Adapt Your Coding Style to LLMs or Get Left Behind
There seems to be a stark divide in the developer community regarding the effectiveness of LLMs for coding. Some are disappointed, others quietly reap the benefits of these powerful tools.
I encourage everyone to approach these tools with an open mind and a willingness to experiment. My current go-to tools, Claude Opus 3 and GPT-4o, have significantly boosted my productivity.
To maximize the potential of LLMs in your coding workflow, you must grasp their strengths and limitations. LLMs are trained on vast amounts of data and code from across the web, including GitHub and Stack Overflow. They excel when working with well-established stacks that have a strong community presence. For instance, PHP/Laravel and LLMs are a powerful combination, while more niche or cutting-edge frameworks and languages may yield less impressive results.
LLMs thrive when you adhere to common programming styles and best practices. If your codebase is well-structured and follows conventions, LLMs can generate more relevant and seamless code snippets.
