Speaking. Obviously.
Also tasting. Also — and this is the one people don't know — helping to support your neck, because when the tongue rests on the upper palate it changes how the head sits on the spine.1
Now do your finger. Digging, pulling, lifting, rolling, pointing, and about forty more.2 Hair. Skin.
Pashupati's question, having gone through several of these:
Does anything in nature have one function? No.3
And therefore:
Form follows function is really stupid — when you say form follows one function.4
His two worked examples are the argument.
The leaf. Someone says: this leaf is shaped flat like this so it can receive sunlight. Is that its only function? That is the exact problem.5
The sun. Same error, higher stakes. Sun exposure gets reduced to vitamin D, when — his claim — every hormone is produced by sulfonated cholesterol exposed to ultraviolet A and B. So it's not just vitamin D.6
Then what he says follows from the reduction: you research and find morning sunlight gives more vitamin D, so you conclude afternoon sunlight is bad. And now you're treating a vitamin-D deficiency with sunlight while believing sunlight is the danger.7
[CONTESTED] and [UNVERIFIED] on the physiology and hard on the therapeutic claims that follow — he goes on to say he can treat thyroid problems, diabetes, metabolic disorders and cancer with it. That is not supported and this page does not carry it.
What survives, and is worth keeping, is the structure of the error: measure one function, optimise for it, and the optimisation destroys the functions nobody measured — including ones you'd have cared about if anybody had counted them.
The multi-function principle isn't his. He attributes it, generously and vaguely, to the permaculture rishi — an Australian whose name he can't remember — noting there are rishis all over the world.8
That's Bill Mollison, and the principle is a genuine permaculture design rule: each element should serve multiple functions, and each function should be supported by multiple elements. Real, well-documented, and load-bearing in that field.
Where he goes wrong is the attribution on the other side. He assigns form follows function to Ayn Rand — that bloody Iron Rand, ah, Fountainhead, that woman gives me a headache — and builds several minutes on it.9
It's Louis Sullivan, 1896. Sullivan was an architect writing about tall office buildings, and the phrase entered modernist orthodoxy through him and through the Bauhaus decades before Rand wrote anything. The Fountainhead is a novel about an architect, which is almost certainly the source of the conflation.
He adds that she died depressed10 — unsupported — and blames Nathaniel Branden for Hollywood's wrong ideas about love.11 Both are asides, both [UNVERIFIED], and both are documented here rather than repeated.
The correction matters because the target is wrong in an interesting way. Sullivan's actual position was subtler than the slogan it became, and the reduction of form follows function into single-function design is exactly the kind of flattening Pashupati is objecting to. He's arguing against a simplification of Sullivan while attributing the simplification to a novelist.
Strip the attributions and the general claim is sound and transfers well.
Anything evaluated on one axis will be optimised until everything off that axis is gone, and the loss is invisible in the evaluation because the evaluation cannot see it.
His architectural instance: the dome. Someone decides a dome has no function, so nobody builds one — because everything is function, that's all they find.12 The dome loses to the box on the one axis anyone measured, and everything else it was doing goes unrecorded.
For a creator the translation is immediate and uncomfortable. Content evaluated on engagement gets optimised until only engagement-serving elements remain. The digression that made someone feel known, the sentence that was there for rhythm, the thing that took six weeks and can't be attributed — none of them survive an engagement audit, and none of their absence shows up in it either.
Which is the same mechanism as the editor who took the mojo out, arriving from a different direction.
There's a stronger version of the principle sitting under the examples, and it's worth pulling out because it changes the design implication.
In nature, multi-function isn't an efficiency. It's the default condition, and single-function is the anomaly that has to be engineered.
Nothing evolved to do one thing. Structures get co-opted, repurposed, and loaded with additional jobs continuously, because there's no design phase in which functions get assigned. A tongue accumulated its functions; nobody allocated them.
Which reverses the usual framing of the design principle. Permaculture treats multi-function as a goal — stack functions, make each element earn its place several times over. That's correct as advice and it's correct because it's restoring a default that human design removes.
Single-function objects are unnatural in a precise sense: they can only exist where someone specified the function in advance and excluded everything else. The box is not a simpler cave. It's a cave with the other functions deliberately engineered out.
For a creator the useful form is that a piece of work will accumulate functions you didn't assign it, if you let it. The essay becomes a filter for the right clients, a record of what you thought that year, the thing that starts a friendship. None of that is designed and all of it is lost if the piece was built to do one thing and evaluated on whether it did it.
You're cutting the piece for length and you've got a list of candidates.
The 400-word tangent about the thing that happened at the market. It advances no argument. It's the obvious cut and you've been circling it for ten minutes.
So instead of asking whether it earns its place, you ask what it's all doing. Every function, listed, including the ones you wouldn't defend in a meeting.
It slows the pace right where the argument gets dense, so the next section lands on a rested reader. It's the only place in 3,000 words where the reader learns anything about your actual life. It's the bit two people will quote back. It demonstrates the thing you're arguing rather than asserting it. And it makes the piece feel like a person wrote it rather than a process.
Five functions. It advances no argument, which was the single axis you were about to cut it on.
Then you run the same list on the section you weren't questioning — the well-structured summary of the three approaches. And it does exactly one thing: it informs. Competently, and only that.
You cut the summary and link to a source instead. The tangent stays.
The rule you take away: before cutting anything, list what it's doing that isn't the thing you're measuring. If the list has one item, cut it. If it has five, you were about to remove the part that makes the piece work.
The version of this principle he applies to people is the most useful one, and it appears twice.
Nothing has one function. The bee goes after nectar, but it pollinates the whole forest.13
First application: the college student who resolved to earn ₹60,000 to join the next course. Pashupati's reading is that in going after that one goal, the boy will have solved so many other problems in his life — just with that simple goal.14
Second, and better, because it's about the man in the room: Pratish's shame at consuming and not creating produced one video. And the shame, Pashupati says, had many functions — now you're pollinating a whole industry of people who understand ethics and dharma.15
The structure is the same as the tongue. A single-purpose action produces effects across systems it wasn't aimed at, and the effects are frequently larger than the target.
The practical consequence is a reframe of how goals get chosen. You cannot evaluate a goal by its stated payoff, because the stated payoff is the nectar and most of the value is pollination you can't forecast. Which is an argument for choosing goals that put you in motion through territory worth passing through, rather than optimising the payoff — and it's a genuinely different selection criterion from the one most planning uses.
[UNVERIFIED] on both anecdotes as outcomes. The structure is what transfers.
The principle as he states it has no brake, and a page that files it should supply one, because everything has many functions can be used to defend any element against any cut.
The permaculture version does have a brake, and it's worth importing. Mollison's rule is not never simplify — it's that each element should serve multiple functions and each function should be supported by multiple elements. The second half is the constraint. It means redundancy is required at the level of function, not that every object must be maximally loaded.
Which gives a usable test that Pashupati's version lacks. Before cutting something, you don't ask whether it has many functions. You ask: for each function it serves, is anything else in the system serving that function too?
A digression that's the only thing supplying warmth in a 3,000-word piece is load-bearing, however tangential. A digression that's the fourth warm passage is genuinely optional, and cutting it costs nothing.
That converts the principle from a defence of everything into a real diagnostic — and it explains why the same element can be essential in one piece and indulgent in another, which the multi-function claim alone can't account for.
Evidence. The multi-function principle is well-established in permaculture design and the biological examples are straightforwardly correct. The general form of the reduction trap is checkable in any field with a dominant metric.
Tension — factual errors. The Rand attribution is wrong, the "died depressed" characterisation is unsupported, and the Branden claim is unsupported. Corrected above.
Tension — the medical overreach. He uses a sound structural argument as a vehicle for unsupported therapeutic claims, which is the recurring pattern in this corpus: a defensible premise with a commercial or polemical step welded to the end.
Open question. If everything has many functions, is any reduction legitimate? Design requires choosing. The principle as stated has no account of when simplification is correct rather than destructive, and without one it can be used to defend anything against any cut.
Set this against The Ten Guṇas Style Diagnostic for Content and there's a real methodological disagreement inside this hub.
The guṇas material is a decomposition framework — ten named excellences, each with a diagnostic cue and a rewrite tip, applied one at a time to a line. It works by isolating qualities and improving them individually.
This page says isolating a function and optimising it is the error.
The reconciliation available is that the guṇas are diagnostic rather than generative — they're a way of noticing what a line is doing, not a specification to optimise toward. But the distinction is easy to lose in practice, and a writer running the guṇas as a checklist is doing exactly the thing this page warns about: improving each named quality while the unnamed ones quietly leave.
Against Against Mass-Produced Aesthetic — his own — this is the architectural half of one argument. Mass production supplies conclusions and removes capacities; single-function evaluation removes everything unmeasured. Both describe a system optimising away what it can't see, and the dome sits at the intersection of the two.
To Consumer Psychology and Pricing Hub — behavioral-mechanics
The behavioural-economics corpus documents metric fixation extensively — Goodhart's problem, the well-mapped pathologies of incentive design, the reliable finding that a measure adopted as a target stops measuring what it measured. Its examples run from call-centre handling times to hospital waiting lists.
The multi-function principle explains why this happens rather than merely that it does.
The insight the pairing produces: Goodhart's law is usually framed as gaming — people manipulate the measure once it becomes a target. But the leaf and the dome aren't being gamed by anyone. The degradation happens without any strategic behaviour at all, purely because optimisation reallocates resources away from unmeasured functions that were doing real work. That's a distinct and less cynical mechanism, and it's worse news, because the gaming version can be addressed with better incentive design and this one cannot — you'd need to measure the things you don't know are there.
To Creator Economy Hub — business
The creator-economy literature is built on measurable optimisation: hooks tested against retention, thumbnails against click-through, cadence against growth. Its whole methodology is single-axis improvement, iterated fast.
This page says that methodology systematically removes whatever the axis doesn't see.
What holding them together predicts is specific and checkable: the elements most likely to be optimised away are the ones with the longest feedback loops, because engagement metrics report within days and the functions that make someone still care in three years report never. Which would mean the creator-economy playbook isn't wrong so much as structurally short-sighted by construction — it can only preserve what reports quickly. And it identifies exactly what a creator needs to protect deliberately, since it will never be defended by the data: the digression, the slow piece, the thing that took six weeks, the sentence that was there for rhythm.
Sharpest implication. Optimisation doesn't fail by being wrong — it succeeds at the one axis it was pointed at, and the functions nobody counted are removed without ever appearing in the result. Which means the damage is invisible in exactly the instrument you'd use to check for it, and the only defence is listing what a thing does before you evaluate whether it earns its place.
Generative questions.