The existential dread hits hard when a coding agent finishes a task in 45 minutes that would have taken you a full week of manual labor, and the code actually works on the first pass. It feels like your value as an engineer is evaporating. But the reality is that translating a spec into syntax was never the most valuable part of the job; it was just the part that took the most time.
Why the "AI took my job" feeling is a lie #
The panic usually stems from the fact that we've spent years mastering the nuances of specific languages or frameworks, only to find out an LLM can hallucinate its way through a boilerplate setup faster than we can type. If your entire professional identity is "I can write a clean React component," then yeah, you're in trouble.
But if you look at the actual lifecycle of a project, the "coding" part is the smallest slice of the pie. The real work is deciding what to build, figuring out why the system is crashing in production under a specific load, and managing the messy intersection of business requirements and technical constraints. An agent can write a function, but it can't tell you why your architectural choice from six months ago is now the primary bottleneck for your database.
The gap between an agent-user and an engineer #
There is a massive difference between a junior dev using an agent to build a CRUD app and a seasoned engineer using an agent to accelerate a complex system.
- **The Junior approach:** Prompt -> Copy/Paste -> Error -> Prompt "fix this" -> Hope it works.
- **The Engineer approach:** Prompt -> Review logic for edge cases -> Identify a potential race condition the AI missed -> Refactor the prompt to handle the state correctly -> Deploy.
Software engineering has always been a chaotic mess #
Complaining that AI is changing the industry too fast is funny when you realize this field has never been stable. Every five years, the "industry standard" gets wiped out by a new framework, a new cloud paradigm, or a new language. If you entered software development as a passion, you basically signed up for a lifetime of radical instability.
The only way to survive is to stop clinging to the "way things were done" and start treating these agents as an extension of your brain. The tool is changing, but the problem-solving requirement isn't. If you can't find value in your work once the boring syntax part is automated, you weren't really solving problems anyway—you were just typing.
Next Claude 3.5 Sonnet writes clean code but you still need a brutal testing pipeline →
All Replies （0） #
No replies yet — be the first!