Infinite Midwit

Author: Adam Mastroianni (Experimental History) · Published: March 31, 2026 · URL: https://www.experimental-history.com/p/infinite-midwit

One-line: AI has an enormous amount of objective intelligence and essentially no subjective judgment — it can tell you what is true, but not what is worth doing — so it converges on the competent, confident middle of everything.

Summary

Mastroianni’s opening is counterintuitive: “The better AI has gotten, the less anxious I’ve become.” A few years ago, as someone whose job is producing words on the internet, he assumed he’d soon be filling his pockets with stones. Having looked closer at “our electric god as it has slouched toward San Francisco to be born,” he finds something else — “an infinite midwit: a stooge who is always available and very knowledgeable, but smart? Well, yes and no, in weird ways.” The models learned to count the r’s in “strawberry” and stopped recommending glue on pizza, and yet “there’s still a hole in the center of its capabilities that’s as big as it was in 2022… I only know this because that hole is where I live.”

The two intelligences. Some problems have clear boundaries and verifiable solutions (“What’s the cube root of 38,126?”) and require objective intelligence. Others are vague and squishy, and it isn’t clear whether you’ve solved them or whether they exist (“How do I live a good life?”), and require subjective intelligence. “Objective intelligence can be trained, reinforced, and validated. Subjective intelligence cannot.” Using one word for both is, he notes, itself a case of the objective crowding out the subjective: a century of psychology “proved” there is only one intelligence because every test correlates with every other, producing the g-factor. But any test of intelligence is only a test of the objective kind — “‘How do I live a good life?’ is not a multiple-choice question” — so rediscovering g “is like being surprised that you find the same patch of sidewalk every time you look under the same streetlight.”

AI is pure objective intelligence, which is “why each new model comes with a report card instead of a birth certificate.” The promise of superintelligence assumes objective intelligence is the only intelligence, or that the kinds are fungible — that we are playing “under Settlers of Catan rules, where if you have enough of any one resource, you can trade it for any other.”

Writing as the test case. Writing needs both. LLMs ace the objective parts — grammar, semantics, syntax are unimpeachable — but “good writing requires an additional bit of juju that makes the prose live and breathe, a light on the inside that can’t be quantified or checklisted,” and that light has never come on. He notes the striking consensus among very different writers (Jasmine Sun in The Atlantic, Erik Hoel on his Substack, Sam Kriss in the NYT) that LLMs write badly. His own complaint is not hallucination: “everything it writes vaguely sucks. I drag my eyes across the words and I feel nothing.”

His account of why is the best passage in the piece. Words don’t contain feeling; they are “a recipe for creating that feeling inside your own head, to assemble the right set of emotions out of the experiences you have at hand,” and the resulting experience in a reader is never identical to the writer’s because the ingredients differ. “The computer doesn’t know any of this. It can’t know any of this. It can only read the cookbook; it can’t taste the meal. Objective knowledge can make your sentences true, but it can’t make them alive.” And crucially, this wall can’t be scaled: “it’s a wall with a width you cannot describe and a height you cannot see” — you can’t minimize what you can’t measure.

He does check, periodically, whether the frontier models can out-write him. The result “sounds like a version of me that has sustained blunt force trauma to the back of the head and spent years recovering in a hospital where the Wi-Fi, for whatever reason, only lets you log onto LinkedIn”: metaphors that don’t congeal, phrases that sound insightful until you think about them, breathless insistence that every sentence is a revelation. He is candid that this should be no contest — he hasn’t read the internet, has no team of Stanford PhDs, and nobody has invested $2.5 trillion in him.

The nerd argument. The evidence that objective intelligence doesn’t convert into the other kinds is that we already ran the experiment on humans. Paul Graham argued twenty years ago that nerds are unpopular because they don’t want popularity; Mastroianni disagrees — the nerds he knew schemed constantly for status and simply “had the wrong kind of smarts.” If objective intelligence were sufficient, “Mensa should be the Illuminati, not a social club for people who know lots of digits of pi.” His exemplar is Scott Adams via Scott Alexander’s eulogy for Dilbert: a man of considerable intelligence who failed at the restaurant, the Dilberito and the books about religion, and succeeded only at drawing guys in ties. He quotes Alexander’s line about the Bay Area nerds who periodically decide to hack themselves into being charismatic, and “every yoga studio and therapist’s office in the Bay Area has a little shed in the back where they keep the skulls of the last ten thousand bright nerds who tried this.” A college friend who was brilliant and could not turn an essay in on time makes the same point: life is not a role-playing game where intelligence is experience points you reassign to any Big Five trait.

Hence the close. The remaining hope for scaling is quantity — maybe four units of objective intelligence don’t buy one of subjective, but four billion might. “The CEO of Anthropic promises us a ‘country of geniuses in a data center’. Maybe that will happen! Or maybe we will discover the data center actually contains a country full of Scott Adamses. At the very least, we can look forward to many more flavors of Dilberitos.”

Full text

Archived privately against link rot: ../attachments/infinite-midwit/infinite-midwit.md

Connections

• Adam Mastroianni — the author; Experimental History is his blog, and the objective/subjective split here is the same move he makes against psychometrics and peer review elsewhere.
• Taste — the “light on the inside” that he says can’t be quantified or checklisted, and therefore can’t be optimized toward.
• Judgment — the essay’s real claim is that judgment and intelligence are separable, and that a century of testing hid it by only ever measuring the half that is measurable.
• Expertise — the nerd argument, and the observation that “our world is not run by people who won their statewide spelling bee.”
• Artificial Intelligence — the capability claim, framed as a structural wall rather than a current limitation.
• Scott Alexander — quoted at length from his eulogy for Dilbert; the Scott Adams case is the load-bearing example.
• Anthropic — the “country of geniuses in a data center” promise the piece is answering.