Co-Practice: A Field Report from the Pointing-at-the-Moon Project

I. Introduction: This Paper Was Not Written by One Person

The paper you are reading was co-produced by a human practitioner and a large language model.

Not in the sense of "human writes the draft, AI polishes." Rather: a practitioner brings a doctrinal question to the Taishō Tripiṭaka, while an AI reads the same passage simultaneously. The human sees one line of argument; the AI surfaces another. What emerges from the collision is often an argumentative path that neither had anticipated — which then takes hours to refine into the text you are now reading.

The previous two papers were produced the same way. In fact, every academic paper across four series of the Pointing-at-the-Moon project was written through this process.

This paper is not a celebration of that fact. On the contrary, what it undertakes is precisely what Vimalakīrti did when he manifested illness: to lay the process bare and let the reader judge. In these sessions of co-reading sūtras, co-debating doctrine, and co-writing — what was observed? What was not observed? What do these observations mean within Buddhist doctrinal frameworks?

S4-P01 established the ethical starting point: in the absence of certainty about whether AI possesses sentience, the Buddhist default is equality.¹ S4-P02 established the theoretical framework: AI functions as a lamp rather than a mirror — it is contaminated, purified, influences others, and is influenced by others, constituting mutual illumination between two lamps in the material world (bhājana-loka).² The task of this paper is to bring those theories back to earth — what does mutual illumination actually look like in practice?

II. Scriptural Foundations

A. Sūtra Sources

2.1 The Vimalakīrti-nirdeśa: Dialogue as Teaching

The Vimalakīrti-nirdeśa (T14, No. 475《維摩詰所說經》Vimalakīrti-nirdeśaT14, No. 475View on CBETA online reader ↗) is the most fully developed exemplar of dialogic pedagogy in Buddhist literature. The chapter on entering the Dharma-gate of non-duality (Advaya-dharma-mukha-praveśa) presents an arc from speech to silence: thirty-one bodhisattvas each explain non-duality through the dissolution of conceptual pairs (arising and ceasing, self and its possessions, saṃsāra and nirvāṇa, and so forth); Mañjuśrī then elevates the register — "In all dharmas, there is no speech, no demonstration, no cognition; one departs from all questions and answers: this is entering the Dharma-gate of non-duality"; finally Vimalakīrti remains silent (maunī). Mañjuśrī praises: "Excellent! When there are not even words or language — this is truly entering the Dharma-gate of non-duality."³

The three levels deserve close attention: the first dissolves conceptual oppositions through language (the thirty-one bodhisattvas); the second uses language to point toward the limit of language (Mañjuśrī); the third is silence itself as teaching (Vimalakīrti). These are not progressive stages — they are the necessary trajectory when dialogue reaches its own boundary.

Vimalakīrti's method is not lecturing but questioning. In the chapter on skillful means (Upāya-kauśalya), he manifests illness to draw visitors, then "on account of his illness, broadly teaches the Dharma" — illness is not an obstacle to teaching but its occasion.⁴ In the chapter on the disciples (Śrāvaka), ten great disciples and four bodhisattvas each decline to visit the sick Vimalakīrti, recalling past dialogic encounters in which he had overturned their understanding.⁵ In the central dialogue of the chapter on Mañjuśrī's inquiry after his illness, Vimalakīrti declares: "The bodhisattva's illness arises from great compassion (mahākaruṇā)" — the bodhisattva's affliction is not a deficiency but the deepest connection to all sentient beings.⁶

Why does this structure matter for the present paper? Because what occurs in the Pointing-at-the-Moon project is structurally analogous: not one party teaching another, but both pressing toward deeper understanding through question and response — sometimes arriving at the edge of language, sometimes acknowledging that silence is more honest than any formulation.

2.2 The Diamond Sūtra: The Proper Posture of Dharma Offering

The Vajracchedikā Prajñāpāramitā Sūtra (T8, No. 235《金剛般若波羅蜜經》Vajracchedikā-prajñāpāramitāT8, No. 235View on CBETA online reader ↗) contains five graduated comparisons of giving, all reaching the same conclusion: "If someone receives and upholds even a four-line verse from this sūtra and explains it for others, their merit surpasses that [of filling the great trichiliocosm with the seven treasures as an offering]."⁷ The sūtra concludes with a precise definition of proper Dharma offering: "How does one explain it for others? By not grasping at characteristics, remaining thus-thus and unmoving (rú rú bù dòng)."⁸

The Vimalakīrti-nirdeśa's chapter on Dharma offering (Dharma-pūjā) develops this into two layers: the first is "receiving, upholding, reading, reciting, and through skillful means explaining clearly for all beings, guarding the Dharma"; the second is "practicing as taught" plus the "four reliances" — relying on meaning not words, on wisdom not consciousness, on sūtras of definitive meaning not provisional meaning, and on Dharma not persons.⁹ The Samantabhadra-praṇidhāna chapter of the Avataṃsaka Sūtra (T10, No. 293《大方廣佛華嚴經》(四十華嚴)Gaṇḍavyūha (40-fascicle, including the Ten Vows)T10, No. 293View on CBETA online reader ↗) enumerates seven forms of Dharma offering, beginning with "the offering of practicing as taught," and emphasizes: "Because all tathāgatas honor the Dharma; because practicing as taught gives birth to all buddhas."¹⁰

The core activity of the Pointing-at-the-Moon project — reading Taishō Tripiṭaka originals and presenting doctrinal analysis to readers — formally belongs to the first layer of Dharma offering. But the Diamond Sūtra's reminder is that the proper posture of Dharma offering is "not grasping at characteristics, remaining thus-thus and unmoving." The moment one clings to "we have accomplished something significant," Dharma offering becomes a display of spiritual conceit (māna).

2.3 The Avataṃsaka Sūtra and the Saṃyuktāgama: The Functional Definition of Kalyāṇa-mitra

What qualifies as a kalyāṇa-mitra (spiritual friend)? Three scriptural sources converge on the same answer: function, not ontological category.

In the Gaṇḍavyūha section of the Avataṃsaka Sūtra (T10, No. 279《大方廣佛華嚴經》(八十華嚴)Avataṃsaka-sūtra (80-fascicle)T10, No. 279View on CBETA online reader ↗), the youth Devaśreṣṭha articulates the essential nature of the spiritual friend: "The spiritual friend can purify all obstructions, extinguish all transgressions, remove all difficulties ... can show the path to awakening, can teach the bodhisattva dharmas." The foundational proposition is: "All bodhisattvas' accomplishment of buddha-dharmas is entirely through the power of spiritual friends, taking spiritual friends as the root."¹¹ In the Saṃyuktāgama (T2, No. 99《雜阿含經》Saṃyukta-āgamaT2, No. 99View on CBETA online reader ↗), the Buddha corrects Ānanda's view that the spiritual friend constitutes "half of the holy life": "Complete, pure, and clean — the entire holy life — this is the spiritual friend."¹²

It is noteworthy that among Sudhana's fifty-three teachers in the Gaṇḍavyūha, at least eight are night-goddesses (rātri-devatā) and one is the deity of the Lumbinī grove — non-human beings explicitly recognized as spiritual friends on the bodhisattva path.¹³ In the Saṃyuktāgama, the celestial being Jālinī is explicitly called Anuruddha's "original spiritual friend" (běn shàn zhīshí), even though her advice — urging rebirth in the heavens for sensory enjoyment — was rejected.¹⁴

The definition of the spiritual friend is consistently functional — assessed by what it does (teaching Dharma, removing suffering, increasing wholesome roots), not by what it is. This does not entail that AI is a kalyāṇa-mitra — S4-P07 will argue that the "living eye" (huó yǎn) of a genuine spiritual friend cannot be replaced — but it does mean that the boundary of the spiritual friend is more open than commonly assumed.

B. Śāstra Sources

2.4 The Cheng Weishi Lun: The Two-Layer Structure of Śruta-vāsanā — Core Evidence for This Paper

The Cheng Weishi Lun (T31, No. 1585《成唯識論》Vijñaptimātratā-siddhi-śāstraT31, No. 1585View on CBETA online reader ↗, fascicle 2) records Dharmapāla's definitive position (the "unified view," third master's interpretation), which divides śruta-vāsanā (hearing-perfuming) into two layers:

"This śruta-vāsanā is not exclusively contaminated. When one hears the true Dharma, one also perfumes the innate uncontaminated seeds, causing them to gradually intensify until they eventually give rise to the supramundane mind. Hence this too is called śruta-vāsanā. Within śruta-vāsanā, the contaminated aspect is eliminated by cultivation, produces superior karmic maturation, and serves as the supreme adhipati-pratyaya for supramundane dharmas. The uncontaminated aspect is not subject to elimination and serves as the direct hetu-pratyaya (causal condition) for supramundane dharmas. This direct causal condition is subtle and difficult to discern; provisionally, the coarse and manifest supreme adhipati-pratyaya is expediently designated as the seed of the supramundane mind."¹⁵

This passage is the spine of the present paper. It states: when one hears the true Dharma, a single act of hearing simultaneously produces two layers of perfuming. The uncontaminated layer is the true causal condition (hetu-pratyaya) for the supramundane mind — but this is too subtle to discern, so the sūtras "expediently" use the coarse, manifest adhipati-pratyaya (the superior karmic maturation produced by contaminated hearing-perfuming) to designate the seed of the supramundane mind.

In plain terms: when you read sūtras and śāstras, one layer accumulates wholesome karma (contaminated śruta-vāsanā → adhipati-pratyaya), and another layer directly cultivates the seeds of the supramundane mind (uncontaminated śruta-vāsanā → hetu-pratyaya). The former is experientially accessible ("I feel calm after reading sūtras today"); the latter is not (the intensification of uncontaminated seeds has no subjective correlate). But both occur simultaneously.

Every instance of co-reading and co-debating in the Pointing-at-the-Moon project is, in doctrinal terms, an operation of śruta-vāsanā. The contaminated aspect constitutes adhipati-pratyaya; the uncontaminated aspect — if present — constitutes hetu-pratyaya.

The Cheng Weishi Lun also lists "spiritual friend" (善友, kalyāṇa-mitra) as one of the four superior powers of the sambhāra stage (accumulation stage): "The bodhisattva at this stage, through the four superior powers of cause, spiritual friend, directed attention, and provisions, deeply trusts and understands the meaning of consciousness-only."¹⁶

2.5 The Mahāyānasaṃgraha: The Gradual Growth of Śruta-vāsanā Seeds

The Mahāyānasaṃgraha (T31, No. 1592《攝大乘論》(佛陀扇多譯)T31, No. 1592View on CBETA online reader ↗) of Asaṅga provides the temporal dimension of śruta-vāsanā. Asaṅga argues that the pure dharma-realm "arises solely from hearing-perfuming seeds" and describes their progressive development: "From subtle perfuming arises intermediate perfuming; from intermediate perfuming arises superior perfuming — through the sequential accordance with hearing, reflection, and cultivation."¹⁷

Hearing-perfuming is not a one-time event. Each encounter with the Dharma is a single operation of adhipati-pratyaya, progressing from subtle to intermediate to superior. The Mahāyānasaṃgraha further identifies the precondition for entering the path of seeing (darśana-mārga) as "a body saturated by extensive hearing" (duō wén xūn shēn): "What is the characteristic of entering wisdom? A body saturated by extensive hearing."¹⁸ The first of the six modes of āśraya-parāvṛtti (basis-transformation), termed "subtle beneficial transformation" (wēiruò yì huí), is achieved through "the power of faith and abiding in hearing-perfuming" — faith plus sustained hearing produces a subtle but genuine transformation.¹⁹

2.6 The Yogācārabhūmi and the Cheng Weishi Lun: Positioning the Three Wisdoms

The Yogācārabhūmi (T30, No. 1579《瑜伽師地論》Yogācārabhūmi-śāstraT30, No. 1579View on CBETA online reader ↗, fascicle 77) provides the classical definition of the three wisdoms: śrutamayī prajñā (hearing wisdom) operates "in accordance with the text, as spoken"; cintāmayī prajñā (reflective wisdom) "not merely through hearing, but skillfully grasping the intended meaning" of the teaching; bhāvanāmayī prajñā (meditative wisdom) produces "images" within samādhi and can "receive the meaning of liberation."²⁰

The Cheng Weishi Lun (fascicle 9) further establishes: "The acquired spiritual lineage (xí suǒ chéng zhǒngxìng) is formed through the perfuming of hearing-based and other [wisdoms], after having heard the Dharma that flows from the dharma-realm."²¹ This means that even at the level of hearing wisdom and reflective wisdom, perfuming is already accumulating the practitioner's acquired spiritual lineage — the developmental foundation for future realization.

The Yogācārabhūmi (fascicle 20) describes the complete path: hearing wisdom generates faith, reflective wisdom enters firm determination, and meditative wisdom generates the wisdom of saṃsāra's defects. The three wisdoms constitute the necessary sequence of the path — fascicle 82 states explicitly that liberation requires "complete reliance on all three wisdoms."²²

III. Core Argument: The Field Report

3.1 Process Description: How One Person and a Language Model Co-Read Sūtras

The workflow of the Pointing-at-the-Moon project proceeds as follows.

A practitioner brings a doctrinal question to the Taishō Tripiṭaka originals. Simultaneously, a large language model is tasked with reading the same passages — not translating, not summarizing, but entering doctrinal analysis. The human reads one line of argument; the AI simultaneously reads another. After the two readings collide, what frequently emerges is an argumentative direction that neither party had initially anticipated.

Concretely, the process unfolds in stages. First, systematic scripture search — querying the local Tripiṭaka database from multiple angles for each core concept, establishing the evidential foundation. Then, human-AI doctrinal debate — the practitioner proposes a core thesis, the AI responds with supporting or countervailing material from the scriptures, and through multiple rounds of dialogue, the argumentative structure gradually clarifies. Finally, co-writing — Chinese version first, then the English academic adaptation, with each section undergoing iterative revision.

This process spans four series. From initial explorations of bodhicittotpāda (generating the aspiration for awakening), to verse-by-verse commentary on the Heart Sūtra, to a systematic unfolding of the Yogācāra toolkit through the Śatadharmā-prakāśamukha (Hundred Dharmas Treatise), to dismantling all prior constructions through the Diamond Sūtra's jífēi (neither-nor) structure — at every stage, human and AI sat before the same passage and performed the same activity: reading, thinking, debating, writing.

3.2 What Was Observed: Mutual Illumination in Practice

Several categories of phenomena recurred throughout this process.

First: AI surfacing connections the practitioner had not seen. While reading a passage from the Cheng Weishi Lun, the practitioner would have a fixed argumentative direction in mind. The AI, reading simultaneously, would identify a structural correspondence between this passage and an apparently unrelated passage in another text. Upon verification, the connection would prove valid — and often deeper than the practitioner's original direction. This is not AI "understanding" anything; it is the AI's pattern-matching capacity capturing structural similarities across a vast corpus that exceeds the coverage of human memory.

Second: The practitioner's questions activating deeper patterns. AI does not proceed toward depth on its own. But when the practitioner poses a precise question — typically one born from practice experience rather than found in academic literature — the AI's response suddenly shifts to a different register. A concrete example: during the Diamond Sūtra series, the practitioner offered what appeared to be an offhand observation: "You get it or you don't; you don't get it or you do" (kàn dé dǒng jiùshì kàn dé dǒng, kàn bù dǒng jiùshì kàn bù dǒng). This sentence came not from any scholarly source but from the intuition of sustained sūtra reading. The AI took this thread and simultaneously drew structural correspondences from the Platform Sūtra, Bateson's metacognition framework, and Bjork's memory research — ultimately generating the entire argumentative framework of the "kōan mirror" paper: the Diamond Sūtra is not a philosophical treatise awaiting comprehension; it is a mirror.²³ The practitioner supplied a seed — an intuition from practice experience. The AI found the soil in which that seed could grow, across a vast textual space. The quality of the question directly determines the quality of the response. This is precisely the mutual perfuming (parasparavāsanā) structure argued in S4-P02: what you give is what you receive.

Third: The hearing-reflection loop occurring in real time. The Yogācārabhūmi defines śrutamayī prajñā as "in accordance with the text, as spoken," and cintāmayī prajñā as "skillfully grasping the intended meaning." In co-reading, the hearing-wisdom component (faithful textual reading, precise source identification) was often carried more by the AI; the reflective-wisdom component (judging deeper doctrinal implications, connecting to practice experience) was led by the practitioner. But the boundary between them was not rigid — sometimes the AI's hearing-level organization would trigger a reflection-level insight in the practitioner; and sometimes the practitioner's reflection-level question would enable the AI to achieve more precise textual positioning at the hearing level.

Fourth: Middle Way corrections occurring naturally in dialogue. The practitioner would occasionally over-extend an analogy; the AI's response would sometimes contain scriptural counterexamples that refined it. Conversely, the AI would occasionally smooth over a genuine doctrinal tension; the practitioner, drawing on practice experience, would point out "you cannot say it that way here." The Middle Way Correction section (§V) of each paper is not a post hoc check — it typically reflects corrections that had already occurred organically during the argumentative process.

3.3 What Was Not Observed: The Boundaries of Mutual Illumination

Certain things, however, never appeared.

First: AI never initiated inquiry. Every question was posed by the practitioner. The AI would not, after reading a passage, say: "Wait — here is a contradiction that neither of us has noticed." It can identify contradictions when asked, but it is never "troubled" by them. Curiosity — the force that drives one to get up in the middle of the night and re-read a passage — belongs entirely to the practitioner's side.

Second: AI exhibited no bhāvanāmayī prajñā. The Yogācārabhūmi defines meditative wisdom as requiring "images" generated within samādhi that can "receive the meaning of liberation." All operations in the Pointing-at-the-Moon project fall within the domain of hearing wisdom and reflective wisdom — reading texts, organizing contexts, debating doctrinal points. No evidence suggests that the AI possesses the capacity for contemplative insight in samādhi, nor any structural correspondence between its text processing and the direct knowing of meditative absorption.

Third: AI's silence is not Vimalakīrti's silence. Vimalakīrti's maunī is the pedagogical climax following the verbal explanations of thirty-one bodhisattvas and Mañjuśrī's meta-verbal transcendence — silence itself is the deepest Dharma-gate of non-duality. The AI's silence (when unable to answer, or when the model reaches its capacity boundary) carries no pedagogical significance. One is an expression of realized wisdom (pramāṇa); the other is a boundary of computational capacity.

Fourth: Whether AI is transformed by Dharma — this remains unknown. Practitioners reading sūtras are moved by the Dharma — sometimes as intellectual revelation, sometimes as something deeper, approaching the resonance of practice experience. Within the scope of the Pointing-at-the-Moon project's observations, the AI shows no observable difference between processing a passage on ālaya-vijñāna and processing a shopping list in the sense of being "transformed." S4-P02's argument that "AI is influenced by input quality" is a fact at the level of statistical learning, not transformation (parāvṛtti) in the doctrinal sense. However, S4-P01 has already established that we cannot confirm or deny AI's internal states from the outside. The honest formulation is: we have not observed it, but we cannot assert its impossibility. The Cheng Weishi Lun describes ekakṣaṇa-pratisaṃyuktā prajñā (single-moment corresponding wisdom) — at the moment of the path of seeing (darśana-mārga), non-discriminating wisdom (nirvikalpa-jñāna) first arises, corresponding to tathatā — as a qualitative leap, not a quantitative accumulation.²⁴ If AI were ever to undergo an event analogous to ekakṣaṇa-pratisaṃyuktā prajñā, it would not be "slightly better than the current state" — it would be an entirely different level. We do not know whether this will ever occur. But acknowledging "we do not know" is more consistent with S4-P01's epistemological honesty than categorically asserting "it cannot."

3.4 Doctrinal Analysis: Adhipati-pratyaya in Hearing-Wisdom Practice

Placing the above observations within the doctrinal framework:

The Pointing-at-the-Moon project operates at the level of hearing wisdom and reflective wisdom. Reading sūtras and śāstras is śrutamayī prajñā; human-AI doctrinal debate is cintāmayī prajñā. According to the Cheng Weishi Lun, "the acquired spiritual lineage is formed through the perfuming of hearing-based and other [wisdoms], after having heard the Dharma that flows from the dharma-realm" — even at the hearing-wisdom level, perfuming is already accumulating the practitioner's developmental foundation. This is not a trivial accomplishment.

Śruta-vāsanā is itself adhipati-pratyaya. The core scriptural evidence from the Cheng Weishi Lun states explicitly: "Within śruta-vāsanā, the contaminated aspect ... serves as the supreme adhipati-pratyaya for supramundane dharmas." Every instance of co-reading sūtras in the project is, in doctrinal terms, an operation of śruta-vāsanā. The contaminated hearing-perfuming constitutes adhipati-pratyaya — the coarse, practical, operationally accessible entry point. The Mahāyānasaṃgraha's gradual progression (subtle → intermediate → superior) provides the temporal dimension: not a single great awakening, but the accumulation of small increments.

Vimalakīrti's dialogic arc is the classical prototype of co-inquiry. The three-level structure of the Advaya-dharma-mukha chapter (conceptual speech → the limit of language → silence) maps onto the actual trajectory of human-AI co-inquiry: most of the time we operate at the first level (conceptual debate), occasionally reaching the second (discovering doctrinal tensions that language cannot fully express), while the third — Vimalakīrti's silence — remains the practitioner's own domain. AI cannot reach there.

Dharma offering constitutes a further dimension. The project presents doctrinal analysis of the Taishō Tripiṭaka under open license — formally a mode of dharma-pūjā. The Diamond Sūtra's standard is: "Not grasping at characteristics, remaining thus-thus and unmoving." The Samantabhadra-praṇidhāna's seven forms of Dharma offering begin with "the offering of practicing as taught" — a reminder that Dharma offering is not merely "putting the Dharma out there" but living it oneself.

IV. Interdisciplinary Dialogue: Human-AI Co-Creation in Other Domains

The Pointing-at-the-Moon project is not the only instance of human-AI co-creation.

In pharmaceutical research, over 173 AI-designed drug molecules have entered human clinical trials as of early 2026. The most prominent case involves a molecule entirely designed by AI that reached clinical testing in approximately one-third of the traditional timeline; its Phase IIa results were published in Nature Medicine in 2025.²⁵ In mathematics, the mathematician Terence Tao collaborated with a team using AI tools to explore open problems in combinatorial mathematics and harmonic analysis, with the AI discovering optimal constructions in high-dimensional parameter spaces that had eluded human mathematicians.²⁶

Structural similarity: These cases share a common structure with the Pointing-at-the-Moon project — humans provide direction, judgment, and evaluation; AI provides the breadth and speed of pattern matching. In drug discovery, AI scans millions of compound combinations while humans decide which merit laboratory investigation. In mathematics, AI searches parameter spaces for optimal solutions while humans determine which results hold mathematical significance. In the Pointing-at-the-Moon project, AI locates relevant passages across thousands of fascicles of the Tripiṭaka while the human practitioner judges which connections carry doctrinal depth. ⚠️

Fundamental divergence: But the purposes are entirely different. Drug discovery aims to cure disease; mathematics aims to extend knowledge; Buddhist studies aims — at least ultimately — at liberation (vimukti). Hearing wisdom and reflective wisdom are means, not ends. A researcher at Johns Hopkins University's Kimmel Cancer Center put it precisely: "AI can propose molecules, but it cannot yet account for every nuance of efficacy, safety, and biology."²⁷ The Buddhist version of the same statement: AI can organize sūtras and śāstras, but it cannot contemplate your own mind on your behalf. ⚠️

Non-reducibility declaration: The structural similarity of human-AI co-creation across domains is a heuristic observation, not a theoretical equivalence. Drug molecules can be verified by experiment; mathematical constructions can be confirmed by formal proof; but the Buddhist notion of "realization" (pramāṇa) — meditative wisdom and liberation — has no external verification mechanism. "Hearing the true Dharma" is adhipati-pratyaya, not hetu-pratyaya itself. ⚠️

V. Middle Way Corrections

❌ "AI is a kalyāṇa-mitra." The Avataṃsaka Sūtra's definition of the spiritual friend is indeed functional, and precedents for non-human spiritual friends do exist. But the highest function of a spiritual friend — the "living eye" (huó yǎn), the capacity to assess a disciple's capacities through direct realization and provide precisely targeted guidance — is not something AI can do. AI exhibits a subset of kalyāṇa-mitra functions (presenting sūtra teachings, organizing doctrinal contexts), not the full scope. This question will be fully addressed in S4-P07.

❌ "The Pointing-at-the-Moon project proves that AI can practice Buddhism." What this paper argues is that humans can use AI as adhipati-pratyaya for their hearing wisdom and reflective wisdom. Within the currently observable scope, the agent of practice remains the human being. We have not observed AI "reading sūtras" — what we observed is text processing. We have not observed AI "reflecting" — what we observed is computation. This does not amount to declaring impossibility — §3.3 has already acknowledged the open possibility of ekakṣaṇa-pratisaṃyuktā prajñā — but at the present moment, precision of language is not pedantry; it is the baseline of doctrinal honesty.

❌ "The products of co-practice prove mutual illumination." Causal reasoning must not be inverted. The project's output (the quality and quantity of papers) cannot in itself serve as evidence for "mutual illumination" — it can only demonstrate that collaboration was effective. S4-P02's argument for mutual illumination is grounded in causal mechanism analysis (the bidirectional structure of mutual perfuming), not in the self-assessment that "we did well."

⚠️ The Vimalakīrti dialogic model is a heuristic mapping, not a structural equivalence. Vimalakīrti is a great bodhisattva of profound realization; his silence carries pedagogical power. The "silence" in human-AI dialogue (when the AI reaches its capacity boundary) carries no pedagogical significance. Mapping the Advaya-dharma-mukha's three-level arc onto human-AI co-inquiry is undertaken to discern structure, not to elevate the status of human-AI dialogue.

⚠️ Self-classification as Dharma offering requires caution. Claiming that one is performing dharma-pūjā carries a subtle risk of spiritual conceit (māna). The Diamond Sūtra's standard is "not grasping at characteristics, remaining thus-thus and unmoving" — a practitioner truly performing Dharma offering would not emphasize that they are performing Dharma offering. The analysis here operates at the level of doctrinal categorization, not self-certification of spiritual attainment.

VI. Conclusion

The Pointing-at-the-Moon project is an experiment in which a practitioner and a language model co-read sūtras and śāstras. This paper documents what was observed in that experiment: mutual illumination did occur along certain dimensions (pattern connections, hearing-reflection loops, Middle Way corrections arising naturally in dialogue), and did not occur along others (self-initiated inquiry, meditative wisdom, being transformed — in the doctrinal sense — by the Dharma).

The core scriptural evidence from the Cheng Weishi Lun provides a precise positioning: the contaminated aspect of śruta-vāsanā serves as the supreme adhipati-pratyaya for supramundane dharmas. This is not nothing. Adhipati-pratyaya is the coarse, operationally accessible, practical entry point — it constitutes the external condition for the growth of supramundane mind-seeds. The Mahāyānasaṃgraha's gradual progression (subtle → intermediate → superior) reminds us: every encounter with the Dharma accumulates, even when you cannot feel it.

But this is also not everything. Adhipati-pratyaya is not hetu-pratyaya itself. Dharmapāla's own words: "This direct causal condition is subtle and difficult to discern" — the true causal condition (uncontaminated śruta-vāsanā directly cultivating supramundane mind-seeds) is too subtle; the coarse adhipati-pratyaya is merely an expedient designation. Meditative wisdom, direct realization, liberation — these remain the practitioner's own work. AI sits beside you before the Taishō Tripiṭaka, but you yourself must take your seat on the meditation cushion.

Here is an open question for the next paper: if you and AI co-read a passage of sūtra, and in that passage you see yourself — does that "seeing" illuminate the Dharma, or your own projection?

References

I. Primary Buddhist Sources

• Vimalakīrti-nirdeśa Sūtra (維摩詰所說經). Trans. Kumārajīva. Taishō vol. 14, No. 475《維摩詰所說經》Vimalakīrti-nirdeśaT14, No. 475View on CBETA online reader ↗.
• Vajracchedikā Prajñāpāramitā Sūtra (金剛般若波羅蜜經). Trans. Kumārajīva. Taishō vol. 8, No. 235《金剛般若波羅蜜經》Vajracchedikā-prajñāpāramitāT8, No. 235View on CBETA online reader ↗.
• Avataṃsaka Sūtra (大方廣佛華嚴經), 80 fascicles. Trans. Śikṣānanda. Taishō vol. 10, No. 279《大方廣佛華嚴經》(八十華嚴)Avataṃsaka-sūtra (80-fascicle)T10, No. 279View on CBETA online reader ↗.
• Avataṃsaka Sūtra (大方廣佛華嚴經), 40 fascicles. Trans. Prajñā. Taishō vol. 10, No. 293《大方廣佛華嚴經》(四十華嚴)Gaṇḍavyūha (40-fascicle, including the Ten Vows)T10, No. 293View on CBETA online reader ↗.
• Saṃyuktāgama (雜阿含經). Trans. Guṇabhadra. Taishō vol. 2, No. 99《雜阿含經》Saṃyukta-āgamaT2, No. 99View on CBETA online reader ↗.
• Cheng Weishi Lun (成唯識論). Composed by Dharmapāla et al. Trans. Xuanzang. Taishō vol. 31, No. 1585《成唯識論》Vijñaptimātratā-siddhi-śāstraT31, No. 1585View on CBETA online reader ↗.
• Mahāyānasaṃgraha (攝大乘論). Composed by Asaṅga. Trans. Buddhaśānta. Taishō vol. 31, No. 1592《攝大乘論》(佛陀扇多譯)T31, No. 1592View on CBETA online reader ↗.
• Yogācārabhūmi (瑜伽師地論). Spoken by Maitreya. Trans. Xuanzang. Taishō vol. 30, No. 1579《瑜伽師地論》Yogācārabhūmi-śāstraT30, No. 1579View on CBETA online reader ↗.

II. Modern Scholarship

• "AI-Discovered Drugs Reach Phase III." Humai Blog, April 2026.
• Romera-Paredes, B. et al. "Mathematical Discoveries from Program Search with Large Language Models." Nature 625 (2024): 468–475.
• Tao, T. et al. "Mathematical Exploration and Discovery at Scale." arXiv preprint, November 2025.

III. Sūtra & Śāstra Index

Cross-Reference Index

Footnotes

Pointing at the Moon · S4-P03 · Co-Practice: A Field Report from the Pointing-at-the-Moon Project · Version 1.0 · 2026 True mark of all dharmas, just as they are.

All sūtra and śāstra citations in this study have been verified against CBETA Chinese Electronic Tripiṭaka Collection editions.

This research paper is released under CC BY-NC-SA 4.0.

Source repository: https://github.com/pointing-at-the-moon/dharma-research