while True: learn()

What the game is

while True: learn() is a puzzle game from the studio Luden.io about building machine-learning systems out of boxes and wires. The framing is a joke that holds up: you are a programmer who discovers your cat is better at coding than you are, and you set out to build a system that translates cat into human speech. Along the way you take freelance contracts, and later you can run a startup as its technical lead.

Each level is a diagram. Data enters on one side, has to leave the other side sorted, labeled or predicted correctly, and in between you place processing blocks and connect them with lines. Contracts are modeled on recognizable machine-learning problems such as image recognition and self-driving cars. The game is sold on almost every platform and the terms differ: Steam and Xbox sell it outright, while the iPhone and iPad version is a paid app at $4.99 with no in-app purchases and no advertising, rated 9+ by Apple. Shopping inside the game uses in-game money only.

What you actually do

A client states a goal: sort these pictures into two piles, or predict this value from that data. You drag blocks onto the board, wire them together and press run, and a simulation streams data through your pipeline and reports the accuracy it reached and how long it took. If that is not good enough, you rearrange. Money from finished contracts buys better hardware and better algorithms, which unlock harder contracts. Pressure comes from three constraints at once, accuracy and time and budget, and later levels only pass if you satisfy all three. Getting better looks like recognizing the shape of a problem before building anything.

What it exercises

• Builds a pipeline where each stage's output is the next stage's input.
• Trades accuracy against speed and cost instead of maximizing one alone.
• Recognizes a repeated problem shape and reuses a solution that already worked.
• Reads a failed run to find which stage caused the error.
• Names what a model, a dataset and a training step actually are.

Why we rate it Moderate Worksheet score 66 / 100

1. FitHow much real knowledge or skill does the play itself make you use?70weight 45%

   The vocabulary of machine learning is the puzzle, not a wrapper around it: to pass a level you route data through a model, notice that accuracy is limited by your inputs, and pick a different approach. That is a real domain reduced to a small set of levers, which is why fit lands at 70. It is not higher because the blocks abstract away everything mathematical, so a player learns the shape of the field without any of its substance.

2. TransferWill what you practise here work outside the game?62weight 30%

   Someone who plays this arrives at a real explanation of machine learning already holding the words and the general flow, which makes an explanation land faster than it otherwise would. That is genuine, but it is vocabulary and orientation rather than a usable skill, and no study shows it lasting. At 62, transfer credits recognition and intuition, not anything closer to writing code or doing the mathematics.

3. PracticeDoes the game make you use that skill again and again, as it gets harder?65weight 25%

   There are many contracts and they all use the same building and rewiring action, so a player returns to the core idea dozens of times across the campaign, and optimization pressure keeps finished levels worth revisiting. That volume earns 65. It does not go higher because the campaign is finite and the block set stops growing well before the end, so the practice has a ceiling.

What the evidence says

Editorial onlyNobody outside our desk has checked this. It is our reading of how the game plays.

The level is Editorial only, based on Luden.io's own game page and the Apple App Store listing. Those confirm the developer, the visual node-based programming, the cat and startup framing, the range of platforms, and the mobile terms: a paid app at $4.99, no in-app purchases, no advertising, rated 9+. Luden.io states that the game lets a player learn how machine learning works; that is the developer's claim, not a measured result, and no independent study of the game exists. What can be said with confidence is narrower and checkable: the play forces a player to use the vocabulary correctly to pass a level.

• Official game page ↗