You're identifying something real that job descriptions haven't caught up to yet.

The honest answer: **it's both, and it matters which direction you lean.**

**You're doing engineering if:**
- You're evaluating trade-offs (why this architecture over that one)
- You catch logical errors the AI misses
- You understand failure modes and edge cases
- You're making decisions about what *not* to build
- The review phase is where real thinking happens, not just cleanup

**You're doing PM with a terminal if:**
- You're mainly describing requirements and accepting outputs
- The review is syntax checking and "does it look right"
- You can't articulate *why* the AI's approach works or doesn't
- You're essentially a specification writer

The uncomfortable part: **both are real jobs, but they pay differently and have different demand curves.**

The engineering version (deep review, architectural decisions, knowing *why* things fail) is still rare and valuable. The PM-with-terminal version is becoming commoditized faster.

What likely matters for your career:
- Can you debug AI-generated code at the logic level, or just at the "does it compile" level?
- Do you understand the domain well enough to catch when the AI's solution creates future problems?
- Are you learning more or executing more?

If it's the first three—you're still engineering. You've just shifted from implementation to verification, which is actually higher-leverage work. If it's the last category, yeah, that's different and riskier long-term.

What's your ratio actually looking like?