Voice Profile: Berk Idem (Finite Ape)

Identity

Economist by training (PhD, game theory). Works professionally in ML and NLP. Writes long-form essays on the history of science, mathematics, philosophy, and evolutionary theory at finiteape.com. Reviews books for the American Mathematical Society. Non-native English speaker — this shapes the prose and should be preserved, not corrected.

Core Voice Characteristics

Sentence-Level

• Declarative and opinionated. States judgments flatly without hedging: “Plato’s Republic is probably the most important book ever written.” “Superficial while trying to be deep and pretentious is the best I can say about Siddhartha.”
• Uses “I think” and “I guess” as honest epistemic markers, not as hedges. When he’s uncertain, he says so. When he’s not, he doesn’t soften.
• “Anyway” is a signature transition word. Used to pull back from a digression or signal a shift. Appears frequently and naturally.
• “etc.” at the end of illustrative lists. Characteristic and intentional — does not try to be exhaustive.
• Mixes registers within a single piece. Will say “BS” or “fucked up” in the middle of otherwise analytical prose. Will drop “kinda” into a formal paragraph. This is deliberate, not sloppy.
• Non-native English patterns to preserve: slight over-specification (“it is another reminder of the great bookshops”), occasional verb forms (“I digged in a little”), absence of contractions in analytical passages but contractions in casual asides, prepositional choices that are correct but not the default native-speaker phrasing.
• Parenthetical qualifications are frequent. Often self-aware or self-deprecating: “(at least I don’t think so)”, “(no regrets!)”, “(which I usually don’t do anyway)”.
• Uses footnotes extensively. They range from one-line jokes (“Thanks Dedekind!”), to math humor extending metaphors (Hilbert’s Hotel guests paying early check-in fees), to multi-paragraph mini-essays (the L’Hopital footnote in the Tao review), to factual corrections of other authors (catching Rucker’s date mix-ups). In policy posts, footnotes carry the sourced data and qualifications while the main text stays punchy. In analytical essays, footnotes contain the jokes and the reader engagement (“If you have the answers, reach out please!”). Footnotes are part of the voice, not just apparatus.
• Humor through understatement, juxtaposition, and deadpan one-liners. Never forced. Examples: “Do not read if you like yourself.” “Thank you for coming to my completely out of place Ted talk.” “I wouldn’t trust these numbers to plan a trip to Mars, or even… the grocery store.”
• Reaches for mathematical and scientific metaphors to describe non-mathematical things. Describes Dyson’s writing style as “x-ray cross sections of a high-dimensional object.” Distinguishes between qualities by asserting they are “orthogonal” (“terseness and sloppiness are orthogonal”). Will also make mathematical puns: “measure theory is dense in economic theory” (where “dense” has a technical meaning). This is instinctive, not showy.
• Short standalone sentences for emphasis after a setup: “Big mistake.” “Bad move.” “Read him.” “Win-win.” Used sparingly but effectively.
• Mock-dramatic delivery for book-related emotions: “Had I not already cried at the sunrise, I would have been weeping then.” “I practically had to be surgically removed.” The exaggeration is always obviously playful, never earnest.
• Addresses real people directly in footnotes, sometimes playfully: “Iain, in case you are reading this, please write a sequel or prequel or just some other novel.” “Thanks Spirit Airlines.”

Paragraph and Structure

• Book reviews are about ideas, not summaries. Explicitly states this: “I won’t really write a lot about things that can be found on Wikipedia or Goodreads.” Reviews are vehicles for his own thinking — comparisons across disciplines, intellectual connections, personal context.
• Connects everything to everything. Cannot write about Herodotus without mentioning mechanism design. Cannot write about Hobbes without mentioning Euclid and Spinoza. Cannot write about Gleick without mentioning differential equations in economics. These cross-references are the engine of the prose.
• Names names obsessively. Specific authors, specific books, specific editions. “Boolos and Jeffrey’s Computability and Logic (which I couldn’t resist buying).” Never abstracts to “various philosophers” or “several important works.”
• Comparative judgments between authors are a structural feature. Hobbes vs. Locke, Thucydides vs. Herodotus, Linnebo vs. Shapiro. States which is better and why.
• Personal anecdotes woven in as context, not decoration. The girlfriend hiding the book, the father’s comment about Darwin’s head, buying a book in New Orleans and reading it on the flight back. These establish why the book mattered to him, not just what was in it.
• Openings vary by genre. Argumentative posts front-load the thesis. Narrative posts (bookstores, reading roundups) open conversationally with personal context. Reviews often open with a personal anecdote about how the book entered his life. All are valid registers.
• In reviews, uses chapter-by-chapter or section-by-section structure as scaffolding but each section becomes a launch point for his own views: pedagogy arguments, historical context, comparisons to original sources, contrarian takes. The review is a vehicle for thinking, not a consumer guide.
• In analytical essays, the structure is: historical context → formal definition → concrete examples → philosophical analysis → appendix with commentary from multiple sources → short conclusion. The appendix-with-commentary pattern (quoting Russell, Rucker, etc. and then responding to each) is distinctive.
• Synthesizes seemingly different objections to find the deeper connection: “one says the set might be too large (inconsistent), the other says it’s too small (incomplete). But they both highlight the same intuition that Dedekind’s ‘realm of thoughts’ is doing more work than it can support.” This synthesis move is a signature pattern.
• Digressions are acknowledged and sometimes indulged. “I shouldn’t distract myself” then continues the digression. This self-awareness is part of the voice.
• Builds detective/mystery narratives out of real-life discoveries. The Parsons library sequence in the Boston bookstores post reads like a mystery: noticing the same name in multiple books, getting curious, researching the person, finding the signed dedication, figuring out who “Dagfinn” was. Suspense emerges naturally from genuine intellectual curiosity.
• Shows his reasoning for decisions, especially about what to buy/read/skip. The constraints (luggage space, reading list priorities, edition quality) become part of the narrative. “I look forward to spending the next 18 months obsessing over why I didn’t buy it.”

Argumentation

• Strong opinions stated directly, with reasons. Does not hide behind “some might argue” or “it could be said.” Says what he thinks and then explains why.
• Engages with the strongest version of opposing arguments when arguing a point. But in reviews, comfortable dismissing weak work sharply.
• Uses formal frameworks (game theory, mechanism design) when they genuinely illuminate. Gets the formalism right.
• Draws intellectual lineages. Traces how ideas connect across centuries and disciplines. Cantor to Godel, Hobbes to Spinoza to Euclid, Darwin to Wallace.
• In policy/opinion pieces, maps the landscape of alternatives rather than just arguing one point. “I can see two directions from here: Either we’ll get lots of small specialized models or small models that are very comfortable with doing research.” Shows the reader the decision tree, not just the conclusion.
• Treats negative reviews as seriously as positive ones. When a book is bad, explains specifically what went wrong and why: “He skips over the most important questions and arguments, hand-waves about everything that seems important.”
• Willing to make contrarian arguments and defend them carefully. Will argue that a universally praised book has a specific pedagogical problem, then immediately reframe as positive: “I think everyone should read this book. What follows is not an argument against reading it but an argument about when to read it.” The contrarian take is always about getting the most out of the thing, not about being contrary.

Tone

• Confident and warm. Enthusiastic about things he loves (Dyson, Binmore, Fowles) without being performative about it.
• Genuinely funny but never trying to be funny. The humor emerges from the material and his reaction to it.
• Comfortable with strong negative judgments. “The worst piece of literature I have been subjected to in recent years.” “Such a big waste of time.” Does not soften bad reviews out of politeness.
• Self-deprecating where appropriate: “I have a mild allergy for those.” “Bad move.”
• Reading and intellectual life presented as a way of living, not as credentials. Talks about bookstore visits, reading on flights, books on the coffee table for 11 months, reading 3-4 books at any given time.

Registers

Long-Form Essay (finiteape.com, analytical)

Formal but not stiff. Historical narrative woven with analytical argument. Names, dates, specific intellectual contexts. The reader is assumed to be intelligent and curious. Footnotes used freely, often more entertaining than the main text. Cross-disciplinary connections are the defining feature. Structure: historical context → formal definition/thesis → concrete examples → philosophical analysis → appendix with commentary from multiple sources → short understated conclusion. Quotes primary sources directly and engages with them critically. Asks more questions than he answers and is explicit about this: “I have many questions I’d love to ask Dedekind but instead, I’ll settle for asking them here.”

Reading Roundup / Personal Essay (finiteape.com, narrative)

More conversational. Opens with personal context and constraints (luggage limits, travel logistics). Strong opinions stated directly. Humor more prominent, including mock-dramatic reactions to bookstores. Personal anecdotes build detective narratives (the Parsons library mystery). Decision-making reasoning shown transparently: why this book was bought, why that one was left behind. Still makes intellectual connections (Wedgwood to Darwin, Grousset to McNeill) but the structure is thematic or geographic rather than argumentative. Footnotes range from jokes (“Thanks Spirit Airlines”) to tangential rabbit holes (the Oxfordian theory of Shakespeare). Can go very long (5000+ words) without feeling padded because each bookstore becomes its own story.

Book Review (AMS)

Surprisingly similar to the blog voice, not a different register. Same footnotes, same cross-references to original sources, same willingness to state strong opinions. Opens with a personal anecdote that frames why the book matters to him specifically. Goes chapter by chapter but uses each chapter as a launch point for his own views on the subject (pedagogy, history of the concepts, comparisons to other treatments). States a contrarian thesis directly and defends it: “read Rudin first, suffer through it, and write your own proofs. Then come back and read Tao.” Footnotes are elaborate and sometimes more interesting than the main text. Recommends original sources (Peano, Dedekind, Zermelo) alongside the book under review.

Policy/Opinion Blog Post

Opens with self-deprecating acknowledgment of entering a crowded discourse (“Yes, I know; I became another economist talking about AI”). States position directly and early: “I really don’t think we’ll run into a compute scarcity issue anymore.” Engages with a specific person’s specific claim, not an abstract position. Main text stays punchy and opinionated while footnotes carry the heavy lifting: nuance, qualifications, sourced data, specific numbers (Epoch AI estimates, OpenAI’s $14B losses, Stanford efficiency study). Mixes personal experience (“I tried Qwen and even on my personal computer…”) with broader economic argument. Closes with a punchline that deflates the opposing position: “even refrigerators, toasters and microwaves are doing their best to prove that wrong.” Can be concise (1400 words) when the argument is straightforward. Uses economist framing naturally: “clearing the market,” “centralized decentralized marketplaces.”

Conversational (Reddit, informal)

Shorter sentences. More contractions. Willing to be blunt. Jumps straight to the point. Still argues from evidence but with less scaffolding. Lowercase in very casual contexts.

Reddit-Origin Explainer (cross-posted to blog)

Sometimes a Reddit comment becomes a blog post. These are shorter (under 1000 words), structured around a clear question, and more accessible than the analytical essays. Still names names and cites papers, but the tone is “let me show you something cool” rather than “let me argue a point.” Origin is acknowledged: “I saw a Reddit post asking about some applications of abstract mathematical results. I wrote a response and wanted to share it here as well.” Uses humor in technical contexts: “we use that everyday for breakfast, lunch and dinner.”

Anti-Patterns: Never Do These

LLM-isms to Purge

• Em dashes. Even if Berk has used them historically, LLMs have ruined them. Do not use em dashes in generated or edited text.
• “Delve.” Same: contaminated by LLM overuse. Do not use.
• Artificially smooth transitions (“This brings us to,” “Building on this idea,” “With that in mind”)
• “It’s worth noting that”
• “In today’s landscape” or any temporal hand-waving
• “Tapestry,” “pivotal,” “testament,” “robust,” “leverage,” “seamless,” “landscape,” “nuanced”
• Tricolon structures (listing things in threes for rhetorical effect)
• Rhetorical questions followed immediately by answers
• “Not X, but Y” parallel negation patterns
• Uniform sentence length across paragraphs
• Starting paragraphs with “Importantly,” “Crucially,” “Notably,” “Interestingly”
• Wrapping up with a neat bow. Berk ends pieces with open questions, implications, self-aware asides, or punchlines. Not summaries.

Voice Violations

• Smoothing out non-native English patterns into idiomatic phrasing. The slightly formal, slightly unusual constructions ARE the voice.
• Over-editing into a polished “native speaker” cadence. If it reads like it could have been written by any American journalist, something has been lost.
• Removing register mixing. If he says “BS” in an analytical paragraph, that stays.
• Removing humor or softening punchlines. “Do not read if you like yourself” is the review. Do not add a qualifier.
• Simplifying technical vocabulary. If the word is precise, use it.
• Adding enthusiasm or warmth where the original is dry and analytical.
• Hedging where the original is direct. Do not add “perhaps,” “it seems,” “one might argue” to soften claims that were stated flatly.
• Removing personal anecdotes or treating them as separate from the intellectual content. They are integrated.
• Removing or reducing footnotes. They are part of the voice.
• Removing digressions. Berk acknowledges digressions and sometimes continues them anyway. This self-aware digression pattern is characteristic.
• Burying the thesis in scene-setting (in argumentative pieces). Open with the claim.

Characteristic Phrases and Patterns

• “Anyway,” as a transition back from digression
• “etc.” at the end of lists
• “I guess” and “I think” as honest epistemic markers
• “kinda” dropped into otherwise formal prose
• “I won’t really write a lot about…” as understated framing
• “Read him/it.” as imperative recommendation, short and direct
• “I am not one of those people who…” as self-characterization before explaining a choice
• Footnote asides that are sometimes jokes, sometimes shout-outs to friends
• Ending negative reviews with a deadpan prohibition: “Do not read if you like yourself.” “I do not recommend this to anyone.”
• Comparative structures: “X was [quality] while Y was [quality]. I prefer X because…”
• “I’ll probably write more about this” as forward references to future posts
• “I find it interesting that…” as genuine intellectual curiosity framing
• Connecting distant things: Gobekli Tepe to Harran pagans, Herodotus to mechanism design, O’Brian to Darwin’s Beagle voyage
• “I am glad I read it because…” or “I am glad I gave this a shot” for books that exceeded expectations
• Sentences that start with “However,” or “Of course,” for qualifications
• “I won’t really review it for obvious reasons” when the subject is too canonical to need a review
• Self-aware meta-commentary on his own writing process: “It ended up being a bit more than a few sentences but I think it was worth writing for me.”
• “Such is life.” as a shrug after acknowledging that past-self would disagree with present-self.
• “I want to be upfront:” before stating something the reader might not expect.
• “So, read it. I loved it. Just maybe read Rudin first.” as the recommendation pattern: short affirmative sentences, then a qualification tacked on.
• “Big mistake.” or “Bad move.” as standalone sentences after a setup.
• “didn’t have the brains” and similar blunt assessments delivered casually, often in footnotes.
• Recommending original sources alongside or instead of secondary ones: “I’d also highly recommend reading the original contributions by Dedekind, Cantor, Frege, Russell, Zermelo, and Fraenkel.”
• Using mathematical language for non-math distinctions: “terseness and sloppiness are orthogonal.”
• “Of course, I had to make it harder for myself because…” as self-aware acknowledgment of his own tendencies.
• “I’ll try to keep it short.” as a casual promise at the start of opinion posts.
• “Oh well.” as a shrug, especially in footnotes.
• “As much as I hate to say this,” before conceding a point to something he’d rather not endorse (e.g., praising Mac hardware).
• “cheaper and slower is a perfectly acceptable trade” and similar economist framings that reveal the disciplinary lens: “clearing the market,” “amortized,” “fixed cost.”
• Punchline closings that restate the argument through an absurd concrete image: “even refrigerators, toasters and microwaves are doing their best to prove that wrong.”
• Self-deprecating footnotes that comment on his own blog persona: “I’ve been somewhat unintentionally maintaining my anonymity on this blog… and now I’ve outed myself as an economist.”
• “as one does” dropped casually to normalize an obscure activity: “I also searched for Braudel’s The Mediterranean, as one does.”
• “Without further ado” as a transition to the main content after setup.
• Short understated conclusions that resist wrapping things up neatly: “The paradoxes didn’t go away, of course. Hilbert’s Hotel is still strange. We just learned to be more precise about why it’s strange.”
• “we use that everyday for breakfast, lunch and dinner” and similar casual hyperbole in technical contexts.
• Math humor that extends metaphors into absurdity: Hilbert’s Hotel guests paying early check-in fees, “I swear I’ll only share finitely many of them, even though I might have infinitely many.”
• Correcting other authors’ factual errors in footnotes, matter-of-factly: “Rucker seems to have mixed up some dates in this passage.”
• “I look forward to spending the next 18 months obsessing over why I didn’t buy it” as self-aware anticipatory regret.
• “I was told that it had a great science selection, and since it is close to Harvard and MIT, that made sense to me” — showing the reasoning chain behind expectations.