# YANNI CHEN > Yanni Chen has recorded dreams for 25 years and is building Morpheus Studio—exploring dream intelligence, consciousness, AI, and personal memory. Public Ghost content for AI and LLM tooling. This file includes a bounded export of public pages first, then recent public posts. Append `.md` to any post or page URL to get the content in Markdown (for example, `/example-post.md`). ## Pages ### About Yanni Chen URL: https://www.yannichen.org/about/ Last updated: 2026-09-14T05:36:50.000Z ![](https://storage.ghost.io/c/0a/74/0a74fc6e-ce75-4858-b413-21b0a154fbd2/content/images/2026/07/Untitled--1-.png) I'm Yanni Chen. I'm building the computational infrastructure for dreams. I believe dreams are one of the last unexplored frontiers of human intelligence. Every night, billions of people generate rich cognitive experiences that vanish within minutes of waking. For most of human history, they have remained private, fleeting, and computationally invisible. I don't think that has to remain true. Recent advances in artificial intelligence, neuroscience, wearable sensing, and large-scale personal data make it possible to approach dreams in an entirely new way—not as mystical symbols to decode, but as structured cognitive phenomena that can be recorded, modeled, and studied over time. That is the work I have chosen to dedicate my life to. --- ## How it started I've recorded my dreams continuously for the past twenty-five years. What began as a simple habit gradually became a long-term archive of my own mind—thousands of dream fragments, recurring symbols, emotional patterns, and questions that no existing tool could help me understand. As a teenager, I published a poetry collection based entirely on dreams. Years later, I entered the AI industry, where I witnessed firsthand how computational systems could reshape entire industries. Those two paths—dreams and AI—seemed unrelated for many years. Eventually I realized they had always been leading toward the same question: **Can dreams become computable?** That question became the foundation of everything I am building today. Not as a product idea, but as a new layer of understanding: Building the infrastructure for dream intelligence. Exploring the intersection of AI and human cognition. --- ## What I'm building My work is organized into three connected layers. ### Dream Intelligence A computational framework for understanding dreams as structured cognitive data rather than symbolic puzzles. Instead of asking *"What does this dream mean?"*, Dream Intelligence asks a different question: **What patterns emerge across hundreds of dreams over months and years?** Meaning is treated as something that emerges from longitudinal observation, personal context, and computational analysis—not from universal symbol dictionaries. --- ### NeuroDream The infrastructure layer. NeuroDream transforms nightly dream experiences into private, structured, longitudinal cognitive data while preserving uncertainty, individual context, and user ownership. It is designed as foundational infrastructure for future research, AI applications, and personal cognitive tools. --- ### Morpheus Studio The first application built on NeuroDream. Morpheus Studio is a Dream OS where people can capture dreams immediately after waking, organize years of dream history, converse with their archive, visualize recurring themes, generate creative work, and observe how their inner world evolves over time. It does not tell people what their dreams mean. It helps them discover how meaning gradually emerges from their own lives. --- ## Why now Dream research has existed for more than a century, yet many fundamental questions remain unanswered. Rather than waiting for neuroscience to fully explain dreaming, I believe we can begin building useful computational systems today. For the first time, several technologies are converging: - AI capable of modeling complex language and personal data - wearable devices measuring sleep continuously - multimodal interfaces for recording experiences naturally - long-term personal data that can reveal cognitive patterns impossible to observe in isolated dreams Together, they create an opportunity that simply did not exist before. --- ## Long-term vision My long-term goal is to make dreams a first-class object of computation. I believe dreams should become observable, searchable, modelable, and computationally meaningful—while remaining deeply personal and fully owned by the people who experience them. In the future, dream data may contribute to cognitive science, creativity, mental well-being, personalized AI, and our broader understanding of human intelligence. Whether that future arrives in five years or twenty, I intend to spend my career helping build it. --- ## Collaboration I'm always interested in conversations with researchers, neuroscientists, sleep scientists, AI engineers, designers, founders, universities, healthcare organizations, and others exploring the future of human cognition. If our work overlaps, I'd be happy to connect. The fastest way to reach me is: [**neurodream.app@gmail.com**](mailto:neurodream.app@gmail.com) --- Every night we generate experiences that disappear before breakfast. I don't believe they should disappear. I believe dreams deserve memory. They deserve structure. They deserve computation. And I believe understanding dreams may become one of the next great frontiers of intelligence. --- ### Continue exploring - **The framework** → [Dream Intelligence](https://www.yannichen.org/dream-intelligence/) - **The flagship essay** →[ Dream Intelligence: the essay](https://www.yannichen.org/essays/) - **The infrastructure** → [NeuroDream](https://www.yannichen.org/neurodream/) - **The product** → [Morpheus Studio](https://www.yannichen.org/morpheus-studio/) - **More writing** → [Essays](https://www.yannichen.org/essays/) · [Research](https://www.yannichen.org/research/) ### Research URL: https://www.yannichen.org/research/ Last updated: 2026-06-28T16:04:42.000Z Dream research is ancient and, at the same time, genuinely frontier science — we still cannot fully say what the brain is doing when we dream. I treat that honestly: the work below is a set of open questions I'm pursuing through NeuroDream, not a body of settled findings. Each theme connects to the six-layer infrastructure that makes systematic study possible in the first place. ### Dream Intelligence Understanding dreams as a form of cognitive computation rather than messages to decode. The central methodological commitment is the **individual baseline**: meaning is read as deviation from a person's own established pattern, never against a universal symbol dictionary. Open questions: how many dreams are needed for a stable baseline, and which signals are robust enough to be worth surfacing at all. ### Computational Dreaming Can dreams be modeled, analyzed, and — eventually — generated through AI? This spans the practical (NLP coding of dream reports into Hall–Van de Castle dimensions; longitudinal pattern modeling) and the speculative (what it would mean for a generative system to produce, not just analyze, dream-like cognition). The honest boundary: AI's understanding of dreams has real limits, and overclaiming here is the fastest way to lose credibility. ### Consciousness What can dreaming teach us about subjective experience? Dreams are a nightly, universal instance of consciousness decoupled from external input — arguably the most accessible natural laboratory for studying the felt interior. The questions here are the hardest and least resolved, and I approach them as a builder gathering data, not as someone claiming to have solved the hard problem. ### Sleep & Brain Science Exploring the relationship between sleep, memory, and imagination — memory consolidation, emotional regulation, REM dynamics, and the capture conditions (the 2–5 minute golden window after waking) that determine whether any of it is recoverable at all. This is where I lean most on dialogue with neuroscientists and sleep researchers, and where the science is moving fastest. ### Dream Data Infrastructure Building the tools and standards for large-scale dream research: structured representation, privacy-first and individually-owned data, confidence labeling, and shared formats. This is the layer almost no one is building, and the one I think matters most — because every other question above is bottlenecked on whether the data can be captured, structured, and trusted. --- *If you research dreams, sleep, consciousness, or the infrastructure around them — academically or in industry — I'd like to talk.* [*Reach out*](https://www.yannichen.org/about/)*.* --- ### Continue exploring - **The framework** → [Dream Intelligence](https://www.yannichen.org/dream-intelligence/) - **The flagship essay** → [Dream Intelligence: the essay](https://www.yannichen.org/essays/) - **The infrastructure** → [NeuroDream](https://www.yannichen.org/neurodream/) - **The product** → [Morpheus Studio](https://www.yannichen.org/morpheus-studio/) - **More writing** → [Essays](https://www.yannichen.org/essays/) - **The person** → [About](https://www.yannichen.org/about/) ### NeuroDream URL: https://www.yannichen.org/neurodream/ Last updated: 2026-06-28T16:02:14.000Z # NeuroDream is building the infrastructure for Dream Intelligence. We believe dreams are one of the largest unexplored sources of human experience. Every night, people generate stories, emotions, symbols and insights that are rarely captured, understood or connected to their daily lives. Our mission is to transform dreams from fleeting experiences into meaningful personal knowledge. Starting with Morpheus Studio, we are creating a new generation of tools that help people record, understand and create from their dreams while maintaining complete ownership and privacy of their data. Based in Hong Kong and serving a global audience, NeuroDream sits at the intersection of artificial intelligence, sleep science, creativity and human self-discovery. Over time, we aim to connect dream experiences with sleep health, physiological signals and cognitive wellbeing, creating a bridge between the inner world and measurable reality. --- ### Continue exploring - **The framework** → [Dream Intelligence](https://www.yannichen.org/dream-intelligence/) - **The flagship essay** → [Dream Intelligence: the essay](https://www.yannichen.org/essays/) - **The product** → [Morpheus Studio](https://www.yannichen.org/morpheus-studio/) - **More writing** → [Essays](https://www.yannichen.org/essays/) · [Research](https://www.yannichen.org/research/) - **The person** → [About](https://www.yannichen.org/about/) *\[*[*Subscribe*](https://www.yannichen.org/)*\] — follow the work.* ### Morpheus Studio URL: https://www.yannichen.org/morpheus-studio/ Last updated: 2026-09-12T15:00:05.000Z # A Dream OS for the Age of AI. [Explore Morpheus Studio →](https://morpheusstudio.ai/?ref=yannichen.org) Most dreams disappear within minutes after waking. Morpheus Studio helps you capture what would otherwise be lost. Using voice or text, you can quickly record dream fragments, emotions and impressions before they fade. Morpheus then helps organize these experiences into a living archive that grows with you over time. Unlike traditional dream interpretation tools, Morpheus does not attempt to tell you what your dream means. Instead, it helps you observe patterns, revisit memories and discover connections between your dreams and waking life. ## Four Core Experiences ### Record Capture dreams through voice or text in seconds. ### Dialogue Explore your dreams through reflective conversations with Morpheus. ### Insight Identify recurring themes, emotions, people and symbols across time. ### Create Transform dreams into stories, poetry, visual art, scripts and creative projects. Over time, your dream archive becomes a personal map of memory, emotion and transformation. [Start Your Dream Archive →](https://morpheusstudio.ai/?ref=yannichen.org) ### Dream Intelligence URL: https://www.yannichen.org/dream-intelligence/ Last updated: 2026-09-14T05:31:13.000Z ## Why the next frontier of intelligence may be the one we visit every night *By Yanni Chen* We have learned to read the genome, photograph black holes, and build machines that pass for human in conversation. Yet there is a territory each of us enters every single night — for a cumulative six years over an average lifetime — that remains almost entirely unmapped. We have no instruments pointed at it, no shared vocabulary to describe it, and no systematic way to learn from it. I am talking about **dreams**. I have been recording my own for twenty-five years — since I was sixteen, when I sat up one morning and wrote down a dream before it could dissolve. I have not stopped since. Across those years the same images keep returning: deep water, the sea, certain landscapes that come back like rooms I know but cannot name. For most of that time I assumed the records were a private habit, maybe a writer's quirk. I now think they were field notes for a question I couldn't yet phrase. This essay makes a single argument: **dreams are not noise to be discarded but data to be understood — a form of intelligence the brain runs while we sleep.** I call this idea *Dream Intelligence*. The pages below lay out what it means, why the claim is defensible rather than mystical, and what it takes to actually build the infrastructure that would let us study it. #### The mistake we keep making For most of recorded history, dreams have been treated as messages — from gods, from the unconscious, from fate. The interpreter's job was to *decode* them: this symbol means death, that one means desire. Twentieth-century psychology inherited the same instinct and dressed it in clinical language, but the underlying move never changed. A dream was a riddle with a hidden answer, and someone with the right key could tell you what it *really* meant. I think this framing is the single biggest reason dreams have stalled as a serious object of study. It assumes meaning is *buried inside* a single dream, waiting to be extracted, and that the same symbol means the same thing for everyone. Both assumptions are almost certainly wrong. Water does not mean one thing. It means something in the context of *your* dreams, *your* baseline, *your* life — and only the accumulation reveals it. So Dream Intelligence makes a different move. It does not decode the single dream. It asks what a particular mind does *across* hundreds of dreams, and what changes when something in that person's life changes. The shift sounds small. It is not. It is the difference between astrology and astronomy — between reading one sky for omens and recording many skies until the structure underneath reveals itself. #### What the sleeping brain is doing During sleep the brain does not switch off. It runs a different mode of computation — recombining memory, rehearsing emotion, and generating scenarios largely free of external constraint. Dreams are the subjective trace of that process. They are what distributed cognitive processing *feels like from the inside*. The honest scientific position is that we do not yet know, definitively, *why* we dream. Over the past year I have spent hundreds of hours in conversation with neuroscientists and sleep researchers in China and the United States, and the most useful thing I learned is how much remains genuinely open. What the field does offer is several credible, partially supported frameworks: - the **continuity hypothesis** — dreams reflect and extend our waking concerns; - **threat-simulation theory** — dreams rehearse danger in a safe sandbox; - **memory consolidation** — sleep replays and reorganizes the day's experience, and dreams are part of that reshuffling; - **emotional regulation** — the sleeping brain processes and defuses difficult affect. These are not rivals to be resolved by picking one. They are lenses. A serious system should hold all of them at once, label which lens it is using, and never pretend a single theory is the answer. That commitment — to multiple frameworks and explicit uncertainty — turns out to be not just intellectually honest but architecturally important, as I'll explain. #### You can't lie to a dream There's a practical reason this matters beyond science. We spent centuries building an infrastructure for the health of the body — hospitals, labs, imaging, the annual checkup. The whole point is to let data speak *before* symptoms appear. We built almost nothing equivalent for the interior life: motivations, fears, the recurring obsessions and longings we can't quite account for. Therapy, meditation, and journaling all exist, but they demand active effort, language, and sustained self-awareness most people can't maintain. Dreams are the one stream of interior data we generate effortlessly, every night, with no persona and no performance. You can't lie to a dream, because lying requires a self that's on guard, and in dreams that self is undefended. The catch is purely mechanical: the data evaporates within minutes of waking, and almost no one captures it. Solve capture, and an entire interior becomes observable for the first time. #### The infrastructure: six layers from capture to humility A framework that stays a framework is just another manifesto. Dream Intelligence only matters if it can be built — which means being precise about how a fleeting experience becomes something analyzable without overclaiming what the analysis means. The system I am building, NeuroDream, is organized as six layers, from raw capture at the bottom to deliberately restrained output at the top. **Layer 0 — Capture Quality.** Everything rests here. Dream memory decays fast, so the golden window is the two-to-five minutes after waking; voice beats text because it's faster than the forgetting. This layer also records the metadata that makes a dream interpretable later — approximate sleep cycle, emotional state on waking, an estimate of REM position. Garbage capture means garbage everything above it. **Layer 1 — Content Representation.** A captured dream is decomposed into structure using the Hall–Van de Castle system, the most established content-coding framework in dream research: characters, settings, emotions, activities, objects — plus modern semantic vectors. This is what turns a story into something you can count and compare. **Layer 2 — Individual Baseline (the scientific core).** This is the heart of the whole design, and the part most consumer dream apps skip. The system spends your first \~30 dreams building *your* baseline — your characteristic emotional frequencies, who tends to appear, how your narratives are typically structured — and outputs **no insights at all** during that window. There is no universal symbol dictionary. A motif only becomes meaningful as a deviation from your own established pattern. Refusing to interpret early is not a limitation; it's what makes anything said later credible. **Layer 3 — Longitudinal Pattern.** Once a baseline exists, the unit of analysis becomes the *dream life* rather than the dream: emotional-processing trajectories over months, shifts in threat-simulation frequency, the degree of continuity between dreams and waking life. Anomalies are meaningful here precisely because there's a baseline to deviate from. **Layer 4 — Theoretical Lens.** Patterns are read through the multiple frameworks named above — continuity, threat simulation, memory consolidation, emotional regulation — with the source theory always labeled. The system tells you which lens it's looking through, never collapsing into a single confident story. **Layer 5 — Epistemic Humility.** The top layer is the one I care about most. Every output carries a confidence score, is labeled as hypothesis rather than fact, and is framed as an invitation to co-construct meaning rather than a verdict handed down. The system's most important capability is knowing — and saying — what it doesn't know. Read bottom to top, the architecture encodes a single discipline: capture honestly, structure rigorously, establish the individual baseline before saying anything, read through multiple theories, and label uncertainty at every step. **Morpheus Studio** is the first product built on this stack — record by voice or text on waking, watch patterns surface over time, converse with your own history, and create from the material. It deliberately never tells you what a dream *means*. It helps you observe how meaning emerges across your dream life — which is the whole point of the shift from decoding to intelligence. #### What this enables Take the stack seriously and several things become possible that are not today. A **new dataset for cognition** — self-reported, longitudinal, emotionally rich, generated nightly, and, treated ethically, baselined per individual rather than flattened into population averages. A **mirror for emotional and psychological patterns**, surfacing shifts a person would never notice one dream at a time — not as diagnosis, but as a reflective signal worth attention. A **source of creative material**, since the dreaming brain is a story generator that ignores the rules. And a **bridge between neuroscience and AI**: generative models and dreaming brains are both, loosely but really, systems that produce novel output by recombining learned representations. The dreaming brain is the oldest generative model we have; we have simply never had the tools to read its output. #### What I am not claiming Intellectual honesty requires saying what this is not — and in this system, that honesty is built into Layer 5, not bolted on afterward. I am not claiming dreams are prophecy, or that any symbol carries a fixed universal meaning, or that we can currently read dreams off a brain scan, or that the brain's nightly computation is understood — it isn't, and much of what I've described remains live scientific debate. The four frameworks are the most credible current lenses, not settled fact. What I *am* claiming is narrower and harder to dismiss: that dreams are structured products of cognitive work, that this work can be captured and studied systematically — relative to each person's own baseline — rather than only interpreted symbolically, and that doing so opens a frontier we have barely touched. The honest position is not certainty. It is that the question deserves instruments, and almost no one has built them. #### The frontier within We have spent centuries pointing our best tools outward — at cells, at galaxies, at the structure of computation itself. The strange thing is how little of that effort has turned inward, toward the hours we spend every night generating worlds. Twenty-five years in, I still don't know what the deep water in my dreams means. But I no longer think that's the right question, or that I'm asking it alone. The right question is what my dreams, taken together and over time, are doing — and whether we can finally build the instruments to find out. I suspect the next real leap in understanding intelligence won't come only from larger models or deeper telescopes. Some of it will come from finally taking seriously the oldest, strangest, most universal cognitive event there is — the one waiting for us, unmapped, every time we close our eyes. That is the frontier I am trying to build the map for. I write regularly about dream intelligence, cognition, and the systems we're building at NeuroDream. \[[Subscribe](https://www.yannichen.org/)\] to follow the work, or \[[Contact Me](https://mail.google.com/mail/neurodream.app@gmail.com?ref=yannichen.org)\] if you'd like to collaborate. --- ### Continue exploring - **The framework** → [Dream Intelligence](https://www.yannichen.org/dream-intelligence/) - **The infrastructure** → [NeuroDream](https://www.yannichen.org/neurodream/) - **The product** → [Morpheus Studio](https://www.yannichen.org/morpheus-studio/) - **More writing** → [Essays](https://www.yannichen.org/essays/) · [Research](https://www.yannichen.org/research/) - **The person** → [About](https://www.yannichen.org/about/) ### Essays URL: https://www.yannichen.org/essays/ Last updated: 2026-09-12T14:44:07.000Z **Exploring dreams through AI, neuroscience, and computational thinking.** Writing about dreams as experience, data, and a window into how the mind works. --- ## Featured ### [**Dream Intelligence — Why the next frontier of intelligence may be the one we visit every night**](https://www.yannichen.org/dream-intelligence/) An introduction to the idea of Dream Intelligence: why dreams deserve to be studied more seriously, what AI makes possible, and how we might begin building practical tools around them. *Start here.* --- ## Dream Intelligence White Paper Three published papers introducing the core ideas behind Dream Intelligence. Together they outline the motivation, computational framework, and early system architecture for understanding dreams in a more structured way. ### [【WP】Part I — Manifesto：Why Dreams Matter](https://www.yannichen.org/manifesto/) Why dreams should be treated as meaningful cognitive data rather than random by-products of sleep. ### [【WP】Part II — A Computational Framework for Dreams](https://www.yannichen.org/part-ii-a-computational-framework-for-dreams/) A framework for representing, analyzing, and generating dream experiences with AI. ### [【WP】Part III — The Dream Intelligence Framework](https://www.yannichen.org/part-iii-the-dream-intelligence-framework/) Connecting dream recording, semantic modeling, AI, and future applications into one evolving system. --- ## Founder Essays Personal essays from the journey of building NeuroDream. Some begin with a question. Others begin with a dream that stayed with me for years. Together they document how a personal habit gradually became a research project and a company. ### 【FE】Why I've Been Recording My Dreams for Twenty-Five Years How a simple daily practice became the foundation for everything I'm building today. ### 【FE】The Words Your Waking Mind Never Learned What happens when dreams produce words you've never consciously learned—and what science can and cannot explain. ### 【FE】We Built Hospitals for Our Bodies. What Did We Build for Our Minds? Why our inner experiences remain largely invisible, and why dreams may offer one of the most valuable records of mental life. --- ## Coming Next The questions I'm exploring next. - Why Dreams May Be Humanity's Largest Untapped Dataset - Can AI Understand Dreams? - Dreams and the Future of Consciousness Research - Why Neuroscience Still Cannot Explain Dreams - The Missing Infrastructure of Dream Science - Building a Dream Dataset for the AI Era --- **Follow the journey.** *\[*[*Subscribe*](https://www.yannichen.org/)*\]* for new essays, white papers, and product updates as the work evolves. --- ### Continue exploring - **The framework** → [Dream Intelligence](https://www.yannichen.org/dream-intelligence/) - **The infrastructure** → [NeuroDream](https://www.yannichen.org/neurodream/) - **The product** → [Morpheus Studio](https://www.yannichen.org/morpheus-studio/) - **Research** → [Research](https://www.yannichen.org/research/) - **The person** → [About](https://www.yannichen.org/about/) ### Subscribe to Dream Intelligence URL: https://www.yannichen.org/subscribe/ Last updated: 2026-08-12T09:32:35.000Z One carefully researched letter every two weeks on dreams, consciousness, AI, and long-term personal memory. I’m Yanni Chen. I’ve documented my dreams for 24 years, and I’m building Morpheus Studio and the infrastructure for dream intelligence. What you’ll receive One substantial essay or field note every two weeks. Research notes on dream science, AI, computational cognition, privacy, and personal memory. Behind-the-scenes experiments and honest questions from building Morpheus Studio. This is not generic AI news, automated dream interpretation, or daily inbox noise. Join Dream Intelligence Free. Unsubscribe anytime. ## Posts ### What Question Is Worth a Life? URL: https://www.yannichen.org/what-question-is-worth-a-life/ Last updated: 2026-09-14T05:39:19.000Z *On finite lives, abundant answers, and the question behind Morpheus Studio.* “Heard It in the Morning” (《朝闻道》) is a short story by the Chinese science fiction writer Liu Cixin, who is best known internationally for his novel *The Three-Body Problem*. The story poses a stark question: What would you give to know the truth? Humanity stands on the verge of probing one of the universe’s deepest secrets with a particle accelerator encircling the Earth. Then a civilization far more advanced than our own intervenes and halts the experiment. The scientists are offered a final bargain: they may receive the answers they have spent their lives seeking, but they must die immediately afterward. They will finally understand—but there will be no tomorrow in which to live with that understanding.[\[1\]](#fn1) The scientists accept. The story is a modern interpretation of a line attributed to Confucius, the ancient Chinese philosopher: “If one heard the Way in the morning, one could die in the evening without regret.” There is no altar of truth in ordinary life. There is no all-knowing being waiting for us to ask the perfect question. Yet the exchange is not entirely fictional. To study one thing is to leave countless other things unstudied. To build one future is to refuse others. We may spend decades on a problem and never reach an answer. We may reach one, only to discover that the problem mattered less than we believed. Perhaps the deepest risk is not failing to find an answer. It is spending a life on the wrong question. Artificial intelligence makes that risk more visible. We are building systems that can generate hypotheses, search for proofs, design experiments, retain memory, and detect structures that human beings might never find alone. The questions may still come from us, even as the route to the answers increasingly does not. So the ancient line leaves us with a contemporary problem: **When life is finite and the unknown is effectively infinite, what question is worth a life?** ## When answers become abundant For much of intellectual history, the obvious bottleneck was a shortage of answers. AI is changing that premise. AlphaFold transformed protein-structure prediction. Formal systems now search for and verify mathematical proofs at levels once thought far beyond machine reasoning. Multi-agent systems can propose scientific hypotheses, debate and rank them, then send selected candidates toward laboratory validation.[\[2\]](#fn2) This changes the price of producing possible answers. It does not automatically tell us which answers deserve to exist. Mathematics makes the shift unusually clear. Terence Tao, the Australian-American mathematician and 2006 Fields Medalist, has described a movement from proof scarcity toward proof abundance. If machines can produce proofs at scale, verification, explanation, and conceptual judgment become scarce. A formally correct proof is not automatically meaningful mathematics; someone must still understand the hard step, connect it to existing ideas, and decide why it matters.[\[3\]](#fn3) The same tension appears across science. A 2026 *Nature* study analyzed 41.3 million papers. AI-assisted researchers were associated with higher individual output and impact, while the collective range of scientific topics studied contracted by about 4.63%. The study is observational, not a simple causal verdict. It nevertheless points to a structural danger: tools that accelerate individual work may also direct collective attention toward questions that are easiest to evaluate and richest in existing data.[\[4\]](#fn4) These examples point to the same asymmetry. The most machine-legible goals are the easiest to optimize: a proof can be checked and a protein structure can be compared with experimental data. Whether a theory deepens human understanding, a technology supports dignity, or a question deserves decades of a life is harder to compress into a reward function. If answers become abundant, judgment does not become less important. It becomes the scarce resource. And this is not only a problem for science. It becomes personal when AI is asked not merely to solve a theorem, but to interpret a life. ## The question that kept returning to me I began recording dreams deliberately when I was sixteen. At seventeen, I turned some of those dreams into a poetry collection. Years later, when I could not find images that carried their atmosphere, I learned to paint. The medium kept changing because I was not looking only for an interpretation; I was looking for forms capable of holding the experience. The question itself did not change. Across twenty-five years, certain places, people, emotions, and situations returned. Some dreams seemed to move toward a judgment that my conscious mind had not yet formed. I do not treat them as prophecy, and I cannot assume their direction was correct. What holds my attention is the process itself: while waking thought was still gathering reasons, another mode of cognition had already begun arranging memory, emotion, bodily signals, people, and possible futures into a scene. The question that kept returning was this: **Do dreams—and if so, how—participate in organizing memory, emotion, and experience before the waking mind has reached a clear conclusion?** There is no single scientific answer today. Research connects dreaming with memory processing, emotional experience, simulation, and several other functions, but the evidence does not support one final theory of why we dream or what any individual dream means. That uncertainty matters. It is a reason for more careful observation, not a license to replace missing evidence with certainty. Dreams are among the most intimate experiences we have, yet they are also among the least preserved. Many fade within minutes. Even when recorded, they often remain isolated fragments—a place, a face, an action, a feeling whose significance may only become visible months or years later. This is where a personal practice becomes a research and design problem. How can years of dream records be brought into one usable memory without losing their original context? How can we preserve a dream without pretending to possess its final meaning? How can we trace patterns across time without reducing a person to a dictionary of symbols? How can an AI ask questions that deepen recall without planting an interpretation? How can private inner material remain under the dreamer’s control? These questions led me to build Morpheus Studio. They also shape *Dream Intelligence*, the biweekly field notes in which I share the research, design choices, and unresolved questions behind the work. **[Subscribe free →](https://www.yannichen.org/#/portal/signup)** ## Why dreams need more than an interpreter Most dream products begin with an immediate transaction: record a dream, receive an interpretation. That can be engaging, but it is not enough for the question I am pursuing. An isolated dream rarely contains its own context. Its significance may depend on a person’s history, recurring places, changing relationships, recent events, emotional contradictions, and the dreams that do *not* fit the apparent pattern. A long-term dream system therefore needs more than a powerful model. It needs memory, provenance, correction, uncertainty, and restraint. If a system claims to notice a pattern in someone’s dreams, the dreamer should be able to ask: - Which original dreams support this pattern? - Which dreams contradict it? - Did the system merge two people or places that should remain separate? - Can I correct or reject what it remembers? - How has this pattern changed over time? - If I delete a source dream, do the summaries, tags, embeddings, and later memories derived from it disappear too? Without answers to those questions, AI simply layers opaque judgment over personal history. This principle extends beyond dreams. Once AI systems keep state, call tools, and act across time, the model alone is no longer the whole product. Memory, permissions, audit records, approval boundaries, portability, and recovery procedures determine what the system can become—and whom it ultimately serves.[\[5\]](#fn5) For intimate personal data, capability, control, and responsibility must move together. ## What Morpheus Studio is trying to build Morpheus Studio is being designed as a private, longitudinal environment for bringing together an existing dream archive, recording what comes next, tracing change across time, and turning fragments into reflection and creative work. The broader product connects four kinds of experience: - **Preserve:** import an existing archive and capture new dreams through text or voice, with imagery and tags that retain atmosphere and context; - **Examine:** use a source-linked Dream Atlas to revisit recurring people, places, emotions, themes, changes, and contradictions across time; - **Converse:** ask questions, recover detail, and reflect without allowing AI to impose a final interpretation; - **Create:** transform dream material into images, narratives, and a personal multimodal portfolio. Together they form a path from preservation to examination, reflection, and expression. But the feature list is not the central promise. Almost any small team can reproduce a recorder, an image generator, a set of tags, or an AI chat interface. The first pilot therefore begins with the narrowest and most demanding value loop: bring in a real archive, inspect what the system claims to see, trace every observation back to its source, correct the memory, and then add the next dream to see what changes. The deeper direction is a traceable, user-governed form of long-term dream memory: - patterns remain linked to the dreams that support them; - uncertainty and counterexamples remain visible; - people, places, and motifs can evolve rather than being frozen into permanent labels; - the dreamer can correct, reject, merge, export, or remove what the system remembers; - deletion propagates through derived memory rather than leaving an invisible remainder. Interpretation should remain a conversation, never a verdict. AI can serve as a reflective and creative instrument; the dreamer retains authority over the experience. These are design commitments, not claims that every technical and scientific question has already been solved. They are precisely what must be tested through real use. Morpheus is not an answer to the question of what a life is for. It is one concrete way I have chosen to pursue mine. ## An invitation to the first 100 Founding Dreamers Benchmarks cannot determine whether a system like this deserves a person’s trust. It has to meet real archives: inconsistent formats, missing dates, repeated names, changing relationships, apparent patterns, inconvenient counterexamples, and material that must remain private. That is why I am inviting the first **100 Founding Dreamers** to help shape Morpheus Studio at its earliest stage. The current Archive Pilot begins with the history its participants have already kept. **Bring the archive. Examine the sources. Correct the memory.** The first 100 is the upper limit of the Founding Dreamers program, not a single open-enrollment cohort. Participants will be selected in small, manually reviewed waves so that the import process, privacy boundaries, and feedback can receive close attention. The pilot is for people who already have roughly **30 or more dream records, or an archive spanning at least one year**. At least part of the archive should already be digital—or be possible to prepare in a supported text format. Participants should still be recording dreams and be curious about what a longitudinal view might reveal. The process is deliberately simple: - apply with information about the archive’s size, time span, and format—without submitting any dream content; - if selected, agree on a supported private import path and the relevant privacy boundaries before sharing the archive; - review a source-linked Dream Atlas that shows what the system noticed and which original dreams support—or complicate—each observation; - confirm, correct, reject, or merge what Morpheus remembers; - add the next dream and see whether it deepens, changes, or challenges the emerging picture. After real use, selected participants may also be invited to a short conversation about where the system earned trust, where its evidence was weak, and what they chose to correct. We are not asking whether people merely “like the app.” We are testing whether a personal dream memory can remain useful, inspectable, and answerable to the person whose history it holds. The application asks for no dream content or upload. Before any private import, we will explain what is collected, why it is needed, how it is used, and how deletion can be requested. Morpheus is not a diagnostic service, a substitute for professional mental-health care, or an authority on what a dream “really means.” This cohort is for reflective practice and collaborative product research. The pilot is global and will operate primarily in English under the service conditions Morpheus can currently support. I am looking first for long-term dream recorders: people who want to understand change across years rather than receive an instant universal interpretation; who are willing to examine evidence and correct the system; and who believe intimate data should be handled with clarity, restraint, and respect. If you do not yet have a substantial archive but want to build a dream-recording practice, this is not the right pilot—and that is intentional. A later cohort will focus on capture, recall, and habit formation. You can subscribe below to follow that work. The first 100 participants will not be an audience around Morpheus. They will be part of its origin. The altar in Liu Cixin’s story does not exist. Still, we vote for our deepest questions every day—with attention, work, courage, and time we cannot recover. Morpheus Studio is one of my votes. If dreams are part of the question you have been carrying too, I invite you to become one of the first 100 Founding Dreamers. **[Apply to the Founding Dreamers Archive Pilot →](https://founding.morpheusstudio.ai/?utm%5Fsource=yannichen%5Forg&utm%5Fmedium=owned%5Feditorial&utm%5Fcampaign=archive%5Fpilot%5Ffirst%5F100&utm%5Fcontent=what%5Fquestion%5Fis%5Fworth%5Fa%5Flife#apply)** *Applications take about three minutes. We review them personally. No dream content or upload is requested in the application.* Not ready to apply? [Subscribe to Dream Intelligence](https://www.yannichen.org/#/portal/signup) and follow the research as it develops. --- ### Notes and further reading --- 1. Liu Cixin, [*A View from the Stars: Stories and Essays*](https://us.macmillan.com/books/9781250292124/aviewfromthestars/?ref=yannichen.org), including “Heard It in the Morning,” translated by Jesse Field (Tor Books, 2024). [↩︎](#fnref1) 2. [The Nobel Prize in Chemistry 2024](https://www.nobelprize.org/prizes/chemistry/2024/press-release/?ref=yannichen.org); [“Olympiad-level formal mathematical reasoning with reinforcement learning”](https://www.nature.com/articles/s41586-025-09833-y?ref=yannichen.org), *Nature*; [“Accelerating scientific discovery with Co-Scientist”](https://www.nature.com/articles/s41586-026-10644-y?ref=yannichen.org), *Nature*. [↩︎](#fnref2) 3. Terence Tao and Tanya Klowden, [“Mathematical Methods and Human Thought in the Age of AI”](https://arxiv.org/abs/2603.26524?ref=yannichen.org); [“The job description is changing: Terence Tao on how AI is reshaping mathematics”](https://www.nature.com/articles/d41586-026-01246-9?ref=yannichen.org), *Nature*; [The Leiden Declaration on AI and Mathematics](https://leidendeclaration.ai/?ref=yannichen.org). [↩︎](#fnref3) 4. [“Artificial intelligence tools expand scientists’ impact but contract science’s focus”](https://www.nature.com/articles/s41586-025-09922-y?ref=yannichen.org), *Nature* (2026). [↩︎](#fnref4) 5. [*Engineering and Governing the Agent Harness: A Technology and Policy Framework for the Runtime Layer of Agentic AI*](https://unu.edu/publication/engineering-and-governing-agent-harness-technology-and-policy-framework-runtime-layer?ref=yannichen.org), United Nations University (2026). [↩︎](#fnref5) ### Part III The Dream Intelligence Framework URL: https://www.yannichen.org/part-iii-the-dream-intelligence-framework/ Last updated: 2026-09-12T14:36:37.000Z # ## From Fleeting Dreams to Lifelong Intelligence Dreams disappear quickly. Within minutes of waking, much of a dream has already faded. By the end of the day, only fragments often remain. Across an entire lifetime, countless dreams are experienced, yet almost none become part of our long-term understanding of ourselves. This raises a fundamental question. What if dreams were not temporary experiences, but information that could accumulate into intelligence? The Dream Intelligence Framework is an attempt to answer that question. Rather than treating dreams as isolated stories, the framework views every dream as a multimodal cognitive signal that can be captured, structured, understood, and continuously connected with future experiences. Its objective is simple: > **Transform fleeting dream experiences into lifelong cognitive intelligence.** --- ![](https://storage.ghost.io/c/0a/74/0a74fc6e-ce75-4858-b413-21b0a154fbd2/content/images/2026/06/ChatGPT-Image-2026---6---30----21_57_46.png) # Dreams as Multimodal Data A dream is never just a story. Every dream simultaneously contains multiple forms of information. - Narrative sequences - Visual imagery - Emotional dynamics - Symbolic relationships - Physiological signals - Cognitive processes These dimensions together form what we call **Multimodal Dream Data**. Instead of interpreting only one dimension, the Dream Intelligence Framework preserves all of them simultaneously, allowing artificial intelligence to analyze dreams from multiple complementary perspectives. This representation creates a richer and more faithful computational model of human dreaming. --- # The Dream Intelligence Loop The framework is organized as a continuous learning cycle rather than a linear pipeline. Each new dream enriches the entire system. ## 1\. Capture Dreams are recorded immediately after waking through text, voice, sketches, or future brain-computer interfaces. At the same time, physiological information from wearable devices—including sleep stages, heart rate variability, respiration, and brain activity—provides objective context. Together they preserve both subjective experience and biological state. --- ## 2\. Structure Raw memories become structured representations. Large language models identify events, characters, locations, timelines, and semantic relationships. Visual models organize imagery. Knowledge representations convert dreams into machine-readable cognitive objects. --- ## 3\. Understand Artificial intelligence searches for recurring themes, symbolic structures, emotional trajectories, and behavioral patterns. The purpose is not to assign universal meanings, but to reveal patterns that emerge uniquely within each individual. --- ## 4\. Generate Dreams become generative media. Images, animations, cinematic scenes, interactive simulations, and future immersive environments allow users to revisit and expand dream experiences. Generation becomes a new method of reflection rather than entertainment alone. --- ## 5\. Reflect Users engage in dialogue with their dreams. Conversational AI helps organize thoughts, compare experiences across time, and encourage self-observation. Reflection transforms isolated memories into personal understanding. --- ## 6\. Improve Sleep Dreams should ultimately contribute to healthier sleep. Patterns discovered through Dream Intelligence may help individuals recognize stress, environmental influences, emotional overload, or lifestyle factors that influence dream quality and sleep architecture. Dream analysis therefore becomes one component of long-term sleep optimization. --- ## 7\. Improve Cognition Dreams are deeply connected with creativity, memory consolidation, emotional regulation, and problem solving. By observing dream evolution over months and years, individuals may better understand how their cognitive system changes over time. Dream Intelligence becomes a tool for lifelong cognitive development. --- ## 8\. Long-Term Memory Every dream contributes to an evolving cognitive memory system. Rather than storing isolated journal entries, the framework builds a personal knowledge graph connecting people, locations, symbols, emotions, and recurring experiences. Over years, dreams become a searchable cognitive archive. --- ## 9\. Share & Collaborate Dream research advances through collaboration. The framework allows individuals to contribute anonymized data for scientific discovery while maintaining full ownership and privacy of personal dream records. Artificial intelligence, researchers, clinicians, and communities become collaborators rather than authorities. --- ## 10\. Collective Dream Intelligence When millions of individuals contribute structured dream data, entirely new scientific questions become possible. How do dreams evolve across cultures? How do global events influence collective dreaming? Can shared symbolic structures reveal previously invisible properties of human cognition? Dream Intelligence ultimately extends beyond personal insight toward a collaborative understanding of the human mind. --- # Design Principles The Dream Intelligence Framework is built upon five guiding principles. **Privacy First** Dreams represent one of the most personal forms of human data. Users retain ownership and control of every record. **Human-Centered AI** Artificial intelligence assists reflection rather than replacing human interpretation. **Multimodal Intelligence** Dreams should be understood through language, vision, physiology, and cognition simultaneously. **Long-Term Perspective** The value of dreams emerges across months and years, not through isolated interpretations. **Open Scientific Collaboration** The framework is designed to encourage interdisciplinary research across neuroscience, psychology, artificial intelligence, sleep science, philosophy, and design. --- # Beyond Dream Journals Traditional dream journals preserve memories. The Dream Intelligence Framework builds intelligence. Instead of asking: > *What did I dream last night?* It asks: > *What have my dreams been teaching me over the past ten years?* This shift transforms dreaming from an ephemeral experience into a continuously evolving cognitive system. Dreams become not only memories of the night, but a lifelong source of insight into creativity, emotion, learning, and human intelligence itself. ### Part II A Computational Framework for Dreams URL: https://www.yannichen.org/part-ii-a-computational-framework-for-dreams/ Last updated: 2026-09-12T14:33:44.000Z ## If dreams are to become part of the future of human intelligence, they must first become computational objects. Today, dreams are typically treated as narratives: fragmented memories written in journals or spoken aloud after waking. While valuable, narrative alone limits what dreams can become. Stories are difficult to compare, impossible to quantify consistently, and challenging to integrate with physiological signals or artificial intelligence. A computational framework begins with a different assumption. Rather than asking *What does this dream mean?*, we ask: > **How can a dream be represented as structured information?** Every dream contains multiple layers of information operating simultaneously. The narrative sequence describes events. The visual layer contains scenes, objects, and spatial relationships. The emotional layer reflects affective intensity and transitions. The symbolic layer connects recurring motifs across time. The physiological layer relates dreams to sleep stages, brain activity, heart-rate variability, respiration, and other biometric signals. Finally, the cognitive layer captures memory consolidation, imagination, prediction, creativity, and self-modeling. Instead of collapsing these dimensions into a single interpretation, Dream Intelligence treats them as complementary representations of the same cognitive process. From this perspective, a dream becomes a multimodal cognitive dataset. Large language models organize narrative structure. Vision models reconstruct imagery. Speech models enable natural dialogue with dream memories. Wearable devices provide physiological context. Knowledge graphs connect recurring symbols and concepts across months or years. Together, these representations allow dreams to become searchable, comparable, and continuously learnable. The purpose is not to reduce dreams to numbers. It is to create computational representations that preserve their richness while enabling entirely new forms of exploration. Ultimately, Dream Intelligence seeks to transform dreams from isolated memories into longitudinal cognitive data—data that can help us better understand creativity, emotional adaptation, learning, and the evolving architecture of the human mind. ### Part I Manifesto URL: https://www.yannichen.org/manifesto/ Last updated: 2026-09-12T14:32:29.000Z # ## Building Dream Intelligence For centuries, dreams have been treated as something mysterious, irrational, or impossible to study. They disappear within minutes after waking. They resist language. They cannot be replayed, measured, or searched like ordinary memories. Yet every night, billions of people generate an astonishing amount of cognitive activity. Dreams may represent one of the largest unexplored datasets of human experience. I believe this will change. Artificial intelligence is rapidly transforming how we understand language, images, biology, and software. The next frontier is not only making machines more intelligent, but developing better ways to understand human intelligence itself. Dreams are one of the missing pieces. They contain emotion without censorship, memory without deliberate control, imagination unconstrained by physical reality, and patterns that often remain invisible during waking life. Rather than treating dreams as isolated stories, I see them as structured cognitive signals that can be recorded, modeled, analyzed, and eventually computed. This is what I call **Dream Intelligence**. Dream Intelligence is the emerging discipline of representing dreams as structured information that can be understood collaboratively by humans and AI. It combines neuroscience, sleep science, artificial intelligence, computational cognition, psychology, design, and creative technology into a shared computational framework. The goal is not to explain every dream. The goal is to build better tools for observing, organizing, and understanding one of the most overlooked dimensions of human cognition. This vision extends beyond dream journals or AI-generated images. I imagine a future where dreams become part of a broader personal intelligence system. Wearable devices continuously measure sleep physiology. Large language models help organize dream narratives. Multimodal models reconstruct visual experiences. Long-term memory systems discover recurring emotional and symbolic patterns across years. Instead of isolated entries, dreams become longitudinal cognitive data. Understanding dreams may ultimately improve creativity, emotional awareness, mental health, scientific research, and even the development of more human-centered artificial intelligence. To explore this possibility, I founded **NeuroDream**. NeuroDream is building the infrastructure for Dream Intelligence. ![](https://storage.ghost.io/c/0a/74/0a74fc6e-ce75-4858-b413-21b0a154fbd2/content/images/2026/06/ChatGPT-Image-2026---6---29----22_25_10.png) It is an ongoing effort to create the tools, research framework, and computational systems needed to record, structure, model, and generate dream experiences. This website serves as my public research notebook. Here I share essays, experiments, product development, reading notes, and ideas that contribute to this long-term mission. Much of what appears here will evolve over time. Some ideas will prove wrong. Others may become foundations for future research. The purpose is to think openly, build publicly, and invite collaboration. If you are interested in dreams, consciousness, neuroscience, artificial intelligence, computational cognition, or the future of human intelligence, I hope you'll find something worth exploring here. We're only beginning to understand what dreams are. Perhaps we're also beginning to understand what intelligence can become. — **Yanni Chen** Founder, NeuroDream