Instead of everyone telling their personal anecdote, we should look at the actual research.

Asking ChatGPT to summarize the state of AI coding research that’s not done by one of the providers, and I feel like I got the sanest summary of the actual state of things:

> If you weight independent work more heavily: current best evidence says don’t expect speedups for senior devs in complex, well-known codebases without process changes; expect security pitfalls unless you add guardrails. Gains look more reliable for novices, unfamiliar code, and templated tasks

Seems the most sane PoV on the actual benefits I’ve seen.

Actual link if you want to pick apart GPT-5s summary

https://chatgpt.com/share/68c56214-1038-8004-b1ed-359393eb15...