Neural Networks Write Telegram Code: Developers Recommended for Dismissal

Developers of the official Telegram client for macOS have stated that almost all client-side code for the messenger is now written using neural networks. According to one of the engineers, specialists who continue to write code manually should be fired.

Artificial Intelligence Achievements in Development

Significant technical tasks have already been implemented with the use of AI. In just a few prompts, the MTProto 2.0 protocol for iOS and macOS clients was rewritten in Rust. Link to the project: https://github.com/TelegramMessenger/mtproto-engine

• A new TON blockchain explorer was created in one day
• The Claude model adapted the iOS client for the foldable iPhone Duo overnight

This development approach is shared by 80 to 90% of Telegram programmers. Colleagues who call the neural network results "slop" are considered specialists stuck in 2023.

Quality control of AI-generated code is assessed as higher than manual development. Vulnerabilities in such code are also proposed to be found and fixed using artificial intelligence.