What Happened
Justin Jackson published an essay on March 7, 2026 examining how Claude Code is reshaping team roles at software companies. The core observation: when an AI coding agent is good enough that product managers can ship pull requests without developer involvement, the traditional division of labor starts breaking down.
Jackson documents specific cases. One founder reported PMs wanting to write code themselves instead of filing tickets. A software company president said their Head of Product now ships PRs directly, bypassing developers entirely. One executive stated they are no longer hiring specialists because "in this new era, generalists win."
The piece references Kent Beck's widely-shared observation that captures the anxiety: "The value of 90% of my skills just dropped to $0. The leverage of my remaining 10% went up a thousand."
Jackson calls the current state a "Mexican standoff" where engineers, PMs, and designers each believe they can now perform the others' roles. His worry is that left unchecked, this dynamic produces jockeying for position, animosity, and jealousy between team members.
His proposed solutions include AI-driven pair programming between PMs and engineers, and reviving the 37signals two-person team model to channel the new capabilities into collaboration rather than competition.
Why It Matters
This is the first serious examination of Claude Code's second-order effects on team structure. Most coverage focuses on individual productivity gains. Jackson is asking the harder question: what happens when everyone on a team suddenly has overlapping capabilities?
If you manage or work on a software team using Claude Code, the pattern is probably already familiar. The PM who used to write detailed specs is now opening pull requests. The designer who used to hand off mockups is now prototyping in code. The pressure to prove value shifts from "can you do your job well" to "can you do everyone's job."
For individual contributors, the implication is clear: deep specialization alone is no longer a defensible position. The remaining value is in judgment, taste, and the ability to coordinate complex work across domains.
Our Take
Jackson is right about the diagnosis but undersells the severity. This is not just a team dynamics problem. It is a fundamental restructuring of how software gets built.
The two-person team model he suggests is not new, but it is newly viable. When Claude Code can handle 80% of implementation work, you genuinely do not need five specialists. You need two people with good judgment and complementary perspectives.
The real risk is not that teams will fight over territory. It is that companies will use this as justification to cut headcount without adjusting expectations. "You have Claude Code now, so you can do the work of three people" is already a common refrain. The teams that navigate this well will be the ones that redirect the freed-up capacity into higher-quality output rather than just doing the same work with fewer people.
If you are using Claude Code on a team today, Jackson's essay is worth reading in full. The competitive dynamics he describes are real, and the sooner your team has an honest conversation about them, the better.