[steveBK123]

The cost/ROI vs productivity thing I see on my team / hear from others is mostly that people are leveraging LLM code to do lower value work they previously never would have done. Lots of "I finally improved my test coverage" and "so I finally got to that ticket thats been in the backlog 2 years" and "I finally did that big refactor" and "I spent 2 weeks on the codebase making it easier for LLM to understand it" kind of stuff.

I have heard fewer anecdotes from people I know of the form "I do my job faster, shipping features 50% faster using LLMs". I think this is a project lifecycle thing too. An existing 5 year old project with users, stakeholders, product team, etc probably is going to null out any coding speed up.

I recall a very "agile" team I was on where we met with users weekly to get their requests, demo the changes made, and usually half the meeting was them changing their mind about what they said last week. I don't think delivering features 2x faster and meeting twice a week would have worked.

[orionblastar]

https://youtu.be/WuhAaMSXD9A AI costs more than using Humans. In 2030, they might have lower-cost tokens.