The new frontier models are passing SWE-bench with huge gains and getting integrated everywhere. We all use Copilot/Claude/etc. now. Be honest: How much of your day is writing new code vs. reviewing, orchestrating, and securing what the AI spits out? Is the role of “Software Engineer” fundamentally shifting?
Been here for three major hype cycles. The best it does is boilerplate faster. The cost of debugging a hallucinated edge case in production, plus the security audit time? It’ll net zero out the velocity gain. I’m sticking to my senior IC role of writing and owning the critical path. Show me the 3-year P99 uptime chart, then we talk.
It’s not about new roles, it’s about making us 2x more productive for the same salary. The business doesn’t see “orchestration” or “security”; they see 50% faster feature delivery. We’ll just be doing two jobs: writing the 50% hard stuff, and fixing/securing the 50% easy stuff. Technical debt is just going to balloon.
| This is the final move to declarative engineering. We define the intent, the LLM generates the implementation, and our platform validates correctness. We’ll be writing less code, but the code of our tests and CI/CD pipelines will become the most critical asset. We need better tools for governance and provenance of machine-generated code. |
|---|