Yield curve for engineers has inverted

Agentic coding tools have inverted the traditional software engineer yield curve. Until recently the value that a software engineer brings to an organization was a monotonically decreasing function of the engineers age.

This is not ageism, it is observation. The general decline in value with age was the result of a multitude of factors.

Younger engineers were more familiar with emerging technologies (an overall trend itself the consequence of many disparate forces), they tend to accept lower salaries, and are typically more willing to put in extra effort as they build their careers.

Older engineers on the other hand will have expertise in mature technologies (necessarily so as expertise takes time to acquire), they will demand higher salaries for that expertise, and generally have more demands from the life side of the work life balance.

Contribution minus cost equals value, and that is an equation that up to now favors youth. Bob Dylan had this old paradigm pretty much nailed with his classic words,

Your old road is rapidly agin’
Please get out of the new one
If you can’t lend your hand
For the times they are a-changin’

Indeed, the times they are a-changin, agentic coding tools have inverted the software engineer yield curve.

Agentic coding tools magnify the value of experience, wisdom has caught a bid.

Suddenly, the knowledge acquired maintaining a massive JVM code base is transferable, via LLM coding tools, to an equally massive Python code base even if the engineer has only passing familiarity with Python. Understanding of the architectural broad strokes and where the necessary and unnecessary complexities lie transcend the particulars of implementation language. No junior will be on the lookout for back-pressure in a queue, but the senior will focus the LLM agent on such issues, guiding its willingness to code and knowledge of syntax to produce solid work.

This is not belief, I’ve observed this phenomena personally and among colleagues. I have used “hands free coding” (keep your vibes to yourself) to build complex image processing pipelines, multi-faceted web scrapers, and React CRUD apps. I was able to produce full and functional software using agentic coding tools because over a 30 year career I developed a foundation of software organizational principles and along the way I have picked up just enough knowledge of OpenCV, BeautifulSoup, and React to prompt on point.

I have seen otherwise brilliant junior engineers struggle with the firehose of LLM generated code. Asking for too much, chasing down irrelevant details, spitting up self fulfilling over commented tests. These are the symptoms of an engineer too young to drive the train and going off the rails.

If you are an aged engineer embrace these new tools, you’re the biggest beneficiary. If you are a junior engineer, realize you wield great magic and you don’t want to end up like the Sorcerer’s Apprentice. If you are responsible for getting things done and product out the door, take a cue from the more recent words of Corb Lund,

I want old men making my whiskey
I want old men singing my blues
And I want old men teaching my horses
’Cause there’s just some things young men can’t do
Like the old boys do