Bengaluru-based Indian developer Ashutosh Rath has released a console application called Atrophy, a tool that aims to help programmers keep their own coding skills in shape as they increasingly rely on AI agents to write code.
Bengaluru-based Indian developer Ashutosh Rath has released a console application called Atrophy, a tool that aims to help programmers keep their own coding skills in shape as they increasingly rely on AI agents to write code.
Atrophy evaluates a developer’s skills using a system similar to chess’s Elo rating, and regularly offers short exercises in five areas:
The exercises cover Python and JavaScript, have three levels of difficulty, and thanks to seeding generation, new options are issued each time. First, the user passes a basic test — one exercise from each direction, which takes about 25 minutes — and then receives a starting rating. Then, Rath advises doing short (5–10 minutes) drills two to three times a week; the application itself chooses an exercise from the direction that has been neglected for the longest time.
Each of the five directions is rated separately and starts at 1200 — there is no hard minimum or maximum, so the rating may well fall below the starting value. Inactivity in the application does not lower the direction rating, but it reduces the system’s confidence in its accuracy.
Separately, Rath suggests going through drills with an AI assistant once a month — the results of such exercises are recorded separately so that you can see the gap between what the developer can do with the help of AI and what he is still able to do on his own.
The author emphasizes that Atrophy is not an anti-AI tool. According to him, the application measures the gap between capabilities with and without AI, since these skills can imperceptibly «rust» without any warning.
