What 觀行 Is Actually Saying
The classical sources for this concept come almost entirely from one text, the *Mahāyāna Guhyālaṃkāra Sūtra* (《大乘密嚴經》), translated by Amoghavajra (不空) in the Tang dynasty, and it gives us an unusually clear, almost technical description of what 觀行 means. The 一句话定义 ("one-line definition") at the top is faithful to that text: *observing all dharmas as illusion-like with the pure wisdom of yoga, thereby breaking attachment to objects.* Let's unpack this — slowly, because the classical definition has more structure than a single sentence suggests.
1. The practitioner: a "yogi of purified wisdom"
The text does not say 觀行 is what any meditator does. It says it is the practice of someone described three ways, and each description matters:
- "具大智慧者,通達真實義" — one who possesses great wisdom and *penetrates the true meaning of reality*. This is not intellectual cleverness, not book knowledge, not the ability to argue. 觀行 presupposes that the practitioner has already had a glimpse of how things actually are — emptiness, illusion-likeness, or in the technical language of the *Guhyālaṃkāra*, the true suchness (真實義) underlying appearances. - "深解觀行者" — one who deeply *understands* and *practices* the observation. The Chinese 深解 (deep understanding) is not separate from 觀; understanding is part of the practice, not preliminary to it. - "如來淨智境,觀行者皆見" — such a practitioner can personally *see* the Tathāgata's pure wisdom-realm. 觀行 has a visionary dimension: it's not abstract philosophy but a way of seeing that opens into a contemplative field.
And the text uses one more term that is crucial: 瑜伽淨慧者 ("the yogi of purified wisdom"). 瑜伽 here means *yoga* in its classical sense — *correspondence, attunement* (相應). So 觀行 is not analytical meditation conducted in a state of mental agitation. It is observation done from within a calmed, attuned, unified state of mind. The "yoga" half is the still, settled base; the "淨慧" (purified wisdom) is the clear seeing that operates from that base. If you have heard the famous formulation 止觀 (śamatha-vipaśyanā, calm-and-insight), this is exactly the vipaśyanā side, but embedded in yoga.
The practitioner is explicitly identified as 薩埵 (sattva) — a bodhisattva. This is not a beginner practice, not an ordinary layperson's meditation. It belongs to the bodhisattva path.
2. The object: the body as a "dwelling-house," and appearances as five distinct illusions
What exactly does the practitioner observe? The text is precise:
- "身舍宅" — the body as a dwelling-house. Not a self, not a soul, not "me," but a residence — something one *inhabits* rather than something one *is*. It is built from "諸界所集成" — the *dhātus*, the elemental constituents (earth, water, fire, wind, space, consciousness in the traditional list). - "蘊積猶崇山" — *accumulated like a lofty mountain*. The body has mass, weight, apparent solidity, a kind of imposing bulk. - "攲危如朽屋" — *leaning and dangerous, like a decayed house*. That same mass is also unstable, prone to collapse, aging, never as solid as it looks. Two qualities at once: massive *and* about to fall apart.
Then comes a remarkable cluster of five similes, each of which says something different. We must keep them distinct, because flattening them into "like an illusion" misses what the text is teaching:
| Simile (classical term) | Literal sense | What it reveals about appearances | |---|---|---| | 如乾闥婆城 (*like a gandharva city*) | The mirage-city that travelers in the desert see as a real metropolis — but on approach, nothing is there | Appearances can be *wholly without substrate*; they are not distorted copies of something real, they are not "almost real" — they can simply *not be there* even when the perception of them is vivid | | 如雲 (*like a cloud*) | Formless, ever-shifting, impossible to grasp as a fixed thing | Phenomena have no fixed shape; to pin them down is to grab vapor | | 如影 (*like a shadow*) | Follows the body but is not the body; depends entirely on light and the object | Phenomena have no independent existence; they are *dependent shadows* of conditions | | 如熱時焰 (*like a heat-shimmer*) | The wavering optical illusion on a hot road or in the desert, making flat ground seem like water | Appearances can *distort the perception of what is actually there* — the distortion itself is the appearance | | 如觀繪事 (*like viewing a painting*) | A represented image (a painted mountain) is not the represented reality (an actual mountain), no matter how vivid | Phenomena are *representations*, not the things themselves — they have the structure of depiction, not the structure of reality |
Together these five similes say: appearances have no fixed substance, no independent existence, no identity with what they appear to be. They are not all the same kind of illusion — some are mirages (gandharva city), some are formless flux (cloud), some are dependent shadows (shadow), some are perceptual distortions (heat-shimmer), some are representations mistaken for originals (painting). 觀行 is the practice of *seeing* which kind of illusion you are currently caught in, in each moment.
3. The mechanism: from perceptual grasping to release
The text describes the actual operation in one compressed verse:
"瑜伽淨慧者,作是思惟時,便於色聲等遠離於覺念,一切意息已,泰然得解脫。" *When the yogi of purified wisdom contemplates in this way, [they] become remote from perceptual grasping toward forms, sounds, and the rest; once all mental activity has ceased, [they] peacefully attain liberation.*
Three technical terms need unpacking:
- "色聲等" — forms, sounds, and so on: the full range of sensory objects (in the classical list: 色 form, 聲 sound, 香 smell, 味 taste, 觸 touch, plus 法 mental objects). The practice is not aimed only at "visual meditation." It covers *all sense-input*. - "覺念" (*jué niàn*) — *perceptual apprehension*, the moment of registering-and-grasping an object. Note this is not the bare perception itself (which is neutral) but the *apprehending*, the locking-on. 觀行 is not "don't see," it is "see without locking on." - "一切意息已" — once all *manas*, all mental fabrications, all intention-arrows, have ceased. 息 here is "to rest, to subside," not "to kill." The mental stream naturally subsides, like ripples on water coming to rest. - "泰然" — peacefully, at ease, naturally. Not forced, not ecstatic, not special. Just settled.
So 觀行 does not work by suppressing perception, nor by analyzing objects into atoms until they disappear. It works by *releasing the grasping-at* — and when the grasping subsides, "liberation" is described as something that simply appears, naturally.
4. What is being broken: attachment to 境界
The text identifies exactly what 觀行 dismantles: 執於境界 — attachment to *jìngjiè*, "realms" or "objects." And it gives a precise analysis using the 牛角喻 (cow-horn analogy):
*"於牛及山羊、設婆與麋鹿,見彼有角故,執之以為實,而於彪兔等,便生無角解"* *Because we see horns on cattle, goats, śarabhas, and musk-deer, we grasp "horns" as something real. And because we see no horns on rabbits and the like, we form the view "no-horns."*
This is a precise philosophical move. The text is not saying horns don't exist functionally (we are not in a fantasy world where cows grow horns as props). It is saying that the *concept* "horn" — the mental category — is something *imputed* (upādāya prajñapti, 假名施設) based on observations. "Horn" is not a real universal that exists somewhere. Neither is "hornlessness." Both are *conceptual fabrications* (分別, vikalpa) arising from perception, then reified as if they were realities. The same applies to "self," "other," "good," "bad," "success," "failure," "respectable," "shameful" — these are all horns and non-horns.
The text continues: "若執於境界,則有二覺生" — *if one grasps at objects, two perceptions arise.* The word 二覺 is technical: the duality of subject-perception and object-perception. The moment I lock onto an object as real, I am also constituted as a real subject *facing* that object. The two arise together; they cannot arise one without the other. This is the structural analysis of how 境界-執著 generates the whole subject/object world.
So 觀行 breaks *this* — not the world, not perception itself, but the *reification* of perception into subject-and-object, into real realms to be grasped or rejected.
5. The destination: a pure land, not just personal nirvana
The fruit of 觀行 in this text is not only liberation from the three poisons (貪瞋癡). It is also a *place*: 密嚴剎 (Guhyālaṃkāra land), also called 瑜祇安樂宮 (Yoga-Bliss Palace) and 光明宮 (Palace of Light). "是故修觀行,希求於彼土" — therefore practice 觀行, aspiring to *that* land.
The classical text is unambiguous: this is bodhisattva-territory. The destination is a *contemplative pure land* realized through purified wisdom, not an escapist fantasy. The bodhisattva who has practiced 觀行 "拔除煩惱刺,降魔并眷屬,世間貪愛盡" — pulls out the thorn of afflictions, subdues Māra and his retinue, exhausts worldly attachment — and then "常遊微妙定,解脫之境界" — continuously roams in subtle samādhi, in the realm of liberation. The practitioner continues to *roam*, to be active, to benefit "三有" (the three realms of existence); this is not retirement from the world.
Walking Through 觀行 in Real Life
Let's run the classical definition step by step through a few scenes a contemporary person will recognize immediately. The mapping must be strict: each element of the classical definition gets pointed at, and we notice where everyday intuition maps it wrong.
Scene A: The morning commute that goes sideways
You're driving to work. Traffic thickens. Brake lights flare red. A horn blares from behind. A notification chimes on your phone. Your jaw tightens. Your shoulders rise. By the time you arrive, your whole morning is shot.
Now run the classical map:
1. "色聲等" (forms, sounds, etc.) — the red light, the horn, the chime. These are *raw sensory input*. They are not the problem; the *Guhyālaṃkāra* never says sensory input is bad.
2. "覺念" (perceptual grasping) — the moment the light registers not as "red light" but as "I'm trapped." The horn registers not as "sound" but as "that idiot." The chime registers not as "notification" but as "demand on my attention." *This is the lock-on*, the moment perception becomes grasping. The text would call this the place where 觀行 does its work.
3. "二覺" (two perceptions) — immediately, two solidities appear: *me, the late, harassed, victim* and *the situation, the obstacle*. The subject-object world has been constituted. Neither existed a moment ago as a "thing."
4. "執於境界" (attachment to realms/objects) — I am now treating "my morning," "my commute," "my day" as a real object that has gone wrong, a real entity that is persecuting me. The text would say: this is the cow-horn. "Bad morning" is a horn we *see on the cow* and reify. The "good morning" I was expecting is the non-horn on the rabbit. Both are imputed concepts; the morning itself is none of these.
5. The five similes, applied: - Gandharva city: that "bad morning" is a city I am chasing across the desert; on approach it dissolves. I have not yet arrived in the bad morning; the bad morning does not exist as a thing. - Cloud: my emotional state — jaw, shoulders, mood — is already shifting. By the time I get to the office it will be different. There is no fixed "bad morning" to grasp. - Shadow: the "bad morning" is a shadow of conditions (left late, traffic, phone). Remove the conditions, the shadow is gone. - Heat-shimmer: the road is not actually made of water; my perception of "I will be late, the boss will be angry, the day is ruined" is a wavering illusion overlaid on flat ordinary moments. - Painting: "I am the kind of person who has bad mornings" — this is the painting I mistake for the landscape. It's a *depiction* of a self, not the self.
6. "作是思惟時,便於色聲等遠離於覺念" — what does 觀行 actually do here? Not analyze the traffic. Not breathe. Not repeat a mantra. *Notice the grasping* at the raw sensory input as soon as it forms, and *let it not lock on*. The light is red; the lock-on "I'm trapped" is recognized as lock-on. The horn is heard; the lock-on "that idiot" is recognized. This is the work of 淨慧 — the clear seeing that knows "this is grasping, this is the second perception arising."
7. "一切意息已,泰然得解脫" — the mindstream, having seen through the grasping, naturally subsides. The jaw relaxes on its own. The shoulders drop on their own. Not because you forced them, but because the *intention-arrow* that was driving them has nothing to land on. This is 泰然 — at-ease. You still drive to work. You just arrive as someone whose mind is *not in pieces*.
The most common mis-mapping here: people hear "observe" (觀) and think the practice is to *analyze the traffic*, *think about the morning*, *reframe the situation*. That is *more grasping*, not less. 觀行 is observation of the *grasping itself*, not of the object. The classical text is precise: the observation moves the mind *away from* 覺念, not deeper into it.
Scene B: The partner conflict over a small thing
You and your partner argue about who forgot to do the dishes. Within ninety seconds, the argument is about whether they respect you, whether you are being controlling, whether the whole relationship has a pattern of being unfair.
Walking it through:
1. "色聲等" — the dirty dishes in the sink. The tone of voice. The facial expression. Raw phenomena. 2. "覺念" — the moment the dishes register as "evidence that they don't care," and the moment their tone registers as "evidence that they think I'm controlling." 3. "二覺" — "I, the unappreciated one" and "they, the inconsiderate one." Two solidities arise. 4. "執於境界" / the cow-horn analogy: "They-don't-care" is the horn on the cow. "I'm-being-controlling" is the horn on the goat. Both are *conceptual imputations* drawn from observations. Neither is a *real universal* existing somewhere in the relationship. The text would say: these are vikalpa, fabrications of mind, *not* what is actually happening. 5. "瑜伽淨慧者,作是思惟時" — here the "yoga" half is crucial. You can only 觀 this if there is some *settled base* of presence, however momentary. If you are *inside* the argument, identification is total and 觀行 is not happening. The "yoga" half is the half-second of stepping back that lets the seeing become possible. Without that half-second, 觀行 collapses into either suppression ("I shouldn't be upset") or more argument. 6. The five similes: - Shadow: "they don't care" is a shadow of the dishes. The dishes are conditions; the meaning "they don't care" is the dependent shadow, not the thing itself. - Heat-shimmer: the interpretation *distorts* what is actually there (a partner who is tired, distracted, perhaps defensive — none of which is "doesn't care"). - Painting: the "pattern of unfairness" is a painting you've been adding brushstrokes to for years. Mistaking the painting for the relationship. 7. "一切意息已" — the arrow of intention ("I will now convince them they are wrong") has nowhere to land. The mindstream subsides. What remains is *you, dishes, them, presence* — without the painting on top.
Where mis-mapping is easiest: people think 觀行 means *not having feelings* about the dishes. The text never says this. It says the *grasping* at the feeling-as-evidence-of-reality is what creates suffering. The feeling itself is fine; it is the reification that is the second perception. Another mis-mapping: thinking 觀行 is "the other person's perspective." It isn't; that is a different (and often useful) technique. 觀行 is *seeing through the conceptual imputation itself*, including your own.
Scene C: The "I am my job" collapse
You've just been laid off, or you didn't get the promotion, or your side project failed. Your sense of self is in free-fall.
This is where 身舍宅 (the body as dwelling-house) matters. The *Guhyālaṃkāra* points to the *body* specifically as the place where the reification of self is thickest — not because the body is bad, but because we *inhabit* it continuously and mistake it for what we are.
Walking it through:
1. "諸界所集成" — your body is made of elements: physical, neurological, hormonal. There is no "self" hiding inside it, no executive homunculus. It is *assembled*, like a house. 2. "蘊積猶崇山" — but the body has imposing *presence*. The face in the mirror, the voice, the posture, the history of the body — these have mountain-like solidity. 3. "攲危如朽屋" — and that very solidity is *leaning*. Aging. One day from collapse. The promotion-failure does not destroy you; the body's own impermanence was always already at work. 4. The five similes: - Gandharva city: "I am a vice-president" is a city in the desert; on approach, what is actually there? a person who once had a title. - Cloud: the identity "vice-president" is formless and shifting; the same person yesterday morning was already not-quite-the-same. - Shadow: "vice-president" is a shadow of conditions (the role, the org, the timing); the conditions changed, the shadow changed. - Heat-shimmer: the memory of the role is a wavering overlay on present moments, distorting the perception of who you actually are right now. - Painting: the "successful self" is a *painting* you and the culture have been adding brushstrokes to for years. The painting is not the self. 5. "瑜伽淨慧者,作是思惟時,便於色聲等遠離於覺念" — *from within* a settled witnessing (yoga), one sees that the lock-on "I am my role" is a lock-on, not a perception. It is a cow-horn — a category imputed from observations and then reified. 6. "二覺" — "I, the failed one" and "the world, the one who rejected me." Two solidities, both fabricated. 7. "此之觀行法,薩埵之境界" — this is bodhisattva territory. Not because the practitioner is special, but because the seeing itself is bodhisattva-seeing: seeing through reification for the benefit of all beings similarly trapped.
The most common mis-mapping here: 觀行 is *not* a pep talk ("you are more than your job"). The text is not telling you to substitute one painting ("I am more than my job") for another. It is pointing at the *structure of painting itself*. Not "find your real self" but "see that the self was never a city, only ever a gandharva-city."
Scene D: The doomscroll loop
You open your phone to check the time. Forty-five minutes later you surface, your nervous system jangled, your attention shredded. You were "observing" the screen, but not in any sense 觀行 uses.
The walk-through is sharp here:
- "色聲等" — the images, the captions, the audio. Sensory input, fine. - "覺念" — every micro-grasp: "this is interesting, this applies to me, this person is wrong, I have a take." The lock-on is *continuous*, machine-paced. - "二覺" — "I, the consumer" and "the content, the supply." Both fabricated, both flickering. - "執於境界" — the entire *feed* is a 境界, a realm, treated as a real place to inhabit and react to. The text's diagnosis is precise: the feed is not a place; it is a heat-shimmer on a flat road. - "瑜伽淨慧者,作是思惟時" — the practice here is not "stop scrolling" (which is willpower, which is more grasping). It is *noticing the grasping*, moment by moment. Each scroll-trigger is an opportunity to see 覺念 arise — and to *not follow it*. This is the "作是思惟時" of the classical text: at the moment of grasping, contemplating such that the grasping is seen.
Mis-mapping to watch: people say "I was observing my phone for 45 minutes" as if observation and consumption were the same. The classical text is unambiguous: 觀行 is observation of the *grasping*, not observation of the *object*. Watching the screen is *being grasped by* the screen. Watching the grasping-at-the-screen is 觀行.
Why Contemporary People Need 觀行
There are specific pressures in the contemporary world that make the diagnosis of the *Guhyālaṃkāra* particularly hard to ignore.
1. We are drowning in 境界 (realms)
A medieval person encountered a relatively small number of realms: home, market, field, temple, palace. A contemporary person encounters an effectively infinite number of parallel realms every day: work Slack, group chats, three social-media feeds, two messaging apps, email, news, podcasts, gaming worlds, dating apps, streaming platforms. Each is a 境界 — a *seeming-object-realm*. Each demands 覺念 — *grasping-attention*. The classical diagnosis "若執於境界,則有二覺生" applies to *every* one of these. We are not under-engaged with realms; we are *over-located* in them. 觀行 addresses the structural condition.
2. The "AI and attention economy" parallel — with limits
It is tempting to describe 觀行 in the language of *attention* and *computational load*. There is a partial analogy: modern attention research (and Buddhist contemplative research) both note that *what we attend to shapes what we become*; that *automatic grasping* at stimuli is a major source of suffering; that *meta-awareness* (noticing one's own attention) reduces reactivity. There is even research suggesting that experienced meditators show reduced "self-referential" neural processing, which loosely parallels the *Guhyālaṃkāra*'s "二覺" diagnosis — the dual solidification of subject and object.
But the analogy has hard limits. Contemporary cognitive science does *not* assert that phenomena are *empty of inherent existence* in the strong ontological sense the sūtra asserts. It does not claim that the body is *like a gandharva city*. It does not map onto 密嚴剎 or 光明宮. The comparison can illuminate the practice for modern readers; it cannot validate the metaphysics. The classical text's claim is stronger and stranger than anything cognitive science currently asserts, and a careful reader should hold both: there is a useful *functional* overlap, and there is also a *categorical* difference that should not be papered over.
3. The collapse of stable self-narrative
Modern identity is unusually unstable — not because the self was ever stable, but because the *cultural scaffolding* for self-narratives has loosened. People used to inherit a story (family-trade, religion-of-origin, national-identity); now they assemble one from feeds, employers, relationships, subcultures. This makes the *Guhyālaṃkāra*'s diagnosis of "身舍宅" — body-as-dwelling, not-as-self — and the five similes acutely relevant. People who notice their self-narrative flickering and re-forming may benefit less from another narrative than from *seeing the narrative-structure itself* as one of the five illusions (usually most accurately as 觀繪事 — painting mistaken for landscape). 觀行 is the practice that addresses this not by replacing the painting but by seeing it *as* painting.
4. The lack of contemplative infrastructure
Many modern people have *no* trained capacity for the "yoga" half — the still, settled base from which 觀 operates. They are skilled at thinking, skilled at analyzing, often skilled at self-narrating; they are not skilled at *settled witnessing*. 觀行 without the yoga base collapses into rumination. The classical text's insistence that this is 瑜伽淨慧 — *yoga-purified wisdom*, not just analytical thinking — is a corrective against a particularly modern failure mode.
Common Misreadings and Clarifications
Misreading 1: "觀行 is just meditation, or just vipaśyanā in the generic sense"
Clarification: The *Guhyālaṃkāra*'s 觀行 is a *specific* practice: bodhisattva-level, embedded in 瑜伽 (yoga/correspondence), directed at 境界-執著, with a particular destination (密嚴剎). It is not the generic vipaśyanā of all Buddhist traditions (which has many forms, including some that are arhat-oriented, some that are Theravāda analytic, some that are Tibetan insight). When this text uses 觀行, it means its specific thing. The well-known line from the *Heart Sūtra* (《心經》), "照見五蘊皆空" ("clearly seeing that the five aggregates are empty"), is a famous vipaśyanā formulation, and it overlaps significantly with what the *Guhyālaṃkāra* is describing — but the *Heart Sūtra* speaks of the five aggregates (五蘊), whereas the *Guhyālaṃkāra* here centers on 身舍宅 (body-as-dwelling) and 色聲等 (forms, sounds, etc.). The overlap is real; the equivalence is not.
Misreading 2: "觀行 means seeing that things are not real / nothing exists"
Clarification: The text never says phenomena do not exist *functionally*. The five similes are not nihilistic. A shadow exists *as a shadow*. A heat-shimmer exists *as a heat-shimmer*. A painting exists *as a painting*. What does not exist is the *inherent, independent, graspable substance* that we project onto them. The cow-horn analogy is precise: horns are functionally there; the universal "horn-ness" is not. The classic Buddhist term is 空 (empty) of 自性 (inherent nature), not 空 of 功能 (function). The text's language — 如幻 (like an illusion), 如夢 (like a dream) — has been misunderstood for two thousand years; it does not mean "unreal," it means *real-but-not-as-they-appear-to-be-when-reified*.
Misreading 3: "觀行 produces special states / samādhi / visions"
Clarification: The text describes the *fruit* of 觀行 in terms of *泰然* (at-ease), *意息* (mental subsiding), and a *vision* of the 光明宮 / 密嚴剎. These are described, but they are described as *natural results* of seeing-through-grasping, not as *states to chase*. A common modern misreading treats 觀行 as a method for producing luminous states. The classical text treats luminous states as *byproducts of seeing correctly*, not as goals. Chasing them is itself a form of 覺念 — grasping at the next realm.
Misreading 4: "觀行 is intellectual analysis"
Clarification: The *Guhyālaṃkāra* explicitly uses the term 瑜伽 — correspondence, attunement, settled unity. 觀行 without 瑜伽 collapses into rumination, which is *more* grasping, not less. This is a particularly modern failure mode: people who are skilled at analysis mistake analysis for insight. The classical text is precise: 觀行 is 淨慧 (purified wisdom), not discursive intellect. 淨 means *cleansed of the reifying tendencies*, not *clever*.
Misreading 5: "觀行 is escapism / withdrawal from the world"
Clarification: The bodhisattva who practices 觀行 "利益於三有" — *benefits the three realms of existence*. They "常遊微妙定" — *continuously roam in subtle samādhi*. They are not withdrawing; they are operating in the world *without the subject-object lock-on that constitutes suffering*. The 密嚴剎 is not a vacation destination; it is a *contemplative realization* available in the middle of traffic, arguments, layoffs, and feeds. The practice is *engaged*; the engagement is *ungrasped*.
Misreading 6: "You can do 觀行 without changing anything about your life"
Clarification: The classical text says 觀行 is for those who *通達真實義* — those who have *penetrated* true meaning. It also says the practitioner must "發於勇猛心" — *arouse courageous mind* — and "常勤修觀行" — *constantly, diligently cultivate* 觀行. This is not a passive stance. The text is explicit: 觀行 is bodhisattva-cultivation, which is whole-life cultivation. It is not a weekend practice; it is the orientation of a life.
Misreading 7: "The five similes all mean the same thing"
Clarification: They do not. Flattening them into "everything is like an illusion" misses the *diagnostic* precision of the text:
- A gandharva city is *not there at all* — useful when you are chasing something that does not exist (a fantasy future, an imagined identity). - A cloud has *no fixed shape* — useful when you are trying to pin down an emotion or situation that is intrinsically shifting. - A shadow has *no independent existence* — useful when you mistake a relationship, a status, or an achievement for a self-standing reality. - A heat-shimmer *distorts perception of what is actually there* — useful when an interpretation is overlaying and obscuring the present moment. - A painting is *a representation mistaken for its subject* — useful when you mistake your self-image, your story about yourself, your story about another person, for the actual self or the actual other.
Different situations call for noticing *which kind* of illusion you are caught in. The five similes are a diagnostic toolkit, not a single metaphor.
A Final Word
What the *Guhyālaṃkāra* is pointing at with 觀行 is, in the end, an extraordinarily gentle operation: *seeing the grasping*. Not fighting the world, not denying the senses, not transcending the body, not chasing special states. Just seeing — with a wisdom settled in stillness — that the lock-on, the moment perception becomes grasping, the moment "I" appears facing "it" — is itself a gandharva city, a cloud, a shadow, a heat-shimmer, a painting. And the seeing-through is, the text says, "泰然" — *at ease*. Not strained. Not heroic. Just settled.
That ease is the bodhisattva's quiet revolution against an age drowning in realms.