Have been using Claude AI in day to day work in the past months, and have been felt humbled.

If we have a smaller team maintaining the project, what will be everyone’s role then?

Software Architects + Project Managers + Senior Code Reviewers + Dev Ops Lead, plus a group of contractors repeating “best known method”, is enough for managing a super large C++ code base?

I think that might be it – we don’t want extra “flatten” people in the same layer, to increase the communication cost; and we don’t want to have “regular devs” anymore.

So far this looks very promising to me – I think the tricky part is “best known method” will be improving significantly fast; Ask each team to develop their own BKM seems sick – there should be an upper level team in the “business group” level (entire GPU org, at least!), the team should have strong cross team understanding, deeply touching historical concerns across all teams.

Considering the upper engineer team needs to handle all the legacy concerns, they should be built by experienced engineers especially promoted from internal; and below “Software Architects + Project Managers + Senior Code Reviewers + Dev Ops Lead, plus a group of contractors” could only be available for external hires and “average” engineers.

Another concern for me is that – Idk what can I do to stand out – under this aggressive workflow. I am not touching algorithms anymore, my code review for AI-based PRs only focus on – documents, function signatures, members, and that’s it. My skill set has been shifted gradually from individual contributor to “workflow enabling”, more like a traditional engineering management role.

I think the risk of US debt has pushed the AI stock market because US economy cannot afford failing AI anymore; but also this trend will push companies to switch AI solution asap, and push higher the stock price to the moon, as a risk management.

Let’s see what’s next!