4/P1N2N2/BP3PPP/R1BQR1K1 b - - 1 12"] where Flores says:
> A very serious mistake
then proceeds on lamenting that most players in that position play 12...b5 and recommends Qc5. And yet, with both SF16 and Leela the best move is b5. Worse, Qc5 isn't even in the first 5 options. The only engine that mentions it is Maia 1900, which tries to play as an 1900 ELO human, and it's the third move, with a difference of evaluation of +1. SF16 rates b5 as +0.3 and Qc7 as +0.9!

The book is Chess Structures: A Grandmaster Guide Paperback, by Mauricio Flores Rios (Author), and it was released in 2015.
Also, it's not a matter of an old game that was annotated before the advent of computers, because the game is chessgames.com/perl/chessgame?gid=1720239 from the 2013 World Rapid Championship. So not only did he annotate it recently, I am sure it was a game annotated by multiple people.

I am trying to understand if this is one of those "GM knows better than engines because he understands plans and above ply 40 that's actually a good move" or something like talking out of one's ass. I remind you that this is THE BOOK that anyone references when discussing pawn structures, besides the much older Solis 1976. How can I trust anything in this book now?

Alternately, how can I trust engines ever again?

no, Engines knows ALWAYS Better than any human, especiall for example Stockfish 11 , he sees more than any possible GM knows. there is no need to ask that, if You don't believe, just check Engine Branches and GM Branches. I saw how Engine are Playing, Beauty of Their Logic starts to Feel and Seen after they work some with Positions, then You may mark , how deep they are calculating, Compared to that ,even Human Beauty chess logic, actually can't compare.
None of GM or whatever can even compete with Stockfish 11 or 9 . I saw only One person defeating here Stockfish 8 on lichess, but He made it after Big Preparations, besides engine here is very cut, I don't think He could win 9 or higher, or 8 with more capacity.
it's not question of trust actually, need to apply simple logic, that if something posseses ability to see much more variant than Humans, then It's logic is Better in all cases+ Humans are unstable quite often, but I believe someday Miracles can Happen( I want to defeat some Stockfish xD)

In the original post I said Qc5, but I meant Qc7. Sorry about that.

@TotalNoob69 said in #1:
> How can I trust anything in this book now?

Remember: strategy is different than tactics (at least, it is for us humans). Although a GM may make occasional tactical oversights (in specific positions), what they say about strategy will generally still remain valid.

This coincides with some of my thoughts, @MrPushwood . Plans come top-down, or from the end of a possible game to the current position. Tactics go from the current position towards an eventual end. Tough to analyse moves in these conditions. A move might be very good in furthering one plan, and bad for another. Can one say "this is a great move for the Isolani plan, but a pretty poor move in general?" It is a little bit confusing.

For humans, there isnt a difference between 0.3 and 0.9.

@TotalNoob69 said in #1:
> I remind you that this is THE BOOK that anyone references when discussing pawn structures

Engines are superior no doubt and call calculate variations upto high depth.
Back in those days there was no existence of chess engine. So, there were strong players who backed on moves that have immediate effect cause they weren't computers who can calculate the game far ahead. Hence, you can't neglect the great authors of the game. It's just that when you do comparison, then you act as stupid. Use any but the one that helps you understand better.