Code reviews create too much friction in today's world, and I believe removing 90% of them and making quality assurance the developer's responsibility, supported by their set of agents, is the right way to go.
I'm right there with you, as someone who has taken a lot of responsibility for the code they write in the past, I've never really found code review useful.
But I've found also fighting to demand everyone take that level of responsibility a losing battle. Code review was an invention to solve the problem that not everyone puts a high level of care into the code they write and there are subtle bugs. With AI agents, that gets turned to the extreme - maybe we start to demand all people writing code need to have their army of ai reviewers as well, or maybe we start to see people specialize in building and running agents for code review and not doing much development themselves?
There is no other way, agents can't take on responsibility. That is why software engineers are not going anywhere. But you can't transition to an engineering manager and expect to not have more responsibilities. And with agents, that is exactly where we're heading.
Yeah, I have a hard time imagining working at full agent speed on a larger team within the same codebase. I wonder if we'll break out into sub-specialties even more extreme than what I outlined in the post. The best Claude Whisperers on your team do the development, everyone else builds support tooling to keep them going at full speed.
Code reviews create too much friction in today's world, and I believe removing 90% of them and making quality assurance the developer's responsibility, supported by their set of agents, is the right way to go.
I'm right there with you, as someone who has taken a lot of responsibility for the code they write in the past, I've never really found code review useful.
But I've found also fighting to demand everyone take that level of responsibility a losing battle. Code review was an invention to solve the problem that not everyone puts a high level of care into the code they write and there are subtle bugs. With AI agents, that gets turned to the extreme - maybe we start to demand all people writing code need to have their army of ai reviewers as well, or maybe we start to see people specialize in building and running agents for code review and not doing much development themselves?
There is no other way, agents can't take on responsibility. That is why software engineers are not going anywhere. But you can't transition to an engineering manager and expect to not have more responsibilities. And with agents, that is exactly where we're heading.
I think AI works better for small teams, like 1-3 devs. More ppl == more headaches
Yeah, I have a hard time imagining working at full agent speed on a larger team within the same codebase. I wonder if we'll break out into sub-specialties even more extreme than what I outlined in the post. The best Claude Whisperers on your team do the development, everyone else builds support tooling to keep them going at full speed.