The Coding Part Is Getting Cheaper. Engineering Becomes the Job.
Software is changing, but not in the way most hot takes frame it.
The interesting shift isn’t “AI will replace developers.” It’s that code generation is collapsing in cost for a huge chunk of work that many teams do every day.
And if we’re honest, a lot of that work has been “solved” for a while — just with different tooling.
CRUD has been solved for years
Most products contain a lot of CRUD:
- Create/edit records
- List and search
- Permissions and basic workflows
- Simple dashboards and admin UIs
We already had frameworks that make this cheap. We already had third-party services that absorb it. We already had component libraries and templates that reduce UI work to assembly.
What’s new is not that CRUD exists — it’s that generating the surrounding code is becoming fast enough that the coding part stops being the constraint.
The bottleneck was never writing code
In real teams, the bottleneck has always been:
- Requirements gathering and ambiguity
- Turning product ideas into technical decisions
- Designing database interactions and service boundaries
- Reviewing changes safely
- Integrating with other systems
- Keeping the thing online
That list looks suspiciously like “engineering” rather than “coding.”
Which leads to the shift we’ll likely see: many “dev” tasks move into roles that are closer to SWE + SRE + DevOps — because those are the roles that already center design, review, reliability, and accountability.
Accountability doesn’t disappear
The most important point is the least exciting: humans remain accountable.
Compliance, regulation, and the real world don’t accept “an agent did it” as a process.
When something goes wrong, someone has to:
- Understand what happened
- Fix it quickly
- Explain the impact
- Prevent recurrence
- Own the decision making
That’s why the biggest risk concern isn’t code generation itself — it’s overconfidence from business leaders.
If organizations treat generation as “we no longer need engineers,” they create a new kind of fragility:
- Nobody understands the system well enough to repair it
- Outages take longer to resolve
- Blame becomes a mess
- Reliability and security erode quietly until they explode loudly
Where humans stay essential
If you had to compress it, humans stay essential in three places:
Requirements and design
Turning product needs into technical shape. Making tradeoffs. Designing boundaries. Thinking through failure modes.
Maintenance and review
Quality control. Risk management. Ensuring changes are safe. Spotting subtle issues that “look right” but behave wrong.
Keeping products online
Operations. Observability. Incident response. Reliability engineering. The “it’s 3am and production is on fire” work.
The coding part is getting cheaper. That’s real. But engineering — the discipline of building reliable systems that humans can understand, maintain, and be accountable for — that’s becoming more important, not less.