A company has recently made a public announcement to admit that, intentionally, they are unable to quantify the work done by engineering. And guess what?
The GMO Internet Group discontinued remote work after observing that hourly typing volume had decreased.
The contradiction that writes itself
Here's the kicker that's supposed to be funny before it gets uncomfortably real. The exact industry that is encouraging developers to use Copilot is also evaluating them based on the quantity of their typing.
Copilot is designed to minimize your typing efforts. That's all you need to know.
Therefore, the request is to utilize the AI that codes for you while generating many keystrokes, or we will assume that you are not working hard enough. Two objectives are conflicting.
Yuuki Yamashita, an AWS Community Builder, hit the nail on the head in his breakdown from July 15, 2026, when he said that a company "demanding AI adoption while grading people on typing volume is optimizing for two contradictory metrics at once."
Unfortunately, you are not able to possess both of these.
A keystroke counter measures the wrong engineer
Some of the best pieces of code I have written have involved me writing no code for the first half-hour. Just thinking. Staring into space. Drawing a few pics on a whiteboard. Deleting more than I wrote.
Yamashita expressed this more aptly than I ever could, by saying that "a keystroke counter cannot distinguish between an engineer typing furiously to fix a self-inflicted bug and an engineer staring at a whiteboard for an hour before writing the ten lines that make the whole system simpler."
Read that twice. The metric rewards the person who created the mess and punishes the person who prevented it.
Presence is what you measure when output defeats you
This is not related to GMO. 78% of employers are using ActivTrak, Teramind, or Sapience Analytics type of software to monitor mouse movements, app flicking, and keystroke speed.
That hasn't been management at all. It has been a confession.
I will never forget what Cory Doctorow declared in September 2024 "Every accusation is a confession. When your boss tells you that he thinks that you can't be trusted to do a good job without total, constant surveillance, he's really saying, I only bother to do my CEO job when I'm afraid of getting fired."
The act of surveillance reveals more about the person watching than the one being watched.
The theater has a body count
When being there is used as a measure, that's what people will strive for. Not performance.
In May 2024, Wells Fargo terminated more than a dozen employees for using "mouse jigglers" to falsify activity, according to FINRA disclosures. But everyone learned the wrong lesson. The employees didn't create the game. The metric did.
In August 2024, Denise Prudhomme, an employee of Wells Fargo, passed away at her desk in an office in Tempe, Arizona.
The electronic system counted her as an attendee. However, she remained lifeless at her workstation for four days until an actual person realized.
This is the actual measurement of "presence." A badge swipe. Not a human. Not the output of the work. Not if anyone is breathing.
The receipts say it doesn't even work
If the surveillance was effective, you could have a debate about the trade-offs. But it isn't.
A study conducted in 2024 by Yuye Ding and Mark Ma from the University of Pittsburgh examined S&P 500 firms which had strict RTO mandates. The results showed no statistically significant improvement in financial performance. Productivity either.
What it did produce:
→ 14% higher departure rate among senior employees
→ 23% longer time-to-fill for open roles
Your most talented employees are the first to leave and their positions remain vacant for extended periods, all because of a measure that ultimately doesn't influence the target it's supposed to safeguard.
What I actually think
Assessing results can be challenging. It demands a sense of aesthetics, understanding of the situation, and openness to evaluating the quality of work rather than the quantity.
Tracking the number of keystrokes is a simple task. All you need is a spreadsheet program and a lack of trust in your colleagues.
If leaders choose the easy option, it indicates that they lack the ability to take on the difficult one. That's the bottom line.
If you need a keystroke counter to distinguish good work from bad work, the problem was already there!
My question is, have any of the monitoring metrics your company has embraced actually improved the quality of the work, or did it simply improve how highly skilled people get at gaming the metric?
Top comments (0)