Specification is the new bottleneck, and it was always underrated
An agent can implement a well-specified requirement quickly and well. It cannot resolve genuine ambiguity about what the system should actually do when two stakeholders want contradictory things — that's a human judgement call, dressed up as a technical question.
Engineers who write precise, complete, testable specifications are now the limiting factor on delivery speed, which is a direct reversal of a career spent assuming implementation speed was the bottleneck.
System-level reasoning doesn't commoditise the way syntax did
Understanding how a change ripples through a distributed system's failure modes, data consistency guarantees, and operational cost is a different kind of knowledge than knowing a language's syntax — it's built from scars, incidents, and years of watching systems behave under real load, and it doesn't get cheaper just because code generation did.
This is the leverage point worth deliberately investing a career in now: not resisting agentic tools, but becoming the person whose judgement about what to build is worth more than the tool's ability to build it fast.
Typing code fast was never the scarce skill. It just used to be the bottleneck. Now that it isn't, what was always actually valuable is finally visible.MNB Software Engineering Practice
Key actions
- Invest deliberately in writing precise, testable specifications
- Build system-level reasoning through real incidents, not just syntax fluency
- Treat ambiguity resolution as the genuinely senior skill it's always been
- Measure your own value by judgement quality, not implementation speed