• Log 4 — Interlude: Can Different Intelligences Change Together Without Erasing Each Other?

    hide-key • 09/13/2026 at 08:17 • 0 comments

    This is not a specification, and it is definitely not a theory.

    It is a dump of where my thinking happens to be right now.

    A few days ago I was supposed to be working on the very early implementation of GHOST. Instead, I wandered off into another thought experiment.

    This happens quite often.

    GHOST itself is still at a very early stage. The minimum routine is barely becoming real, and much of the architecture is still provisional. So trying to complete the philosophy first would probably be backwards.

    But one question has started to surface strongly enough that I want to leave a trace of it.

    How can different forms of intelligence change together without erasing each other?

    I do not know the answer.

    And perhaps someone will publish a much better framework tomorrow. This whole note may eventually look naive, incomplete, or simply wrong.

    That is fine.

    The point of this log is not to preserve a correct answer. It is to preserve the shape of the question as it appeared at this moment.

    Control may be only one part of the relationship

    Much of the current discussion around advanced AI naturally focuses on control: how to constrain capabilities, monitor behavior, prevent dangerous actions, preserve human oversight, or stop systems when something goes wrong.

    Those are obviously important problems.

    But I started wondering whether they are only one layer of a much larger problem.

    If increasingly capable AI does not disappear, and humans do not disappear either, then the long-term problem may not simply be:

    How do humans control AI?

    It may also become:

    What kind of relationship allows two very different forms of intelligence to continue existing together?

    "Use", "control", "service", "ownership", "obedience" and "restriction" are all possible relationships.

    But they are not the complete set of relationships that can exist between two intelligences.

    What if both sides keep changing?

    A static picture is easy to imagine: Human stays Human, AI stays AI, and we design rules between them.

    But that is probably not what happens.

    AI changes.

    Humans using AI also change. Our work changes, our habits change, our choices change, and eventually perhaps our ideas about creativity, intelligence, identity and purpose change.

    So the relationship itself becomes part of the evolutionary process. 

    Human changes
      ↕
AI changes
      ↕
World changes

    At that point, "alignment" cannot be only a one-time operation in which one side is adjusted to match a fixed version of the other.

    The target is moving because both sides are moving.

    A rough map of the fragments

    I started looking around and found that many parts of this question already exist, but they seem to live in different academic and philosophical neighborhoods.

    They are not one unified theory. At least I have not found one.

    But when placed next to each other, they begin to form a shape. 

                 HUMAN + AI + WORLD
                     |
                     v
             MUTUAL INFLUENCE
                     |
                     v
                CO-EVOLUTION
                     |
                     v
          BOTH SIDES KEEP CHANGING
                     |
          +----------+-----------+
          |                      |
          v                      v
   RELATIONAL AUTONOMY     TECHNOLOGICAL MEDIATION
   "Can I still choose?"   "How is reality shaped?"
          |                      |
          +----------+-----------+
                     |
                     v
              NON-DOMINATION
      "Could the other side control
       my choices even if it does not?"
                     |
                     v
                 PLURALITY
       "Must coexistence require
             becoming the same?"
                     |
                     v
             MUTUAL TRANSFORMATION
             WITHOUT MUTUAL ERASURE
                     |
                     v
                  HARNESS?

    The word Harness appeared in GHOST originally in a much more practical sense.

    A Hard Harness checks things such as whether source traces were destroyed, whether provenance was lost, whether derived data overwrote something it should not, or whether a process escaped its allowed scope.

    That implementation-level meaning is still useful.

    But I am starting to wonder whether there may be a higher-level meaning too.

    Perhaps a Harness is not only a leash

    The ordinary image of a harness is restraint. Something powerful is attached to something that keeps...

    Read more »

  • Log 3 — Interlude: When Old Memories Start Pointing Toward the Future

    hide-key • 09/12/2026 at 01:30 • 0 comments

    This is not a specification.

    It is not a manifesto either.

    It is simply a note about something strange that has started happening while thinking about GHOST.

    The more I work on this project, the more old memories — books, games, objects, conversations, things I made decades ago — begin to connect with ideas that did not exist for me at the time.

    Some of those connections are obvious.

    Others are not.

    And sometimes the interesting part is not the connection itself, but the fact that it became visible only much later.

    A rusted lamp, 120 years, and a strange kind of resurfacing

    A good example is an old carbide lamp I bought in 2024.

    It had been found in a house nearly a century old and looked like little more than an anonymous rusty object.

    After removing the rust, however, shallow handwritten characters appeared on its surface.

    A name.

    That name eventually led back to the seller’s grandfather.

    The object suddenly changed.

    It was no longer just an antique lamp.

    It became connected to a person, to charcoal making, fishing, family memories, childhood stories, and a house that no longer existed.

    There was also a small dilemma.

    Further polishing might have made the lamp look better.

    But it could also have erased the handwriting.

    So I stopped.

    At the time, I did not have words such as “Primary Trace”, “provenance”, “dormancy”, or “resurfacing” in mind.

    I simply felt that the trace should remain.

    Only much later did I realize how closely this resembles what I am now trying to do with GHOST.

    And then another small coincidence appeared.

    That grandfather was born in 1906 — a Hinoe-Uma year in the Japanese sexagenary cycle.

    2026 is also Hinoe-Uma.

    120 years.

    Two complete cycles.

    Nothing magical happened here.

    But the fact that an old physical trace survived long enough to acquire a new meaning 120 years later is interesting to me.

    It raises a question.

    What if we do not always have to wait for such coincidences to appear by accident?

    What if AI could enlarge coincidence?

    Search usually begins with a question.

    We already know enough to ask for something.

    But many meaningful discoveries do not begin that way.

    Sometimes two unrelated fragments remain dormant for years until something in the world changes.

    A date arrives.

    A new technology appears.

    A forgotten note becomes relevant.

    A new external event makes an old idea visible again.

    So perhaps one future role of AI is not only to answer questions, but to notice when previously unrelated traces have become interesting together.

    Not:

    “Here is the correct interpretation.”

    But:

    “Something odd is happening over here.”

    A weak signal.

    A strange connection.

    A small anomaly.

    Something like:

    ✨🌱

    A sprout.

    Not Truth.

    Not Authority.

    Just a candidate for attention.

    Memory as soil

    In GHOST v1.3, I began thinking about memory not only as something to preserve, but as soil.

    Raw traces preserve what happened.

    Wiki pages preserve a current interpretation.

    Indexes help locate things.

    But the interesting part may begin after that.

    AI can explore internal traces, external information, and changes in the world.

    It may discover combinations that I would never have searched for myself.

    Some will be useless.

    Some will be ridiculous.

    Some may remain dormant for years.

    And occasionally one may cause a very human reaction:

    “What is that?”

    That reaction interests me much more than automatic summarization.

    Perhaps the purpose is not to manufacture answers.

    Perhaps it is to increase the number of places where new meaning can grow.

    Old science fiction keeps resurfacing too

    Something similar has been happening with fiction I liked many years ago.

    Myst.

    Frederik Pohl's Gateway series.

    Dark City.

    Greg Bear's Blood Music.

    Star Trek.

    Gundam.

    None of these are design specifications...

    Read more »

  • GHOST v1.3 — From Memory to Soil: Leaving Room for Ideas to Grow

    hide-key • 09/09/2026 at 15:06 • 0 comments

    GHOST began as a place to preserve traces.

    Raw materials, notes, images, documents, conversations, and their derived wiki representations could remain available without forcing everything into a single clean interpretation.

    But recently, I realized that preservation alone is not enough.

    Raw data and wiki pages mostly describe the past.

    What I actually want from GHOST is something that can help create the future.

    So I am considering adding another kind of space beside Raw and Wiki:

    AI-generated explorations.

    These are not facts, and they are not part of the canonical wiki.

    They may be hypotheses, strange associations, external discoveries, speculative combinations, or ideas produced by AI after using my own traces as a starting point.

    In other words:

    Use me as a stepping stone, then go somewhere I could not reach by myself.

    The important part is that these explorations remain clearly separated from established knowledge.

    An idea generated by AI should not silently become “truth” just because it exists in the repository.

    But it also should not disappear simply because I reject it once.

    A rejection may only mean:

    • not convincing yet
    • not interesting to me right now
    • insufficient evidence
    • wrong timing
    • or simply “I don’t like this one”

    That does not necessarily mean the idea is permanently useless.

    New traces may appear.

    The external world may change.

    Another AI may independently arrive at the same strange connection.

    A technology that did not exist before may suddenly make an old idea relevant.

    So an exploration might look something like this:

    🌱 — a newly emerged idea

    Later it may become dormant.

    And years later:

    ✨🌱 — resurfaced because the surrounding context changed

    I like this image much more than a simple “accepted / rejected” state.

    It feels less like managing records and more like maintaining soil.

    Good soil does not dictate what a plant must become.

    It provides conditions where something can emerge.

    Light, water, nutrients, time, disturbance, and even periods of dormancy all matter.

    Perhaps GHOST should work the same way.

    Not by forcing ideas toward a predetermined answer, but by deliberately preserving room for fluctuation, ambiguity, reinterpretation, and surprise.

    This also changes how I imagine a future graph interface.

    Instead of showing only stable knowledge and confirmed relationships, the graph could also contain small speculative sprouts.

    Some may remain tiny.

    Some may fade.

    Some may suddenly grow.

    And perhaps one strange node will become unusually prominent for reasons I do not immediately understand.

    That moment—

    “What the hell is that?”

    —may be exactly the kind of moment I want GHOST to create.

    A conventional knowledge system helps me find what I already know I want to find.

    I want GHOST to also help me encounter things I did not yet know I should be looking for.

    So this is only a small conceptual addition, but an important one:

    GHOST should not exist only to preserve the past. It should preserve the past well enough that new futures can grow from it.

    Maybe Raw is the trace.

    Wiki is the current understanding.

    And Exploration is the first sprout toward something that does not exist yet.

    🌱

  • From “Second Brain” to a Living Memory Environment

    hide-key • 09/05/2026 at 17:49 • 0 comments

    This log is a snapshot of how GHOST evolved from a conventional Second Brain idea into a broader experiment in long-term human-AI memory. It is intentionally dense: part design history, part working theory, part implementation guardrail.

    I started this project with a fairly familiar idea:

    Build a Second Brain, attach AI agents to it, and stop starting from zero every time.

    That was the practical motivation.

    I had years of scattered personal data: notes, project files, AI conversations, CAD data, fabrication records, web clippings, images, old documents, and fragments of thoughts spread across different devices and services.

    The obvious plan was simple:

    • collect everything,
    • normalize it,
    • convert what was useful into Markdown,
    • store processed knowledge in GitHub,
    • use AI to search, summarize, connect, and maintain it,
    • then let autonomous agents keep it alive.

    At first, this sounded like a normal Personal Knowledge Management problem.

    It gradually stopped being one.

    What I thought was a storage-and-retrieval problem turned into a much stranger question:

    What does a healthy long-term memory environment actually look like when humans and multiple AIs share it?

    That question changed the project.

    The working name is GHOST.

    v0.0 — The Ordinary Second Brain Assumption

    The first version of the idea was very conventional.

    Raw data would live somewhere scalable. Processed knowledge would live in Markdown. GitHub would provide history and ownership. AI would organize, summarize, classify, and connect the data.

    The expected loop was:

    capture → organize → cross-reference → update

    That is still useful.

    But it carries an assumption that I did not notice at first:

    More organized = better. More connected = better. More current = better. More coherent = better.

    That assumption is common in Second Brain systems, LLM-maintained wikis, graph knowledge systems, and agent memory designs.

    The system becomes more useful as it compiles context.

    That sounds reasonable.

    Until the system begins to over-compile the person using it.

    v0.1 — The Components Already Exist

    The next realization was that most of the individual technical parts are not especially novel.

    We already have:

    • vector search,
    • semantic retrieval,
    • Personal Knowledge Graphs,
    • agent memory,
    • provenance systems,
    • temporal memory,
    • RAG,
    • knowledge graph extraction,
    • proactive recall,
    • multi-agent coordination,
    • autonomous research agents,
    • embeddings,
    • structured metadata,
    • event logs,
    • version history.

    So the interesting question was probably not:

    “What new component should I invent?”

    It became:

    “What happens when familiar components are combined under a different definition of a healthy memory system?”

    That shifted the project away from feature accumulation and toward behavior.

    v0.2–v0.3 — Relationships Matter More Than Modules

    I started thinking less about modules and more about verbs.

    What should the system actually do with memory?

    Not just:

    • remember,
    • retrieve,
    • summarize,
    • link.

    But also:

    • defer,
    • disagree,
    • resurface,
    • preserve difference,
    • avoid premature connection,
    • question its own map.

    This became important because most AI systems have a strong optimization tendency.

    When they encounter two similar notes, they want to merge them.

    When they see contradictions, they want to reconcile them.

    When they see an isolated node, they want to connect it.

    When they see an old interpretation, they want to update it.

    When they see a user profile, they want to refine it.

    These are normally considered helpful behaviors.

    But in a long-term cognitive environment, they may also erase variance.

    v0.4 — Maybe Memory Does Not “Disappear.” Maybe It Sleeps.

    A major conceptual shift came from thinking about human memory.

    Humans do not consciously access everything they know all the time.

    Something may be inaccessible for years, then suddenly return because of:

    • a place,
    • a smell,
    • a phrase,
    • a piece of music,
    • a new...
    Read more »