More Tokens, Fewer Engineers? What One Company's Unusually Honest Layoff Note Teaches
Shifting money from salaries to AI compute can make sense, as long as the people who judge the work are the last thing you cut.
Note: This story was written with AI assistance and edited by me.
Imagine a small engineering team that adopts coding agents and, within a few months, ships more changes in a week than it used to ship in a quarter. Everyone is excited. Then the cloud bill arrives, the agents start waiting in line for compute, and the finance lead asks an uncomfortable question: if the machines are doing most of the typing, why is almost the entire budget still going to salaries? That scenario is made up, but versions of that conversation are now happening in real companies.
One of them chose to talk about it openly last week.
What happened
On November 13, 2026, Uncountable, which makes R&D software for scientists developing new materials, chemicals and formulations, let go of 11 engineers, roughly a quarter of its engineering staff. Rather than stay quiet, the company published a page for job applicants explaining the decision. Its co-founder and chief technology officer, Jason Hirshman, wrote that the company had sized its team by turning product scope into a budget and the budget into headcount, and that this approach stopped working during 2026.
According to the company, agents wrote 91% of its pull requests in October 2026, and it merged more than 60,000 of them that month, with a median of 19 minutes from opening to merge. A year before, roughly the same team merged 1,885, with a median of 16.5 hours. The company said its engineers were now held back by a shortage of agent compute rather than by hours in the day, and that because salaries and model usage are paid from the same pool, moving money toward compute meant losing good people. It described the error as its own and said the decision reflected nothing about the individuals.
Around the same time, Israeli business press reported that cybersecurity company Axonius plans to reduce its workforce by 66 people as it reorganises engineering around AI. And a report citing The Wall Street Journal described AI startup Serval offering $30,000 signing bonuses while hiring 100 engineers this fall.
That is the news as reported. What follows is my view.
The honest part of the note
I respect how Uncountable handled the communication. Most companies announce cuts in vague language and move on. This one named the planning mistake, took responsibility for it, and left the explanation online for anyone thinking of joining. That is harder than it sounds, and it gives the rest of us something real to learn from.
The planning lesson is also correct. For decades, engineering budgets have assumed that more output needs more people. Once a single engineer can direct agents that produce many times their own manual output, the budget has a new and growing line item, and leaders who ignore it will either overspend without noticing or starve their teams of the tools that make them productive.
The part I would not copy blindly
My concern is with how other leaders might read this. The easy takeaway is a formula: fewer engineers, more tokens. The company itself argues against that. Its page says that directing agents without understanding the systems underneath produces fragile software, and that its real advantage is the depth and rigor of its engineers. It is still hiring, just fewer people than it planned, and with a high bar.
When output grows that fast, the things that do not speed up are the human ones: choosing what is worth building, noticing the quiet mistake, deciding how the pieces fit together, and taking responsibility when something breaks for a customer. Agents make those skills more valuable. A team that trims the people who hold them, just to buy more compute, may end up shipping faster while understanding less of what it ships.
The company also says most of its merged changes go through without a person approving them, relying on its own automated reviewer. It has clearly invested heavily to make that safe, and I am not in a position to judge their results. For a typical business, though, especially one handling payments, health data or customer records, I would want a named person accountable for every important change until an automated system has proven itself over time.
How I would plan for this shift
If I were setting next year's engineering budget, I would start here.
Put salaries and AI spend in one plan. Reviewing them together each quarter lets you shift the balance slowly instead of all at once.
Track outcomes instead of volume. Bugs that reach customers and the cost of each change that actually ships tell you far more than the number of pull requests.
Reward judgment. Promote the engineers who can explain why a change is correct and who build the checks that keep agents honest.
Give every system an owner. Each agent and each critical workflow should have a person who answers for it.
Be open with your team. People handle change better when they hear the reasoning early and directly.
The other side
It is fair to say that a lean team with plenty of compute might be exactly right for a company like this one, and outsiders rarely see the whole picture. Not every cut labelled as AI is really about AI, and not every company will face the same trade-off. The demand for engineers who use AI well, shown by bonuses like Serval's, suggests the market values skill more than ever. My point is simply about sequence: first decide where human judgment has to stay, then decide how much compute to buy.
What I have learned
Earlier in my career, as a partner, I helped grow a software consulting firm from about $6M to $60M in under three years. Growth never came from raw volume. It came from experienced people who made sound calls and taught others to do the same, and that lesson feels more relevant now than it did then.
A simple place to start
This week, list the five decisions in your engineering process that you would never want made without a person involved. Then check whether your budget and your hiring plan actually protect the people who make them.
About the Creator
Enjoyed the story? Support the Creator.
Subscribe for free to receive all their stories in your feed.
Comments
There are no comments for this story
Be the first to respond and start the conversation.