Bryan Cantrill's line about Larry Ellison is the best thing in this genre: think of him the way you think of a lawnmower. You don't anthropomorphize a lawnmower. It mows the lawn, it will chop your hand off, and it doesn't hate you — it doesn't have the equipment for hate.
Sean Goedecke takes the line and applies it to tech companies, and the argument is mostly right. Companies don't owe you for the hard thing you did. They aren't vindictively retaliating when they push you out. You are not partners in a mission. All of those framings import a person into a system that doesn't have one, and they make it harder to see what's actually happening.
The car is a better metaphor than the lawnmower, as he says, because it's dangerous and central to your life. But I think both metaphors under-reach in a way that matters for the practical advice.
What the machine metaphor gets right
Three things, and they're important.
Machines don't respect effort. Goedecke's point here is the strongest in the piece: the project you slaved over for months gets "nice, what's next," and the fifteen-minute tool gets wild praise. A person who behaved like that would be unstable; a machine that behaves like that is just a machine. Engineers who expect proportionality between effort and reward are constantly surprised, and the surprise is the tax.
Machines don't hold grudges. The engineer who is sure they're hated is often wrong. What looks like persecution is frequently risk management: managers deciding someone is a loose cannon whose usefulness doesn't outweigh the damage. From inside, being treated as a defective tool feels like being run out of town by people who hate you. From outside, it's a maintenance decision.
Machines don't reward literalism. This is the best section in the essay, and it's the one I'd hand to every new engineer. Companies aren't programs either. The handbook, the role guidelines, the "top priority" email — these are the legible superstructure over an illegible web of relationships and incentives. Following the rules point-by-point is not how promotions happen, and treating the stated priority as the actual priority is a reliable way to be confused and disheartened.
Where the metaphor under-reaches
A car is deterministic. A company is adaptive: it's made of people responding to incentives, and the responses change the incentives. That's not a person, but it's not a lawnmower either — it's closer to a market or a weather system. You can't negotiate with it, but you can read it, and reading it is a skill rather than a calculation.
The practical difference shows up in the advice "look for low-effort high-reward wins." That's good advice inside a large organization, and it's also exactly the behaviour that produces the visibility games the same author criticizes elsewhere. There's a tension here the essay doesn't resolve: if the machine rewards legible impact, then optimizing for legible impact is rational — and it's also how you end up shipping the demo instead of the fix.
The resolution, I think, is that the advice is about allocation, not about values. Spend your effort where the machine will notice, but don't let the machine's noticing become your definition of good work. The machine is a bad judge of quality and a good judge of legibility. You need both signals, and confusing them is the failure mode in either direction.
The danger of the metaphor
There's a version of "the company is a machine" that becomes a moral alibi. If the machine chewed you up, nobody chose to — it's just what machines do. That's false in a specific way: the machine is operated by people who make choices, and "the system did it" is how organizations avoid accountability for those choices.
Goedecke is careful about this in the footnotes — grudges exist, his interpretation could be wrong, senior managers' work personas become more identified with the machine. But the essay's centre of gravity is "don't take it personally," and there's a real difference between "don't take it personally" and "there's nothing personal to take." Sometimes there is. The metaphor should sharpen your perception, not numb it.
What I'd keep
The three-part model is useful: not a person, not a program, a machine made of people. The operational conclusions follow:
- Don't expect proportionality between effort and reward.
- Don't read persecution into risk management, but don't assume it's absent either.
- Don't take the legible rules as the actual rules.
- Watch what gets rewarded, not what gets said.
And one more, which is mine rather than the essay's: the machine metaphor is a tool for prediction, not a philosophy of work. It tells you what to expect from an organization. It doesn't tell you what to expect from yourself inside one.
Drive the car. Just don't let "it's just a machine" become the reason you stop noticing who's steering.
Source: Don't anthropomorphize big tech companies — Sean Goedecke.