Most coders want AI to write code faster: I want AI to write FASTER CODE.

Minimaxir's blog is a hub for machine learning enthusiasts, offering tutorials, project showcases, and insights into the latest trends in AI and data science. With a focus on practical applications of machine learning, Minimaxir shares tips, tools, and resources for building and deploying ML models. Developers can learn about deep learning frameworks, natural language processing techniques, and AI-powered creativity, gaining  skills to tackle real-world problems.

Max Woolf's Blog

The post explores the concept of iteratively improving code quality using LLMs like Claude 3.5 Sonnet by repeatedly asking the model to 'write better code.' Initial results of using basic prompts to refine a Python code sample show significant performance improvements, demonstrating an iterative speedup from a baseline of 657 milliseconds to code that runs in about 6 milliseconds after several iterations. Further experiments using prompt engineering, which explicitly guide the LLM with detailed instructions, yield even more optimized code. However, the process also reveals the limitations of LLMs, highlighting the necessity for human oversight to ensure correctness and manage subtle bugs.

Can LLMs write better code if you keep asking them to “write better code”?

Prompt Engineering LLMs For Even More Better Code #

Next Steps For Better LLM Code Generation #

One of my most used prompts in Copilot is when I’ve written something that works but I feel like my implementation is a bit convoluted I select the code and ask it to “simplify this”. Usually works pretty well and gives good results.

I was really expecting the more intentional prompts to do better 100% of the time

So effectively basic interaction and advanced initial prompting equals next level code speeds (and much lower but still extant bugs).

After reading this I’ll be making my code more bro, that’s for sure!