Puzzle: The Computerly confusing solution

It took some time create this puzzle as chess.com analysis just cant comprehend it. 

Also you are facing the black side, and white's pawns are promoting forward.

The analysis board on lichess was also confused (about the eval, not the position [edit because of confusion]) for a while until it realised

AHHH, one of the hardest puzzles I've ever seen and took a few minutes to solve, found some moves for the 1st try but not all.

Good job!

Computer Analysis boards are never confused by weird human trickery - though you can trick the AI-bots in 50 years which have learned "the human way" of things. Today, It is impossible for engines to not know the direction of play as it follows from the registration of the chess piece coordinates in computer memory.

I mean confused about the eval, not about the position

Here is a game I played against stockfish on lichess with this position: https://lichess.org/O7UoFPJY/black#1

So now you are rated 2012? When I played you last week, you were only rated about 750. Yes you mopped the floor with me, wt a 90.5 and two 99.6 game ratings, but I had my opponent game rating set at only plus or minus 200 points. Not at close to 1400 points.

I see! drdos7 has posted a lot of puzzles recently where engines can't get the proper eval or it takes them forever. Yours is quite decent in that respect. StockFish gives sensible move suggestions when running for a few minutes. Just don't expect the initial values to be correct. If you set the Maximum Depth to unlimited it will continue to find better moves and better evals while you sit and do nothing

How 8s this even a puzzle? It's full of inaccuracies and mistakes

I'm afraid that White has a better alternative on his 7th move that stops Black's plans:

Thanks for pointing out the oversight (indeed the computer's eval confused me too!), here is a revised version of the puzzle.

take a look at this variation, I honestly cant find the best moves using the engines anymore, one move b4 it says +10 and then next move is checkmate for black in 8: 8/1Q6/RPPPPpP1/2pPPP2/q1N1K1p1/1P5p/Bp1k4/1n4n1 b - - 0 1

apparently black is winning

Engines are like horses. To win you need to ride them skilfully. What you report is normal behaviour. Engines do not report final evals. They just continue evaluating while you watch the screen. And +10 may change to -M8 even when you play no move at all!

That's much better, I have it as a mate in 16:

But it is not a puzzle because at the end, there were other moves for black to win only, puzzles have to be only moves!

You are too harsh on yourself. Compositions have strict rules even when presented as guidance. When you transgress those, your composition stands no chance of winning a composition award.

Puzzles have no rules whatsoever but most posters believe it is what you can stuff through the throat of chess.com's puzzle interface. Since chess.com has no clue and treats puzzles as some sort of compressed games that is what you get. For instance, the interface substitutes for your missing opponents by orchestrating the counterplay (as entered by the puzzle poster) and plays the solution as one single line. A chess game is always one single line and analysis lines are not part of the game.

Since puzzles are a mess it is advised to follow the composition rules when you create a puzzle (https://www.janko.at/Retros/Glossary/Codex2015.htm). However when posting a composition in chess.com you will always have to compromise. The main action missing is that the puzzle interface does not support solving in the counterplay department. In a composition solving tournament you get pen and paper to write down your solution with the relevant moves by both sides. So actually, no-one is ever solving a composition on chess.com!