It has been quite a while since I am taking help of AI Clients (like ChatGPT, Claude, Perplexity etc.) alongside my own experimentations with local open source LLMs via Ollama GUI, for proof-reading, brainstorming new ideas, bringing the outsider perspective and validating multiple hypothesizes etc.
However, it has been quite recent experience completing my work using AI assisted coding platform (ex. Claude Code, GitHub Copilot, Gemini Code Assist etc.)
My one-line summary for these coding assistants is like this:
“Think of them as recent college graduates, with solid foundational skillsets and enthusiasm, channel their energy with proper guidance to get productive output.”
When I shared this summary with many of my colleagues at work, they actually liked it as decent summarization.
Being an experienced Data Scientist, I have use multiple tools/technologies to solve business problems. However the focus has always been to get quality output over quality coding.
With AI assisted coding, it seems now possible to get quality output with quality and efficient coding. I learn everyday a lot of new concepts, challenger approaches to my thought process and efficient codes which works just fine.
Having said that, I still need to validate logical flow of the code generated by AI tool, ask it follow-up questions to refine further and ensure it works as good as my own version.
Because at the end of the day, I am still responsible for driving critical business decisions and we are yet not there leave it up to AI assistants.