Even in pre-AI world, there were no testers around me for the past many years, and those people who actually tested the final product before the release (their titles were probably something like "release engineer", not "Tester" or "QA") would file a bug if a problem is found, but they were never responsible for the bug I introduced. The commits had my name on them, and people would come to me with questions. If I could not answer the question because I forgot the code, I would sit down and figure out.
Now with AI, at least after a ~year of using AI for development, things haven't changed in terms of responsibility: my commits still have my name on them, and people will come to me with questions. If I cannot answer the question because I forgot the code or never wrote it in the first place, I will sit down and figure out.
As for the reviewer, I never saw anyone being grilled for stamping buggy code. People make mistakes all the time, including when reviewing someone's code. Code reviews are nice to have and catch real issues, but it's naive to think that they could protect from code breakage, regardless if it's post-AI or not.
Makes sense, but since companies are using coding agents more. Like if you see some repos the top contributor is claude. Since PRs are being made by agents, if you do a git blame claude is the one responsible for the change. In that case who's responsible? Reviewer?
Same goes for my workplace. Human-signed commits only. Companies which are not paying attention will learn eventually. Writing code with AI requires a lot of effort from developers. Specs must be crystal clear, testing philosophy must be good, edge-cases rooted out, regression tests written in advance and proven. In the end, a human being is responsible for the final output; be it the dev who controlled the agent, or the CEO who kicked all devs from the team and then told Claude to create an application.
Even in pre-AI world, there were no testers around me for the past many years, and those people who actually tested the final product before the release (their titles were probably something like "release engineer", not "Tester" or "QA") would file a bug if a problem is found, but they were never responsible for the bug I introduced. The commits had my name on them, and people would come to me with questions. If I could not answer the question because I forgot the code, I would sit down and figure out.
Now with AI, at least after a ~year of using AI for development, things haven't changed in terms of responsibility: my commits still have my name on them, and people will come to me with questions. If I cannot answer the question because I forgot the code or never wrote it in the first place, I will sit down and figure out.
As for the reviewer, I never saw anyone being grilled for stamping buggy code. People make mistakes all the time, including when reviewing someone's code. Code reviews are nice to have and catch real issues, but it's naive to think that they could protect from code breakage, regardless if it's post-AI or not.
Makes sense, but since companies are using coding agents more. Like if you see some repos the top contributor is claude. Since PRs are being made by agents, if you do a git blame claude is the one responsible for the change. In that case who's responsible? Reviewer?
It's not allowed where I work; the commits are signed by a human.
Same goes for my workplace. Human-signed commits only. Companies which are not paying attention will learn eventually. Writing code with AI requires a lot of effort from developers. Specs must be crystal clear, testing philosophy must be good, edge-cases rooted out, regression tests written in advance and proven. In the end, a human being is responsible for the final output; be it the dev who controlled the agent, or the CEO who kicked all devs from the team and then told Claude to create an application.
There shouldn't be a difference, no AI tool is responsibile for damage done to end customers.