Proof of Hit
When a built-in one-way hash underwrites taste, creators mine hits instead of inventing them
Opinions of Saturday Night Live fall along a narrow range. Either it’s always been trash or it slowly turned into trash, having been only slightly better in the past. We talk about it like we do airline food, offering up a ritual complaint to display we have taste. Still, it sits near the top of the ladder as a comedy writing credit and is the last stop for improv actors. Odd perches for a show we swear is awful.
How did a supposedly unwatchable show produce decades of lines we quote at every opportunity? It doesn’t matter whether a coworker is making copies or a string of bad choices land us in a van down by the river, it’s applicable everywhere. It has minted careers so reliably that even those who sneer at the institution accept the credential as a passport. A thing can’t be culturally parasitic for fifty years without giving us something in return. Our default interpretation is missing a variable.
I dipped into the archives. There are hidden gems, and I watch the show infrequently, so others like me should have crowdsourced a meritorious view count for the best sketches. A heuristic emerged that strongly implied my time wouldn’t be wasted. An x-year-old sketch should have more than x million views. This proved a reliable way to watch as much of the good stuff as possible with the fewest false negatives.
Over a few months, I made a playlist of my favorite lesser-known sketches, some from other sources like Mad TV. In the end, about fifty clips for the fifty years SNL has run. A million writer-hours of continuous effort gave us a few hours of humor, spread across half as many characters. Even among those, some of my favorites like Debbie Downer, Suel Forrester, Laura Parsons, and Nicholas Fehn grow tiring after the third sketch. Were the haters right? Is it really that bleak?
It’s easy to treat the meager yield as evidence of incompetence. If it is, our disappointment becomes moral. We imagine a counterfactual world where better decisions could’ve produced excellence routinely. If comedy is a factory process, then SNL becomes a facility that lost its ISO 9001 certification. Send in the consultants. If the output is mostly forgettable, someone must be guilty.
What if the opposite is true? They are not failures at comedy, but masters working in a fiendishly difficult field few dare touch. Comedy isn’t like chess, where skill accumulates and the solution space is bounded; practice and learn strategy and you’ll be measurably better each day. It’s more like hunting. You can do everything right and come home emptyhanded because where prey appears is not a function of your hunting skill.
Comedy is one of the few arts where nearly everyone shares an enormous training set. We spend our entire lives steeped in jokes, voices, timing. The medium isn’t purely the domain of an elite class the way architecture or wine or painting try to be. A room of eighth graders can sort sketches into winners and losers, even if they can’t explain why. Our verifier for comedy appears to be installed at a very low level and, while socially influenced, it is largely a black box.
That universal verifier creates a brutal imbalance. Generating a hit is expensive, while recognizing one is nearly free. You can spend weeks writing and rehearsing a sketch only to have the audience decide in ten seconds whether it’s a hit. Spend years developing a character only to have strangers decide instantly whether it will return. Consensus arrives quickly when the verifier is shared among everyone.
A good joke behaves like a one‑way function. Verification is instant, sometimes before you consciously realize, but generation is hard, and it’s strangely difficult to reverse from output back into recipe. You can analyze why something worked, name its moves, even imitate its surface, but none of that guarantees you can produce the next one on command. Laughter tells you an answer is valid, but it doesn’t tell you how to find another answer.
Melody operates under similar rules. Most humans have been marinating in music since infancy, and the verifier is as universal as laughter. The disagreement performed as “taste” is often about tribe and identity, and involves applying those as penalties to entire genres, not about whether one song is better constructed than another. You may love rap and hate country, but if forced to sit through two country songs, I bet you can rank their relative quality, and I bet that ranking matches that of a country lover or critic.
When underlying differences shrink, signaling taste must intensify to create a measurable gap. In artistic endeavors, this is good. Quality things should beat out junk and tastemakers deserve reward for identifying them. It goes awry when differences in rival products are miniscule or the verifier is so universal and instinctual that taste adds no information. For an example of both, ask how many lifetime ad dollars are spent convincing you personally that two recipes of brown, nutmeg-flavored sugarwater are meaningfully different.
Comedy and melody behave a lot like proof-of-work in another newly-invented store of value, Bitcoin. They may be humans’ original cryptocurrency. Finding blocks is mostly brute force, yet all of us share the hash function to trivially verify what is good, and whoever finds a hit is handsomely rewarded. The SNL writer’s room is our largest mining rig—one running at a weekly hashrate. We don’t see the failed guesses in Bitcoin the way we do live in sketch comedy.
Are these characters invented or discovered? The invention frame fantasizes that you can accumulate funny attributes and expect the process to be additive. Discovery is more like stumbling into a constraint set that makes a character self-propelling across situations—a stable engine rather than a single-use premise. A disciplined or skilled creation process doesn’t guarantee output. It is only the hashrate. It increases the number of guesses you can make, the speed at which you can iterate, and the precision with which you can recognize when you’re near a seam.
Constraints and how they limit characters are why I argue comic characters are discovered, not invented. Give a character a voice and you’ve forced a posture. That posture implies a tempo. Tempo limits the scenes they can power without snapping. Each choice kills dozens of plausible ones, and the space narrows rapidly. That narrowing is the opposite an invention-shaped craft, where each move opens up more possibilities rather than drastically reducing future moves. A working character is less an open canvas than a discovered region in a sparse, high-dimensional space.
We search the latent space of all possible characters. Each one is a high-dimensional coordinate with values for attitude, shame, style, status hunger, delusion, social blindness, and a dozen other attributes. Most points in that space, if they are funny at all, are unstable equilibria. Under recursion, they either run away or collapse. By the third sketch, if you can’t pin them down, they’ve likely gone off the rails. A coherent character generates conflict without needing to recycle the premise, stakes, or setting. An ideal character is a strange attractor. It orbits a narrow region in an unpredictable manner. No matter how well you understand it or how many loops you’ve seen before, you can’t predict its next move from its last.
A workable recipe for discovering these points looks simple. Take an archetype, add a specific glitch, and drop the character into a setting that triggers the glitch and watch them refuse to adapt. The best engines create a self-reinforcing loop where the attempt to restore status, dignity, or control only heightens the antics. It is a stable feedback system that doesn’t run away into pure chaos and doesn’t collapse into one note. If it works, you’ll crave seeing that character in other settings, not to repeat their catchphrase but discover their limits, if they have any.
Michael Scott’s cringe comes from status hunger fused with a need to be loved—a man who mistakes attention for intimacy and sincerity for permission. Kramer is childlike sincerity with no concept of shame. He controls the scene by forcing others to obey the social rules he ignores. Borat works because the character is an ideological wedge who makes others expose the limits of their cultural relativism. You can place any in a dozen scenarios and still want to see what they’ll do next.
The reward for finding one of these attractors is as lavish as the space is sparse. Look at Jim Carrey. He arguably took one character, Fire Marshal Bill, and created an entire career around it. Mr. Bean went worldwide rarely speaking a word, undermining theories claiming comedy is too cultural to be universal. Unique comedic characters we want to see more than thrice are so rare and valuable that the market values their discovery at a floor of a hundred million dollars, lifelong recognition, and as many movies and TV series as they can power before collapse.
You can watch Paul Reubens refine Pee‑wee Herman in his documentary. It is not an additive process like oil on canvas or clay where each move expands outward into new territory. He is a sculptor like Michelangelo “chiseling away the superfluous material” by testing posture, voice, sweetness, menace, innocence. Early forms feel discordant, as if the constraints haven’t interlocked yet, and then suddenly the character finds the attractor it had always been orbiting. The final result seems inevitable—how all discoveries look in hindsight.
Most recurring sketch characters die because the premise does all the work. The character is a wrapper used to deliver a trick, and once we know it, its sequel becomes predictable. Whatever humor’s evolutionary function—perhaps letting the clever upend power by forcing an audience to emit involuntary signals of coordinated agreement—its formula is simple. A frame is introduced and then a twist forces the frame to flip. The reanalysis in the new frame causes the laugh. Comedy is the equivalent of reading a garden-path sentence only in situational form. Once we can predict the reframe, it does nothing, and no laughter comes.
Under that pressure, long-running characters get Goodharted into simulacra of themselves, a process dubbed Flanderization. One reliable way SNL dodges that trap is by letting the structure create the comedy. Game-show sketches reliably work because rules create conflict even among thin or archetypal characters. You don’t need a nuanced character when the format itself forces people to reveal how they handle pressure and humiliation. The comedy comes from incompatibility under constraint instead of a chaotic attractor.
Once you accept a domain’s function is one-way, the virtues to admire become iteration, sensitivity to feedback, and ability to grind. Iteration is the only way to succeed in a sparse field. If you want the gold, don’t curse the nearly-empty mine or those toiling in it. Prefer to stay out of the mine? Separate your consumption from the search. Let institutions that can afford brute force do their work, and then take the verified survivors. Your attention is scarce currency too. Spend it compounding the value of the best blocks.