Gergely Orosz

Something I hear very little talk about:

How AI coding tools are so much LESS useful when used on existing, large codebases at work (with custom frameworks, conventions, coding style etc)

... compared to doing greenfield work or side projects

So common for me to hear: "yeah I love it on my side projects, but at work it's 'meh'"

I'm getting details talking with devs at the likes of eg Google, Meta, Microsoft: the companies building some of the best AI coding tools out there!

And yet, for their existing codebases, the usefulness is marginal. Mostly for autocomplete (that has a higher miss rate than for greenfield)

And yes, surely there are workarounds. I just don't hear much of these used or successfully used!

Point is almost all success stories I hear are greenfield ones or small projects, or ones started with these tools

Using on larger one a bigger challenge

https://x.com/clemkeirua/status/1927717022424608923

And yes, a project that started as greenfield by LLMs inevitably will get big enough to not fit in the context window / need refactoring

Once it gets here, it's a different ballgame, and it's not the strength of those tools at all...

https://x.com/xravaggio/status/1927717773964116255

As a specific example: a large tech company (thousands of devs) got a Cursor license for all devs to use.

They checked in a few months later, and ~half the devs stopped using it. Just wasn't as useful *inside the company* This is a "top" tech co!

From https://newsletter.pragmaticengineer.com/p/microsoft-ai-dev-tools 