while True: learn()

The title of the review says it all: it's a puzzle game with a machine learning theme, not a game about ML. The obvious comparison that comes to my mind is Human Resource Machine, which is about programming in assembly, because the two games have a very similar format. However, while the "skills" learnt in HRM can be transferrable (to a certain extent) to actual assembly programming, I can't really agree with the claim in the promo video that with this game you "learn how ML works in real life", because the level of abstraction between the puzzles shown in the game and an actual ML problem is simply too great. You can learn, however, what different terms used in the field of ML mean and you get a handy collection of links to get more information if needed, so the game has indeed educational value. Misleading promo aside, as a puzzle game I found it very funny and addictive, but I have to knock off a couple stars for some issues (mostly due to the UI) such as: - I understand a limit on the number of blocks, but the area on which the blocks can be placed should be larger to avoid clutter; - The reinforcement learning screen is a mess - thankfully the RL puzzles themselves are easy; - Some blocks are poorly balanced and end up being never used (at least of you aim for a gold rating); - When investing in startups, it is very difficult to understand the link between number of customers, incoming workload and revenue and consequently how to best optimize the algorithm (e.g. you can see that there are more customers than you can serve but if you add extra servers you may end up with a lower profit or even a loss despite managing to serve all customers). All in all, the sum of all parts of the game clicks for me but in view of the issues above your mileage may vary.