There is no Spoon - Episode 2 puzzle discussion


Feel free to send your feedback or ask some help here!


Really liked the problem. I can’t wait for some editorial!


The puzzle can be found here:


Two questions:

  1. How can someone’s code? As far as I remember, it should be opensourced.
  2. Can I change my solution and test it just for fun after contest?

  1. You can’t. The open-source statement was an old feature not available in the new site due to the achievement system. We are working on it.

  2. You can. -> "View Repor"t to get access to your code.Then, try the contest in the puzzle section Medium / Hard


You can change the language to english in the footer, much more understandable for me than german


Indeed :thought_balloon: :smile:


Seriously? 1h30? Looks more like the time to adapt an already working algo than coding it without any prior knowledge of it.


From what I experienced yesterday, it was totally possible in 1h30.

First puzzle was a matter of minutes, and second one wasn’t hard if you divided the problem correctly.
I had a solution which didn’t check graph connectivity and crossings after 1h10, graph connectivity was a matter of minutes, crossings too but I lost at least 30 minutes because of an inverted boolean. Then I had to add a strategy for last test because PHP weren’t fast enough while the exact same algorithm in C++ didn’t need it. So without my dumb mistake and the last test timeout, I would have finished in about 1h40 instead of 2h20 I think.

However, it depends on what you’re meaning by “without any prior knowledge of it”. Because though I didn’t know Hashi puzzles, I knew other similar ones and that helped me a lot to find an easy strategy. Also, graph connectivity isn’t easy if you’re not comfortable with graphs, otherwise you can do it without thinking.



My solution currently passes all tests except the last three.
The last one because of crossing links, but on CG and Multiple solutions 2 it says:
“Timeout: The APU needs more than X connections to function.”
X being 30 on CG and 20 on Multiple solutions 2.

The problem is I can see my algorithm setting all links correctly, but the platform tells me it fails even though I see it works in the visual interface.
Can someone explain to me what I’ve done wrong? :smile:

Thanks, Julien


Can you post a screenshot of the visual rendering of your links to help us to understand better what it might be?




“The links must connect all the nodes into a single connected group”, You have more than one connected group of the nodes.


You are missing one of the rules:

Here, you created several groups that are not interconnected.


Oh, I am very dumb then. Sorry for the inconvenience!


@SaiksyApo right now for me at least the viewer is bugged so it only display the first test and not correctly.


We already improve the player for the next release.


I have tried to recode it tonight. I came up with a simple bruteforce method that could pass 9/13 tests in one hour (better than my contest score which only did six), and later added some simplifications to help it get its way up to 11/13 tests.

and that’s all I can get! (my bruteforce just timeout on CG and expert) Looks like I can’t “divide” the problem correctly yet! I guess it is as hard as it looks if you don’t know how to code the required algorithms.


Hello everybody. I just finished this test, three weeks after the contest. I had to write some functions of tree building to pass the crossing (and for others exercices on cg), because while the contest, i was using arrays and did only 54%.
So i got it without tree path, and only logical testing : isolated nodes, link matching, …
On some complicated tests, logic isn’t enough, and some luck is needed, so i watch only the next situation to decide… and luck too ;)… and all tests passed but the CG test.
For this one, i use a trick because it is a sort of drawing and not a logical test as the others, so i have just to test if the map is sort of drawing or not… and it does the work.



what strategy would you apply for the following map:


I’m about to fork the context recursively multiple times and see what path will finish. Does someone has a better idea?