By hiring. Or, rather, not hiring.
At every single company I worked for we had huge trouble finding new people skilled enough to understand our product/codebase and extend it. A task that took a new hire two weeks to do sloppily would take any of our engineers about two hours to do properly. Recently we had one guy build nine prototypes of a test jig over the span of two months and none of them worked reliably. I ended up drafting my version one evening out of boredom (it wasn't even my responsibility) and not only it worked - it ended up being the favorite of our test team.
Needless to say, we had to lay off that engineer. He simply wasn't producing the value we paid for him.
I'm not saying there are 10x programmers. Maybe there's just 1x and 0.1x. Maybe it's 3.33x and 0.33x. But the discrepancy is there, and it's disheartening.
The Lake Wobegon effect is prevalent in almost any self-assessed skill. Nobody wants to admit that they're below average.
Of course, that doesn't change the fact that there is an underlying distribution. Just because everyone thinks they're an above average driver doesn't mean there aren't some people who actually are.
Like revenue generated, or cost saved, or lines of code saved, or time needed to deliver work, based on current or previous work.
Is this bullshit for you?
Honestly? Yes.
I have no idea how many dollars of "revenue generated" or "cost saved" are attributable to me as individual. Or anyone else.
Using lines of code as a performance metric? Get ready to see a multi-line Javadoc on every local variable, or get ready to see some insane Perl-style one-liners, depending on how you're using the LOC number.
"Time needed to deliver work"? Again, what is "work"? If your data consists of JIRA tickets, then get ready for a culture of writing JIRA tickets at a contrived (and inefficient) micro level of granularity.
There's no good way to objectively measure your "value" at your CURRENT company. And what do I do when interviewing you to come here for your next job... just take your word for your measurements at the previous gig?
Out of curiosity, are you a programmer? Because subjectively making stuff up, while under the honest delusion that you're guided by objective numbers, is the single most "MBA" trait there is.
Hypothetical example. Your employer has 1 billion users. You create compression that cuts down storage costs by 4%. Run the math and you see how much you saved.
If you saved $0.01 per user, you saved $10m. If the average salary is 100K and you saved $10m, you performed at 100x.
I agree the potential for "X" performance is hard to spot or measure when one isn't in proximity to big problems.
Things are accomplished by teams, not individuals.
There was always a huge amount of players at work helping the upper execs and sales teams, from the devs to HR.
I guess OP's post was and the article itself are noting that working your ass off and making a big contribution in SV as an engineer can sometimes be a bad deal, especially when the engineer in question really is talented.
With a small experienced team and proper division of lager into vertical slices, that can be done.
In a modern environment, never have I been asked to solve a business problem with:
No internet access, No configured editor, With massive time pressure, In a room full of strangers, In my chicken scratch with a fat marker.
It's not that whiteboard interviews are bad. They're just severely outdated.
They stem from a previous time before it was cheap and easy to interview engineers on actual hardware.
We know that whiteboards are a bad metric. Dozens of scientific studies show that the most effective hiring practice is a general intelligence test coupled with a work sample test.
It's easy enough to create a coding problem as a work sample.
Why do we test a proxy for the work, rather than the work itself?
When I get offered one, I send back a quote to solve their problem, at my standard contracting rate.
Some companies do all day interviews or at least 3-4 hrs, what is really the difference between spending the day in an office or at least few hours versus at home with your tools and setup?
Are you less offended if I give you 2 hrs of problems to whiteboard while I watch?
There is no situation in which a company can not spend at least some time evaluating your skill set. For me the best would be to hire you based off your resume and see how it goes for 90 days. However, its often unfair to the employee - especially if they have a position currently.
Hiring is a skill for both parties. some people are good at the process naturally, other people put time into being good at it, the third group has no clue and has no interest in learning.
How do you measure "X", anyway? During the whiteboard interview process? We all regularly discuss how subjective and broken that process is. Do you measure it after the hire, by tracking the number of story and task tickets closed? Now you just have people gaming the system, with a sprint board full of bullshit.
Everyone wants more money. No one wants to work for less than a certain baseline. Everyone's pissed when they find out what their peers make. So if you're an employee who thinks you're underpaid, then test the market. And you're an employer who routinely offers great candidates many multiples of the market rate, then please let me know where I can send my resume.