AI coding assistants offer tiny gains – real productivity lies elsewhere
Devs spend very little time actually coding. AI coding assistants provide only marginal gains when they are
It's besides the point, *typing is not the bottleneck anyway*
https://blog.robbowley.net/2025/02/01/the-evidence-suggests-ai-coding-assistants-offer-tiny-gains-real-productivity-lies-elsewhere/
Devs spend very little time actually coding. AI coding assistants provide only marginal gains when they are
It's besides the point, *typing is not the bottleneck anyway*
https://blog.robbowley.net/2025/02/01/the-evidence-suggests-ai-coding-assistants-offer-tiny-gains-real-productivity-lies-elsewhere/
Comments
Identifying the value of the tool is important, and can be crucial.
I don't see AI LLMs and coding being a productivity win. I see it as a reducing code + easier maintenance solution.
That's not productivity, that's efficiency.
https://arc.dev/talent-blog/impact-of-ai-on-code/
Most I speak to say they find them helpful for yak shaving, filling out tests, very simple tasks, not much else.
https://www.blueoptima.com/resource/llm-paper-1/
I would disagree. You can use copilot and ChatGPT to create the base for an entire application very quickly including tests etc.
If it's being used on an existing piece of software then I can see how people might think that.
Agree somewhat re: early days, however GPT4 has been around 2 years now, wide adoption of coding assistants at least a year.
Plenty also indicating reaching limits of what LLMs can do.
Good luck with your venture!