DeepLint

Humans write flawed code. Code that is buggy, inefficient to execute and unreadable, making it hard to maintain and share across teams. Debugging costs software companies an estimated 575 billion dollars a year. Until now, the main tools that have been used to tackle flawed code are linters, automated code testing and code review. Unfortunately, linters only catch surface-level issues, automatic testing is expensive to design, implement and run, while code review disrupts flow, forcing context switches on developers and slowing the development and deployment of products. DeepLint is a tool that aims to reduce the cost of buggy code by providing instant code review for developers. At present, it is implemented as a plugin for VSCode. DeepLint uses a large language model trained on GitHub to spot subtle errors in code. It then flags these bugs to developers and suggests fixes.