The worst kind of Product guy is an "idea guy". I think every industry has folks who think they're here for their ideas. What you really want out of a Product guy is someone who can think rigorously about the problem they're solving (even when that problem is something like standardizing email signatures for your company).
The lifecycle of a software product is not an idea and then the developers get to work tapping out code. There is an idea and then a bunch of people talk about that idea and all the problems with that idea and they come up with some worse ideas and some better ideas and eventually they hammer out something that's good enough to start talking with the developers about.
There are a lot of ways to do this part. I joined the workforce in the hayday of Agile and everyone taking week long Scrum Master certification courses. There are good and bad parts to all the ways we've run software teams. I could write up a long blog post on what I think matters, but I won't (yet).
The worst way to build a software product is to skip all of that and go from idea to execution without stopping to think. We've removed all the barriers that made idea guys have to stop and work out what their idea actually is. We don't plan so much anymore. Requirements gathering is done ad-hoc in a prompt input to an LLM when it generates unintended outputs.
We're now in the age where the cost of producing code is cheap[1] and LLMs mean that any random person can vibe code something that looks halfway decent on the surface to most folks. The distance between idea and product has been shortened immensely and we now have a C suite that looks at their Product & Engineering teams and say "What's taking you so long? Codex gave me a whole app and put it online in a few hours."
This may all be my frustration with LLMs, vibe coding, and a generally rough job market surfacing. Maybe folks on other teams in other companies are finding time and value out of backlog grooming rituals and curating a backlog of tickets to work though. Maybe other teams are making architectural decisions looking ahead 5 years (or even 9 months).
Even if this is all something that I'm feeling on my lonesome I do think there is some worth to repeating the phrase that I find myself saying a lot these days "My job isn't to write code. It's to understand the systems that the code describes that satisfies the requirements of the business."
It isn't catchy, but it does explain why Claude can't just do it.
I know that LLMs are not cheap. They have a profound impact societally, environmentally, and monetarily. However, in the short term it looks like you've just replaced 6 full-time employees with a machine that won't tell you your idea is bad and waste your time arguing about it. ↩︎