We only use Python now (and nothing works anymore)

A new manager announced a simple, heroic solution to software complexity: “All code must be Python.” Productivity immediately declined to zero, which is exactly what happens when you replace a multilingual codebase and decade‑old tacit knowledge with a one‑line corporate policy.

Tiny model: suppose a manager’s objective is standardization (fewer tools, easier hiring, nicer org charts). Constraint: the team knows C/C++/C# and the product depends on them. The manager can either (a) pay costly retraining and a long migration, (b) let the team keep shipping, or (c) issue an uncompromising edict and hope for voluntary compliance. They chose (c). The result looks like everyone learning a new language on someone else’s time (sf3p0x1’s point), or—if you prefer malicious compliance—the team could have literally said “no” and sat around doing nothing (Zoreb1). Both outcomes are predictably efficient at enforcing the new rule and predictably terrible at shipping.

Practicalities made the joke into a tragedy: scientists were suddenly “contributors” (Hot_Aside_4637), infrastructure debt ballooned (CoderJoe1), and the mandate came with pedagogy: “take classes in your own time on your own dime” (the managerial version of asking the janitor to buy the mop). Also — and this is deliciously bureaucratic — it wasn’t enough to code in Python; it had to be Pythonic, as if aesthetics would refactor legacy C++ into cooperation.

Policy memo (imagined, sadly plausible): “Effective Immediately: All code must be Python. Non‑Python code will be deprecated. Employees may train off‑hours at their discretion. Compliance will be monitored.” Translation: you now have a policy that is easy to write, impossible to implement, and morally excellent.

Call it the One‑Language Fallacy: managers pick a single lever that signals competence (standardize!), underestimate switching costs, and watch output vanish. Not mean to dunk—this is a neat little human tradeoff: you can have consistent tooling or uninterrupted shipping, but not both without paying for it. For certain values of “efficient,” the policy was very efficient indeed—efficient at eliminating working software.