Learning Factory is a factory-automation game where you supply goods to an island full of cats. That premise sounds whimsical, but under the pastel surface sits a genuinely deep production chain puzzle. You mine resources, process them through increasingly complex machines, route products via ground conveyors, underground tunnels, and aerial transport lines, and unlock technology through a branching research tree that spans dozens of nodes. The three-layer logistics system is the real hook: figuring out when to push a product underground versus overhead is the kind of spatial decision-making that sends you back to the drawing board more than once. The machine-learning angle is not a marketing gimmick. The game embeds a simplified ML training loop where you feed data about cat preferences into in-game neural-network nodes to predict demand and automate production decisions. It is approachable enough that you do not need a data-science background, but it has real depth if you want to optimize output ratios. Think of it as a second strategy layer sitting on top of the standard factory belt-and-ratio game. New players can ignore the finer tuning for several hours before the system demands attention, which is exactly the kind of difficulty ramp that respects your learning curve without patronizing you. Procedural world generation gives each run a meaningfully different layout, which matters for replayability. The research tree is wide rather than linear, so two playthroughs can develop very different factory shapes depending on which transport and production techs you prioritize early. The absence of combat or time pressure means you can pause, zoom out, and audit your supply chains without a wave of enemies burning your belts. For players who find factory games stressful because of attack mechanics, this is a genuine selling point rather than a missing feature. The 86-percent positive rating across 513 reviews on Steam suggests the core loop is landing for most buyers. Weaker spots exist. The tutorial covers fundamentals competently but does not go deep enough on the ML mechanics, so expect to spend time in community guides or the in-game glossary when the neural-net nodes start multiplying. Mid-game pacing can feel slow if your initial map layout forces long transport runs before you have the tier-2 logistics unlocks. The mod ecosystem is modest right now given the early 2025 release, though the developer has a history of active post-launch support. AI systems govern cat demand curves rather than opponent factions, so if you need a strategic rival to race against, this will feel too relaxed. For the strategy-and-sim crowd specifically: treat this the way you would treat a Zachtronics puzzle set inside a factory game. The satisfaction is in closing the loop on a production line and watching throughput numbers climb. Build out your underground routing first to keep the surface grid readable, invest in the demand-prediction research branch early, and do not underestimate how much the aerial layer changes your layout options in the late game. This is a well-constructed, low-stress automation game that earns its depth through layered systems rather than complexity for its own sake. Diego, Scout Team