Nine years into this career and I'm starting to wonder if everything I built, documented, and taught is just a very detailed instruction manual for my own exit. Not paranoia. Just pattern recognition.
1. We're the Best Training Data Money Can't Buy
- Every PR review, every Slack thread where you explain why a design decision was made, every Confluence doc you wrote at 11pm... that's training data
- Not for your junior. For the model that'll replace both of you.
- Real World: GitHub Copilot didn't get smart by reading textbooks. It read your code. Your comments. Your variable names. You literally taught it to think like you.
2. Automation Tools Are Not Neutral
- Kubernetes, Terraform, CDK. We sold these as "engineer productivity" tools. They are. They also systematize expert knowledge into configs that a non-engineer can eventually run.
- Real World: A Terraform template good enough means your manager can spin up infra without you. You wrote that template. Think about that.
3. The Junior Engineer Trap
- We mentor juniors because it's the right thing to do. It also accelerates the timeline.
- Teaching someone to fish is noble. Teaching 10,000 people to fish while Anthropic watches the recordings is something else entirely.
- Real World: Every "how I built a scalable notification system" article you publish is a free masterclass for the model that answers that question faster than you next year.
4. Specialist Skills Are Shrinking Windows
- Five years ago, knowing Kubernetes deeply was a moat. Now it's a checkbox.
- The window between "cutting edge skill" and "table stakes" is getting shorter every 18 months.
- Real World: Remember when writing Terraform modules felt premium? Now there's an AI that writes them from a one-line prompt. Your moat dried up while you were heads-down shipping.
5. The One Thing That Still Holds
- Machines own static workflows. They're catching up on dynamic ones. They're nowhere near ambiguous ones.
- The engineer who spots the wrong problem before anyone codes a single line is still irreplaceable.
- Real World: AI shipped a perfect solution to the wrong requirement last quarter at a company I know. Someone still had to walk into a room and say "we're solving the wrong thing." That someone kept their job.
6. So What Do You Actually Do
- Stop documenting your expertise and start weaponizing your judgment.
- Build things others can't spec, not just things others can't code.
- Be the person in the room who asks "should we even build this." AI can't do that yet. Your manager can't do that either.
- Real World: The engineers who thrive in the next five years won't be the best coders. They'll be the ones who know when not to code.
The machine learns fast. The question isn't whether it'll catch up. It's whether you'll be interesting enough by then that catching you isn't worth it.
Cheers🥂
Top comments (0)