◆/romeoStevens76_analysis
path romeoStevens76_analysis.md

Romeo Stevens (@RomeoStevens76) — A Personae Analysis

Romeo Stevens is the co-founder of the Qualia Research Institute (with Andrés Gómez Emilsson and Mike Johnson, Dec 2018) and the founder of MealSquares, writing publicly since 2008 at Neurotic Gradient Descent (neuroticgradientdescent.blogspot.com) and now from Boulder, CO, where he hosts a weekly Sunday Dhamma Chat on X Spaces with @BuddhistSanta and matches donations for a new spiritual center, @vividvoid's Nameless Mountain. He is a long-running LessWrong contributor whose center of gravity has rotated, over a decade, from Bay Area rationalism into a serious lay-Buddhist teaching practice without ever abandoning the engineer's framing. The 100 most recent posts and 60 reply-conversations analyzed here run from late January through late April 2026.

I. Core Worldview & Mental Models

The base claim: practice as engineering

Romeo's bio is itself the thesis: "spiritual practice is working when it changes your sense of what's possible." Everything else falls out of this. Practice is not symbolic, not aesthetic, not consoling — it is a mechanism whose output you can measure by checking whether your option-space has actually expanded. The corollary, which he repeats in many forms, is that any practice tradition that does not produce that change in option-space has degenerated into ritual, performance, or — his frequent diagnosis — a "miserable-but-stable fixed point."

The most explicit statement of this engineering frame is his Apr 2 2026 tweet (150L / 10RT): "Part of the point of practice is making your mind a pleasant ally by deescalating the internal conflicts that your personality is currently based around as miserable-but-stable fixed points." Personality is rendered as a set of equilibria; practice is the perturbation that lets you find better ones.

The "3m → 5a" core

When he writes about Buddhism, he is unusually opinionated about which Buddhism. The recurring shorthand "3m → 5a" — three marks of existence (impermanence, suffering, not-self) applied to the five aggregates (form, feeling, perception, formations, consciousness) — is, he insists, the actual core mechanic. Everything else is later elaboration:

"Two translation notes about the heart sutra, without which the whole thing appears very mysterious: 1. Emptiness is Mahayana shorthand for the three marks. 2. Form is often used throughout Mahayana texts as shorthand for all five aggregates. 3m→5a remains the core logic." (Apr 14 2026, 67L)

In his replies he goes further: "emptiness as a translation arose out of the people who try to improve instructions via verbing key terms to make it clear they are processes and not things. A lot of mahayana seems like a mashup of several different attempts from diff contexts" (reply to @Twinola2, Apr 14 2026). This is the same posture he took on his blog years ago in (mis)Translating the Buddha — and explains why his pinned tweet is the @dkazand MetaGame podcast where he debates @zhukeepa against perennialism. If 3m → 5a is the actual mechanism, then "all religions point to the same thing" is not a deep insight but a sloppy abstraction collapse.

Mental models he reaches for

A non-exhaustive list, drawn directly from posts and replies in the window:

The unifying gesture: take a phenomenon that the surrounding culture treats as moral, mystical, or political, and reframe it as a mechanism with inputs, outputs, and exploitable failure modes.

Intellectual DNA

Named or quoted in the window: the Buddha (Sariputta and Guanyin specifically — see his Apr 15 reply to @OortCloudAtlas: "most people are inadvertantly eliding the original core logic and source (Sariputta consistently teaches 3m→5a). Guanyin teaches a deepening of the same point to him in the sutra."), Ken McLeod (Apr 11 reply: "the three poisons are the three feeling tones tanha-ized"), Carl Rogers (Feb 6, posting The Necessary and Sufficient Conditions of Therapeutic Personality Change), Robert Kegan, Frank Herbert (he engages a Dune quote on selfhood Feb 27), Connirae & Tamara Andreas (his blog post Core Transformation — an NLP method named in the MetaGame interview). The nameless background influence is the early Pali Canon (anapanasati shows up verbatim in his thread reply: "he trains himself I will breath in, sensitive to the entire body").

His tribe of peers is not the same as his list of cited thinkers; see Network Graph below.

Evolution across the window

Three months is short for evolution, but a few things visibly tighten:

Blind spots

Worth naming, since the rules say not to sanitize:

II. The Dharma — How He Reads, Teaches, and Argues About Buddhism

This is the section that exists because of who Romeo is. Most of his tweets are dharma-coded even when the surface is something else.

The thesis: practice is procedurally specifiable, lineages mostly aren't

Romeo's standing complaint about most modern Buddhism is that it has lost the procedure. The Heart Sutra appears mysterious, he argues, because translators wanted to render Mahayana key terms verbal-and-process-like but ended up with vocabulary ("emptiness," "form") that obscures the underlying mechanic of three marks applied to five aggregates. His Apr 14 thread (67L) is the cleanest version of this critique and it landed because the dharma-Twitter audience already feels the gap between translated text and felt practice.

He is also cheerfully willing to call out the failure mode in his own neighborhood:

"I wish this were a joke and not an actual quote. 'Using this method, one can attain freedom in two or three months.' First question: 'How can I attain freedom even more quickly?'" (Feb 23, 117L)

The joke is on the seeking posture itself — the very pattern that brings people to practice is the pattern that prevents the practice from working. Same gesture in his Feb 25 (66L): "One of the reasons therapy methods become more complicated over time is that the people for whom the simple version worked leave the community."

Practice as deescalation, not transcendence

A tonal note: he never sells enlightenment. The phrase he uses for what practice does is deescalation — of internal conflicts, of needless suffering, of the second-arrow reaction on top of the first arrow. "Part of the point of practice is making your mind a pleasant ally by deescalating the internal conflicts..." (Apr 2). And "if you don't know how to feel good, it is easy to build a load bearing excuse like 'people who feel good don't accomplish important things'" (Feb 27). The valence here is unmistakable QRI ancestry — practice as a phenomenological optimization, not a metaphysical one.

Anti-perennialism as the live argument

The pinned tweet — the MetaGame debate with @zhukeepa — is the public version of a long-standing fight. Romeo's position, distilled from the corpus: religions vary in whether they contain a working procedure, and lumping them together because their endpoints sound similar is a sloppy abstraction collapse of the sort he warns against in many other contexts (the Necker-cube reply, the Feb 28 tweet about LLMs rounding off to the known). His Mar 8 (36L) is the same point applied in reverse: "We refer to animism based religions as primitive, but I'm pretty sure anthropomorphizing everything is actually the lower level of development here."

This is also why he keeps insisting that 80%+ of global Buddhists believe the Buddha was a supernatural entity (self-reply Mar 9): the lineage-and-procedure he cares about is a minority report inside the actual demographic Buddhism, and the mainstream version drifts toward exactly the kind of metaphysics he rejects.

What he says in long-form that he won't say on timeline

In the Apr 14 thread on the viral Feb 3 image (a comic about a bald guy venerating a golden foot), his own clarifying replies are blunter than his standalone tweets: the focal joke is about people venerating realized sages instead of doing the simple practice, and his caption pulls anapanasati directly from the Pali — "he trains himself I will breath in, sensitive to the entire body." On timeline he writes oblique aphorisms; in thread, with a smaller audience, he names the source text.

III. Psychology, Therapy & Communication

The other persistent register is therapeutic. Romeo has been doing one-on-ones with men for a decade ("over the last ten years, since getting into therapy methods, I've done many many one on ones with men at various places in their lives," from the Feb 24 thread), and the posts where he is most mechanically precise are usually therapy-shaped.

Coercion, decision, and the avoided-not-the-task move

The single highest-engagement aphorism in the window is also his cleanest therapy claim:

"You aren't avoiding a task, you're avoiding a decision." (Apr 22, 73L)

He immediately threaded the elaboration as a self-reply: "We specifically retreat from decisions into decisionless distractions. That's why scrolling is appealing." And earlier (Feb 24, 76L): "If you make a decision based on coercion you will resist your own decision. If you make a decision based on freedom you will do it. Think back to a decision you made that you didn't follow through on and find the coercion." His self-reply: "if it's something you feel coerced about but you nonetheless have to do it there is likely unprocessed grief and acceptance to go through."

This is the Core Transformation move — name the underlying outcome, find the resistance, locate the coercion or grief. It is the working-frame behind many of his tweets that read superficially as productivity advice.

The narcissism screen

The Feb 20 LPT (216L / 9RT) — that you can screen for narcissism by giving someone a hypothetical social situation and watching whether they answer it directly or short-circuit trying to back out the implications for them — is his most-picked-up methodological tweet. His own Q&A in the thread is more useful than the original: "It's a hardware level breakdown. Can only be routed around by difficult computation which can't be faked. The closest to faking is the narcissists who get into communication frameworks and will misdirect from the specific to general principles they can rattle off."

Two telling things in the same thread: he openly identifies himself as someone who has had to detect this in himself ("Yeah, myself"), and he locates his data in a specific community ("lots of bay area nonsense"). The technique is described with the same cadence as a debugging session.

Communication patterns and obsessive thinking

Tue Feb 24 (39L) — the long obsessive-thoughts list — is the cleanest single-tweet example of his therapy voice: stop attending to specifically what was said / all the things that might have meant / specific statements you might make in the future; start attending to what outcome you want / what outcome they want / natural conflicts that might arise / how you might make offers / how you might reduce ambiguity by checking understanding in a non combative way. This is direct lift from communication-skills training (there is a whiff of NVC and of attachment-style work) but compressed into a usable post-format.

His Feb 25 (53L) "to my younger self" list is the meta-version: the upstream skills he wishes he'd targeted earlier are Econ, Relational, Contemplative, Creativity, Expert Judgment, Pedagogical. Relational is the one he flags as the surprise — the skill set he didn't know mattered, learned via "peer therapy" and books on attachment styles. The Expert Judgment line, lifted straight from his MealSquares-era career, also resurfaces in the Apr 4 reply to @bryan_johnson: "When you get sufficiently rich, your impact both on yourself and the world is mostly mediated through experts that you pay. So expert judgement becomes a more important skill than other domains."

The Feb 24 gender thread, on its own terms

The single highest-likes long post (430L / 35RT / 17R / 6Q) deserves its own subsection because it is doing therapy and structural analysis at the same time. The argument: men radicalizing into hatred of women are running a hindbrain calibrated for small-community feedback through urban-scale signal ("From the perspective of your hind brain assuming you live in a small community this is wild feedback. The hind brain is telling you that you must be ducking up really bad to be getting insta dismissed by every single woman in the tribe."), and women are running a defensive filter that is accurate-on-average but rejects honest signal as well as the lying signal it was designed for. Both sides are responding rationally to information environments their psyches were not built for.

In thread, his pushback is consistently against moralization in either direction. To @dtrogers_2: "I don't think the past was a wonderland. Every society on earth has abused the hell out of people, many of the dynamics are just more visible now." To @MLwagie: "Sleazeballs at a historic low by actual crime levels but we've given the remaining ones amplifiers!" The frame is informational-environment-shaped-our-priors; the prescription is "step back from the table and figure out what you actually want so that you are willing to filter harshly for it."

IV. Actionable Principles — Rules to Live By, Pulled From the Corpus

Romeo writes a lot of one-liner rules. Picked for being supported in the corpus and for being usable by a reader; each cited inline.

  1. You aren't avoiding a task, you're avoiding a decision. (Apr 22 2026, 73L) — Diagnose work-paralysis by asking what choice you're not making, not which item on the list you're not doing.
  2. If you make a decision based on coercion, you will resist your own decision. (Feb 24, 76L) — When a "should" doesn't follow through, find the coercion in the original framing.
  3. Beware your accounting ontology, lest you mistakenly categorize things you valued as a cost. (Mar 11, 38L) — The unit you measure in shapes what you preserve.
  4. It can be useful to check whether doubling a particular belonging would make life easier. (Feb 9, 62L) — The cheapest ergonomic upgrade is often a duplicate, not an optimization.
  5. Self: that which you are accountable to. (Feb 27, 33L) — When you can't tell what someone is, look at what they treat as binding.
  6. In order to compensate for selection effects, replace [men|women] with '[men|women] in my training data.' (Feb 24, 11L) — A meta-rule for online generalization.
  7. If you're the sort to get caught up in philosophical or conceptual overlays it can be a helpful reminder that they're mostly downstream of behavior, not upstream. (Feb 18, 35L) — Cognitive reframes that don't change behavior are rationalization, not insight.
  8. Keep it simple and allow the subtlety its own season. (Feb 26, 41L) — In practice, premature complexity is a distraction; complexity becomes legible only after the simple version has worked for a while.
  9. People become evil when they can no longer envision a good world that includes the things they want. (Mar 15, 243L / 16RT) — Maintain a coherent picture of the future you want; if it disappears, predict bad behavior, including from yourself.
  10. Late Buddhism: pleasant ally over miserable-but-stable fixed point. (Apr 2, 150L) — The benchmark for whether your practice is working is whether your default mind-state is becoming friendlier, not whether you've achieved a target experience.

V. Rhetorical Style — What Makes the Tweets Work

The dominant form: koan-shaped one-liner with a hidden mechanism

Most of Romeo's banger tweets follow the same structure: a single sentence that appears aphoristic but on inspection reveals a specific psychological or game-theoretic mechanism. Compare:

The reason these work is that the koan surface gives the reader the dopamine of a clean phrase and the embedded mechanism rewards rereading. They are designed for the medium.

A second form: the deadpan literalist gag

These are Twitter-native — they only land if you read them in the voice of a very tired, very online dharma teacher. He doesn't break frame to explain.

A third form: the long-form structured analysis

Used sparingly (Feb 24 gender thread, Feb 25 to-my-younger-self list, Feb 26 ambiguity-in-practice paragraphs, Apr 16 short story "The Breakfast Question" written by Claude). These earn outsized engagement when they show up because the rest of the timeline is so compressed — the long-form posts read as explicit thesis statements after dozens of compressed implications.

What the thread audience hears that he didn't quite say

The Feb 24 gender thread is the case study. The replies divide cleanly into people who heard a structural-incentive analysis (and brought their own examples — "lemon market," "kakonomics," urban density, post-1970s collapse) and people who heard tribal validation ("MGTOW is the way forward," "Social media showed us that all women are trash"). Romeo never engages the second group on their terms; he replies politely to the first. The gap between intended thesis and audience read is real and he visibly knows it.

The Feb 3 Golden Foot comic (759L / 39RT) is the inverse — the joke is about people venerating the realized sage instead of doing the practice, which is a piece of intra-Buddhist criticism, but most of the replies treat it as a generic absurdist meme. He gives the actual referent only when asked directly ("any realized sage" / "he trains himself I will breath in, sensitive to the entire body"). The audience size on a viral image post does not necessarily come with audience comprehension, and he seems comfortable with that.

Reply voice vs. post voice

The reply corpus is warmer and sharper than the post corpus. Sharper because the stakes of being misread are lower in a 1:1 — see the Apr 6 cogsec line in a reply to @AlexKrusz: "'crazy that we can't block egregious abuse'->'what about the demarcation problem tho?' is a sign of enemy memetic action and should trigger a cogsec check." Warmer because there's actual conversation — see the Apr 1 reply to @wholebodyprayer: "You are one of the reasons I haven't abandoned the site. I also have to remind myself that most of the benefit is for the lurkers."

Rare in the corpus but worth flagging: the reply voice will concede in a way the post voice almost never does. To @forshaper, asking what self-love means: he asks them a question back rather than asserting. To @QuetzalPhoenix on talking about the boomers: "And if you listen to them talk long enough you'll start to yearn for the bombs" — a sharper joke than he'd post standalone.

VI. Contrarian & Hidden Takes — Where He Cuts Against His Own Tribe

Against rationalist self-congratulation

The Mar 27 tweet quoted earlier — "It is 2006, I am discovering that elite colleges use the available tools to justify their current way of doing things / 2016, rationalists / 2026, tpot" — is from inside the tribe, not outside. He's a fifteen-year LessWrong author saying his own scene rationalizes as much as the Ivy League scene he aged out of. This is a quiet but real apostasy.

Against AI-safety theatre

The Apr 17 reply to @Kore_wa_Kore: "Yes and also: people yell at anthropic more because anthropic is the most open to being yelled at." This is an unfashionable observation in his neighborhood — he is calling out a selection effect in safety discourse that mostly goes unsaid.

Against Buddhist metaphysics drift

He is a Buddhist teacher publicly frustrated that 80%+ of global Buddhists treat the Buddha as a supernatural entity, who keeps having to write retranslation footnotes against the Mahayana tradition his audience is most familiar with, and whose pinned tweet is a debate against the soft-perennialism that most contemplative teachers default to. "yes, I feel the buddha tried to be clear about this but people started going metaphysical before his body was even cold." (Apr 14 reply to @InformationBot0)

Against the "block individually" antipattern

Mar 6 (70L): "The most insane antipattern baked into the web is that we should each individually decide who and what to block. I not only want automated tracking of people who act in bad faith, I want second order blocking of anyone who supports them." In the thread he doubles down in a way that surprised some of his peers: "this is a feature, not a bug, block the entire low quality block-coalition." This is a position most of his rationalist-adjacent peers would call illiberal; he frames it as a cogsec primitive.

Against the modernism-is-dead lament

Mar 7 (69L): "I like the idea that larping the good parts of a dead civilization can lead to great things. Let's stop lamenting the death of modernism and just start copying what was good about it unironically." A position that breaks with both the post-rat aesthetic right (which prefers irony) and the progressive consensus (which is suspicious of revival). His genuine view is that imitation is a respectable contemplative move.

Three-drinks-after-hours

Per his own framework — what he'd say more bluntly out of public-broadcast register:

VII. Network Graph

Inner circle (the people he addresses by handle as peers, not audience)

Recurring orbit but not inner circle

@exgenesis, @nickcammarata (QRI alum), @davidad, @SamoBurja, @AlexKrusz (he's been arguing with him repeatedly about second-order blocking), @sarabollman, @suchnerve, @TylerAlterman, @doom_gazer, @made_in_cosmos, @Meaningness's adjacent crowd. He is in the dharma-meets-rationalism-meets-tpot interzone and it shows.

Who he replies back to

The signal: when he gets a top reply on a focal tweet, who does he engage? In the Feb 24 gender thread, his serious engagement is with the structural-analysis replies (kakonomics, lemon-market elaborations, urban-feedback explanations) — not the MGTOW replies, which he does not respond to. In the narcissism thread, he engages @cube_flipper with a real concession ("Yeah, myself"). The pattern: he replies back to people who are adding mechanism, ignores people who are adding heat.

Who he amplifies (quote-tweets)

@dkazand (his own podcast), @vividvoid (Nameless Mountain), @taobanker (workplace strategy → spiritual practice analogy), @TylerAlterman (meditation), @suchnerve (autists/genies), @carmenleelau (self-blame as suffering-avoidance), @AnthonyNAguirre (LLMs and intellectual work), @bryan_johnson (gentle pushback on expert judgment), @dwarkesh_sp (Petrarch / dead civilizations), @gvy_dvpont (analog YouTube for kids → analog dharma talks), @aneeshm (Carl Rogers), @george__mack (worldview-from-9-to-13). The quote-tweet is reliably his "yes, and here's the dharma extension" move.

Who he ignores

Mainstream AI-safety voices (other than oblique references), mainstream rationalist-celebrity figures, anyone whose name is a brand. He is in-scene but at one remove from the scene's stars.

VIII. The One Idea He Keeps Rewriting

If you stack his recent tweets next to his decade of blog posts (Threefold Training, Four Pillars of Practice Progress, Core Transformation, (mis)Translating the Buddha, Orientation on the Contemplative Path) and the AI-coding interview at nothinghuman, the same essay is being written again and again, in different costumes.

The essay is: most human suffering is a pattern of mechanism on top of mechanism that you have learned to treat as monolithic, and the work is to factor it back into pieces that can each be addressed by practice.

In therapy clothes: "You aren't avoiding a task, you're avoiding a decision." In Buddhism clothes: "Emptiness is Mahayana shorthand for the three marks. Form is shorthand for all five aggregates. 3m → 5a remains the core logic." In dating-discourse clothes: hindbrain calibrated for one environment, getting feedback from another. In AI-cogsec clothes: "second-order updates just don't propagate with humans. We don't have a native circuit for it" and "if you don't have any metacognition about affordances, then processes can come in and hijack them." In meta-pedagogy clothes: "Keep it simple and allow the subtlety its own season." In single-line aphorism: "spiritual practice is working when it changes your sense of what's possible."

The reason he keeps rewriting it is that it is — for him — the actual operating instruction, and the reason it doesn't propagate (his complaint about second-order updates) is that on the surface every restatement looks like a different topic. The Personae read of his timeline is that he is, very patiently, trying to change that.

Reading Curriculum

Where to go from here, drawn from what he actually cites and uses: