Systems thinkers, who often work at larger organizations than startups and VC funds, love to argue that “the purpose of the system is what it does” (POSIWID). Scott Alexander recently took issue with the idea on his Substack, Astral Codex Ten.
Now, I’m interested in what Scott Alexander thinks not because he seems to have had success here on Substack or in startups, but because he wrote the funniest serial blog I’ve ever read: UNSONG (PDF version). I like his writing, and found this critique of POSIWID particularly thought-provoking.
Judging by the 608 comments as of this post’s publication and Alexander’s creation of a comments section highlights post which has nearly 300 comments, it seems that I was not the only one.
The Origins of POSIWID
POSIWID started as an axiom of sorts among cyberneticians after Stafford Beer coined the term.
The original point of it seems to have been to convey the opinion that the purpose of a system is directly upstream of its effects — as opposed to something that could be teased out by looking at the intent of the designer, or operator, or advocate.
In this sense, POSIWID functions similarly to the duck test (“If it looks like a duck, swims like a duck, and quacks like a duck, then it probably is a duck.”) — but more for people evaluating systems than biologists evaluating birds.

POSIWID is an admittedly cynical approach. However, it has a key advantage: it’s easier to discuss and debate about observed effects than somebody’s unobservable intent.
This gets at Alexander’s chief critique of the concept — POSIWID is nearly always invoked to show organizational failures, never systemic successes. Here are the prototypical institutions he considers:
Cancer hospitals
New York City MTA Regional Bus Operations
The first thing I notice is that most of these are government institutions.
The second thing I notice is that all of these are organizations with more layers than an onion. There are teams of teams of teams within each of them.
His examples are all systems of systems.
Alexander’s critique is right in the sense that what these institutions aren’t doing what they claim is their purpose, but that shouldn’t actually be a surprise because these institutions are not in fact the baseline system that I think POSIWID refers to.
POSIWID seems far more accurate when it’s restricted in scope to systems with discrete components as the next level down. Here are some examples of systems that I consider more appropriate than Alexander’s:
The purpose of the Golden State Warriors is to win NBA games.
The purpose of a space probe’s communications system is to facilitate two-way communication between computers on the spacecraft and the ground.
The purpose of X (the app/website, as distinct from the firm) is to be a time suck by convincing viewers to look at content.
The purpose of a startup that’s raised a pre-seed financing round is to validate problem-solution fit.
That last example is worth going into more detail on.
Startups are a great example of POSIWID
Startups, particularly early-stage startups, are a great example of POSIWID.
These young companies are often hierarchically flat institutions, in that management and leadership are often additional tasks undertaken by founders on top of their sales or development roles. Not having layers upon layers of people makes it easier to assess what each individual at the firm is working on.
It also creates the culture of an institution that is a single team.
Understanding the singular team’s collective focus is important because successful startups seem, from my position as an outsider, to be very unified in their immediate purpose. The team is all working together towards a common objective.
If every single member of the institution is sprinting towards the same goal, and the whole of the institution is in some way related to the sum of its parts, it seems clear that the purpose of the institution is making progress towards that goal.
I see this more consistently in startups’ effort allocation than anywhere else. However, this can be observed elsewhere in startups too, and that’s what makes it interesting.
Example: Build or Buy?
An issue early-stage founders often seem to find themselves grappling with is how to access key components in their architecture, whether they should build or buy.
One perspective is that a startup should buy everything possible initially so it can start quickly, and re-engineer the product to vertically integrate it later.
Another perspective is that the startup should build as much as is practical in-house, so it owns as much of the value chain as possible. Blake Scholl, the founder of Boom, endorsed this approach on X earlier this year:
Decision to do our own engines was probably the single most important one in Boom history. Company most likely would have died waiting for the big guys to get stuff done.
Also being in control of powertrain means we can make it do what we want. Including something pretty cool…
I don’t want to be prescriptive on this issue. Founders should do what makes the most sense for their firm.
Example: Investor Evaluation
POSIWID is also an important consideration from my perspective as a prospective investor.
Consider the hypothetical case of a startup where the founders say the problem they are working on is A, and they spend all their time working on B, which is a related problem.
As an investor, I would evaluate the company as a startup working on B in all the things that matter — market size and dynamics, product maturity, technical plan, etc.
I would also have thoughts about the way the startup communicates.
In this sense too, the Purpose Of the Startup Is What It Does.
If you invest in startups, like my writing, and want to chat about possible career opportunities, don’t hesitate to be in touch. The best way to get ahold of me is via email: