You've read a dozen posts about how AI coding tools are "changing everything."
Most of them treat it as structurally novel.
It isn't.
And that's actually reassuring. (what? no existential dread?)
Every major cognitive tool since the 1950s — word processors, spreadsheets, relational databases, email — hit knowledge work and followed the same pattern:
- Specific tasks got automated (bookkeeping, typing, data retrieval, correspondence)
- Demand for the underlying capability expanded because the cost dropped
- New roles emerged adjacent to the tool
- Skill requirements migrated upward within the domain
- Net employment grew
The spreadsheet didn't end accounting — it ended bookkeeping as manual drudgery and created a generation of analysts who could model scenarios entire teams previously couldn't run.
Sound familiar?
Because Copilot isn't ending software development — it's ending boilerplate as manual drudgery.
(read: just changing the ol' jobbie job)
What's genuinely useful here: cognitive tools historically raised the ceiling on individual capability rather than just cutting headcount. The surviving bookkeeper became an analyst. The surviving junior dev becomes... something with more leverage and more abstract scope.
There's also the Solow Paradox buried in here — early computing productivity gains were real but statistically invisible for years.
If you're watching AI productivity metrics right now and finding them underwhelming, that's a known phase of this curve, not a referendum on the technology.
This is part of the "Hitchhiker's Guide to Knowledge Work" — a series worth reading if you want a grounding framework beyond vibes:
Top comments (0)