Automated energy management - Two people sit at a desk with a laptop, model wind turbine, hard hat, and solar panel, holding a clipboard displaying an energy efficiency chart with coloured ratings from A to G.

Playing Games With AI

Playing Games With AI

Jon Slinn – June 2026

I was on holiday recently and that for my family means games… lots of them. I mean of course sun, swimming, eating too much etc but lots of games as well.

One of the favourites is Castle Combo. Full credit to the designers, Grégory Grard, Mathieu Roussel and their publisher Catch Up Games. It really is great so if you are looking for a present for the game fan in your clan I recommend it.

In short what you are aiming to do is maximise a score where scoring mechanisms can either re-enforce each other or conflict. There are trade-offs and choices to make as you select and lay cards in a 9 by 9 grid (or tableau as the game calls it).

The final score is determined by the rules denoted by the section at the foot of each card with the red numbers.

So in the tableau in question here there is the following image on card two:

Automated energy management - A scroll shows 2 per 4 with the number 2 inside a red wax seal, the word per in black text, and a partially melted gold coin next to the number 4 on a brown background.

This means 2 points for every coin in the purse.

There are 4 coins in the purse (you will see them stacked on the game image below) which is the maximum for this card. So that’s 8 points total at the end of the game.

         * note the ‘4’ on the bag above means that’s the maximum capacity of the bag

 

Automated energy management - Nine illustrated game cards are arranged in a 3x3 grid, each depicting an animal character, icons, and numbers. Two round wooden tokens, marked 1, are on the upper middle card. The cards have colourful backgrounds.

The cards are scored left to right in rows so:

Top row left, 5 x 3 = 15;  top centre, 2 x 4 = 8; top right, 2 x 3 = 6

Middle row left, 8; Centre, 3 x 2 = 6; Middle right, 3

Bottom left,10; Bottom centre, 3 x 2 = 6; Bottom right 2 x 3 = 6

In total that’s 68

Easy enough to do, a 6 year old can have a good go at it. Ok, you can get it wrong as well but only because of silly slip ups. It’s not a hard task.

 

As we here at Optimal Monitoring have been fully AI pilled I decided to be lazy and get Gemini on my Google Pixel to add it up for me. It got it totally wrong. I mean even on the Stuard and Master at Arms where I directed on the coin count (as it was too hard to see on the image) it got it wrong…

Ok I thought. This has to be tested now. So now back at my desk here are the results:

Automated energy management - A table compares AI models on a task, listing the engine, score calculated, and notes. Scores range from 0 to 68, with various issues or successes detailed for each model, such as calculation errors or narrative output.

Now to be fair to Gemini & Grok I don’t have access to their premium models. I am sure they would make a better job of it if I did.

That said GPT proved this is very possible and that even some advanced models such as Opus 4.8 struggled to deliver.

Very unscientific I am sure you will agree but a great example of how quickly you find the edges of these general purpose LLMs. They may be able to write all this in seconds and spell it all correctly, but we poor slow meat computers might still be half a stride ahead in some areas…. for now :o)

 

PS: If you do have access to Grok & Gemini premium please give this a go and let me know the results!

 

Appendix

Full prompt:

The above image is from a game of Castle Combo. The full rules for the game are available here :  https://board-game-rules.com/wp-content/uploads/2025/01/castle-combo_Official-Rules.pdf and on other sites.

The Master at Arms is in position 2 and has 4 coins in her purse

The Stuard is at position 3 and has 3 coins in her purse.

Please can you add up the end of game score for this tableau of cards.