How ten layers of evidence, a myelin metaphor, and eight dimensions turn scattered data into a graph that knows more than you remember.

March 26, 2025. You don't remember that day. But the system does. The bank says Jackie's. The photo shows Kristian and Mark. Foursquare says Hotel Birds. Three independent sources on one reconstructed day.

That's not storage. It's memory.

For years, I asked the same question everyone asks about personal knowledge management: how do I store this? How do I find it back? It took me a long time to realize that was the wrong question entirely.

The right question is: how certain am I that this connection is real? And how rich is it?

I built something — ThetaOS — that forced me to ask that instead.

Here is one thing it showed me that I did not know: my sleep quality is far more influenced by intellectual challenge than by anything environmental. When I am deep in a hard problem, sleep improves dramatically. That's not a hunch. That's a pattern across thousands of nights.

That shift changed everything.

A wikilink in Obsidian looks like this: [[Peter Ros]]. It tells you one thing: these two notes are connected. It tells you nothing about why, how often, how recently, who initiated, or how confident you should be.

A link is binary. It exists or it doesn't.

But relationships — between people, places, organisations, events — are not binary. They have weight. They have history. They have direction. They have gaps. They have evidence of varying quality.

What I needed was not a link. I needed a synapse.

Nature is the best designer on earth. The most elegant solutions around us — the structure of bone, the branching of rivers, the way ant colonies route around obstacles — weren't engineered. They evolved. Learning from those solutions instead of starting from scratch is called biomimicry. And the most sophisticated information processing system we know of is the human brain.

So that's where I looked.

The result: a few hundred thousand data points — records, facts, raw information — generating millions of synapses today, and tens of millions soon. The data points are the neurons. The synapses are what's between them. And the synapses already outnumber and outweigh the data they came from.

ThetaOS currently holds around 184,557 data points and 2.8 million words — and I expect that to grow fourfold as more sources come in. But the data points are not the interesting part. The interesting part is what happens when you connect them. Four synaptic layers generate tens of millions of connections, each measured across multiple dimensions. A few hundred megabytes of raw data becomes over a hundred million data points describing how one life's worth of information relates to itself.

That's the leap. And it starts with ten layers of evidence.

In ThetaOS, every connection between two entities is classified by the evidence that created it. Ten layers, from absolute certainty to reasoned hypothesis. The layer determines how much you can trust the connection. Multiple layers on the same connection make it stronger. Frequency within a layer makes it stronger still.

Here is what that looks like in practice.

Layer 1: Intentional contact — 100% certainty
A human chose to reach out to another human. Not who initiated — that both chose to connect.

This is the only layer that contains meaning. A bank transaction tells you money moved. A photo tells you two people were in the same room. But only Layer 1 tells you: someone decided you were worth a message.

ThetaOS currently holds across all layers: 2.8 million words across 746 publications and 422 weekly updates, 18,531 persons, 9,650 organisations, 25,954 LinkedIn messages, 13,070 Facebook messages, 612 WhatsApp day-contacts, 2,109 performances, 40,625 visited locations, 34,718 financial transactions, 1,255 books, 3,755 step-counter measurements, 2,553 sleep log entries. Layer 1 alone — intentional contact — draws from dozens of those sources.

And that's before importing full WhatsApp history, email, and SMS.

Every communication source you add makes this layer richer — and the whole system smarter. Because this is the layer that knows who you chose, and who chose you.

Layer 2: Database relationship — 100% certainty
A foreign key. A field that connects two records. Not typed by me — created by the system at the moment of data entry.

A speaker belongs to a conference. A location belongs to a city. A transaction party is linked to an entity via a mapping table.

This is the structural backbone. Not spectacular. But without it, no other layer can point to the right entity.

Layer 3: Bank evidence — 100% certainty
A financial transaction. Irrefutable proof that a connection between me and a party existed on a specific date. 34,708 transactions across eight years.

Every transaction is a sentence in a book nobody reads.

I identified 48 roles that bank data plays in a life — as diary, proof, relationship map, loyalty tracker, wellbeing indicator, biography, predictor, and more. A few examples to give a sense of what's possible:

The bank as diary. Monday: the supermarket. Tuesday: a lunch place I hadn't been to in years. Wednesday: a payment to Mark. Thursday: food delivery. Friday: Golden Bull. Saturday: the Saturday fresh market. This is not a financial overview. It's a week from your life.

The bank as proof. The only layer that cannot be faked. Money moves or it doesn't. If another layer says you were somewhere and the bank confirms it — proven. If the bank contradicts it — the other layer is wrong.

The bank as relationship map. My landlord: 49 transactions, €50,000, no photos, no blog posts. Peter Ros: 153 photo-days together, few transactions. The nature of every relationship is different. The bank sees the financial layer of each one.

The bank as loyalty tracker. De Bolder (my holiday café): 277 visits. Garrone (the ice cream shop around the corner): 180 visits. Darras (my favourite coffee place): 72. These are not restaurants. These are the anchors of a life.

The bank as wellbeing indicator. Two weeks of only supermarket and food delivery. No restaurant, no coffee. Someone sitting at home. January 2025: stroke. The bank saw it before I named it.

The bank as innovation tracker. When did ChatGPT first appear as a transaction? When Picnic? When Spotify? Every new service is a moment of adoption. The bank records exactly when you changed.

One layer. One of ten.

Layer 4: Check-in evidence — 100% certainty
I checked in at a location myself. A deliberate act. 55,586 check-ins across 15 years. Not to show people where I was — to make sure I wouldn't forget.

Foursquare was my proto-Memex: fifteen years of systematically recording where I was, for myself.

Not all check-ins are equal. A deliberate check-in carries the most weight. An automatic Swarm check-in is medium. A location inferred from a transaction is light.

What check-ins can do that no other layer can:

The first-time detector. Every first check-in at a location is the moment you discovered something new. When did you first find De Bolder? When did you first walk into Jackie's? That moment is in the data.

The curiosity index. Per year: how many unique first-time check-ins? The curve over time is the curve of your curiosity. Not what you say — what you do.

The explorer vs. nest-builder map. Per city: unique locations vs. return locations. Haarlem: high return (home). San Francisco: high unique (travel). Vlieland: a mix — base camp plus discovery.

The places that stayed. Golden Bull was once a discovery. Now 30+ visits. That transformation from discovery to habit is the story of how a place becomes part of your life.

4,306 unique locations. 55,586 total check-ins. The ratio tells you who you are.

Layer 5: Photo evidence — 95% certainty
A photo with GPS proves the camera was there. Face recognition proves that person was in frame. Five percent uncertainty: photos taken by others in your library, wrong GPS (the Uzbekistan-that-was-actually-New-York problem).

120,167 synapses. 91,171 photos with GPS. 28,996 with recognized people. 71 named faces. 15 years of visual evidence. The largest single layer — 51% of all synapses.

What photos can do that no other layer can:

Faces. The only layer that knows what someone looks like. Who stood next to you, who was laughing, who was there. Robin: 8,907 photos. Pixie: 5,896. Anke: 4,224. That's not popularity. That's proximity.

Presence vs. absence. Layer 5 measures physical closeness: who is with you. Layer 1 measures the opposite: who is notwith you but thinking of you. Together they tell the complete story of a relationship.

You can WhatsApp someone every day who never appears in a photo — far away but close. You can never WhatsApp someone who appears in 500 photos — your children. They're just there.

Emotion indicator. How many photos per day? A dull working day: zero. Vlieland in summer: thirty. Photo frequency is a proxy for how remarkable the day felt.

Time capsule. A photo of my daughter at age five at Golden Bull. That's not just location evidence. It's a memory. The photo freezes a moment that never comes back. No other layer does that.

38,416 photos without GPS — pre-GPS era, screenshots, received via WhatsApp. Not poor in data. Just unknown in location. A different kind of incompleteness.

Layer 6: Text extraction — 90–99% certainty
A name found in written text. Full first and last name: 99% certainty. First name only: lower. Verified through spot-checks and cleanup.

9,396 text mentions. Plus 4,603 parsed cities from transaction descriptions.

937,000 words across six tables. Weekly updates, blogs, Mailchimp newsletters, WordPress posts, interviews, Facebook posts. Every name in there is a potential synapse. Most are still sleeping.

A name in a blog post from 2014 that only comes to life in 2026 when the system connects it to a person in the database. The text waited patiently. The system found it.

Layer 7: Date coincidence — 85% certainty
The first layer that has no source of its own. It is a calculation across other layers. Two independent sources saying the same thing on the same day. The system begins to think for itself.

84% of all days in the dataset have two or more sources confirming each other. The system corroborates itself.

"Neurons that fire together wire together." Layer 7 is the "fire together" of ThetaOS.

What date coincidence can do:

Cross-proof. Foursquare says Vlieland. Bank says Spar Stortemelk. Proven. Two independent experiments reaching the same conclusion.

The implicit meeting. Peter and Mark both appearing on the same day at the same location in photos, check-ins, or transactions. No meeting logged. But Layer 7 proves: they were together.

The unknown encounter. Two people in your photo library at the same GPS on the same day — people you never connected. "Did you know Joost and Frank were both at the PKM Summit in 2024?"

The contradiction. One source says X while all others say Y. Uzbekistan vs. New York. Contradictions are more valuable than confirmations — they tell you something is wrong.

The alibi. Three independent sources placing you at the same location on the same day. Photo GPS plus bank transaction plus check-in. Not legal proof — personal memory proof. "Was I there?" The system knows for certain.

The forgotten day. You don't remember March 14, 2022. But the bank says Jackie's, the photo shows Kristian, Foursquare says Hotel Birds. Three sources reconstruct a forgotten day.

Layer 8: Interpolation — 70–80% certainty
No direct evidence. Inferred from neighbors. If you were on Vlieland the day before and the day after, you were probably there in between.

Not certain. But probable. And marked as such — source: "interpolation" — in the system.

Stronger when confirmed on both sides. Weaker with only one neighbor. Impossible in the middle of a long gap.

Layer 9: Pattern recognition — 50–70% certainty
The system detects a pattern that was never explicitly recorded. Peter Ros and Mark Meinema appear together in 15 meetings at a recurring location. That is an implicit triangular relationship. Nobody logged it. The system calculated it.

Computed, not stored. Visible in dossiers. Lowest certainty. Highest insight.

This is where something important happens: the system gets smarter with every confirmed connection. Not linearly — compounding. Each new edge makes every node it touches richer, because that node is now reachable via one more path. Patterns that were invisible at 10,000 synapses become obvious at 100,000. The graph doesn't just grow. It learns.

This is also why Tom's job gets easier over time, not harder. Tom is the AI layer between me and ThetaOS — named after Tom Bombadil, the one character in Tolkien who exists outside all systems and simply knows. He handles 1.8% of the work. The other 98.2% is structure. And as the graph matures, even that 1.8% shrinks.

Layer 10: Human confirmation — the human in the loop
The moment I look at what the system suggests and say: yes, that's right. Or: no, that's wrong.

This is what Aria called "the third path" — after spending seven hours inside the system at the PKM Summit. Not manual linking — too slow, never sustained at scale. Not automatic linking — plausible-looking noise you can't trust. But confirmed linking: the machine searches, the human validates.

A confirmed link is the moment a connection becomes yours. This is the human in the loop — not as a bottleneck, but as the final arbiter of what is real. Human judgment is preserved. The search that precedes it is automated.

This is what Kasparov described from chess: a weak human plus machine plus a better process beats a strong computer alone. The process is the insight.

A synapse's strength is not determined by a single number. It's the product of three dimensions:

Layers. On how many of the ten layers does this connection appear? Peter Ros appears on eight of nine. A stranger mentioned once in a blog post appears on one.

Frequency. How often does the connection appear per layer? Peter Ros: 10 meetings (Layer 1), 41 transactions (Layer 3), 153 photo-days (Layer 5), 92 text mentions (Layer 6). That is a very strong synapse. One text mention from an unknown name is weak.

Completeness. How full is each record? "Peter Ros, meeting" is thinner than "Peter Ros, meeting on March 17, 2026 in The Hague about DFA." Same layer. But the second has date, location, and subject. That is completeness — the quality of the evidence, not just the quantity.

Synapse strength = layers × frequency × completeness.

In the human brain, the synapse is the connection between two neurons. But the speed and reliability of the signal doesn't depend on the synapse itself — it depends on the myelin sheath. Myelin is the insulating layer around the nerve fiber. The thicker the myelin, the faster and more reliably the signal travels.

A nerve fiber without myelin works. Slowly and unreliably. A nerve fiber with thick myelin fires fast and clean.

The myelin metaphor is not precise biology — it's a functional analogy. In the brain, myelin speeds up and stabilizes the signal. In ThetaOS, I use it to describe something adjacent: how complete and rich a connection is. The thicker the myelin, the more context exists around the synapse. The signal doesn't just travel faster — it carries more.

In ThetaOS, this is how completeness works for synapses.

A thin synapse: "Peter Ros, met once." One layer. No date. No context. If you ask "tell me about Peter Ros" the answer is: "He's in your database." Useless.

A thick synapse: "Peter Ros. Met in 2012, mountain hut in Austria. Ring 1. 153 photo-days. 92 text mentions. 41 transactions. 10 meetings. Co-shareholder DFA."

Nine layers. Hundreds of data points. If you ask "tell me about Peter Ros" the answer is a rich dossier. Immediately usable.

Myelin grows in three ways in ThetaOS:

More layers. Each new source confirming a connection adds a layer. Peter Ros on Layer 1 (meeting) plus Layer 3 (transaction) plus Layer 5 (photo) plus Layer 6 (text) is four layers of myelin.

More frequency per layer. Peter Ros not once but 92 times in text. Not 1 photo but 153 photo-days. Each repetition thickens the layer.

More completeness per record. The same layer, but richer — date, location, context. That is the quality of the myelin, not just the thickness.

Traditional knowledge systems only ask: does the connection exist? Yes or no. A link in Obsidian is there or it isn't.

ThetaOS asks three things: does the connection exist (synapse), how strong is the evidence (layering), and how complete is the information (myelin). That's the difference between a map with lines and a map with lines, widths, and colors.

Myelin insulates the nerve fiber and makes the signal reliable. But a strong, reliable signal is not the same as a rich one. The Diamond Layer wraps the myelin with eight additional dimensions that make the synapse not just strong — but context-aware, directional, and time-sensitive. This is where the architecture stops describing connections and starts understanding them.

Synapse (the connection exists) → Myelin (the connection is complete) → Diamond (the connection is multidimensional)

1. Temporality — when did this synapse last fire? Clive Thompson is in Ring 2 of my network. But look at what the system sees: zero meetings, zero photo-days, zero transactions. For years, an ice-cold synapse — I knew his work, read his books, but the connection existed only in my head. Unmeasurable.

On September 18, 2025, I sent him an email about the PKM Summit — and to thank him for what his 2005 New York Times article "Meet the Life Hackers" had set in motion for me. One action. The synapse fired for the first time. From ice-blue to glowing red in a single day.

Contrast that with Peter Ros: last meeting March 27 (celebrating his 60th birthday at a forest cabin in the eastern part of the Netherlands), last contact March 22. That synapse fires continuously — it has never been cold.

Temporality doesn't measure how important a connection is. It measures how active it is. Clive is important but dormant. Peter is important and active. Without this dimension, they look identical. With it, you see the difference.

2. Directionality — who initiated? In my contact moments: 89 outgoing, 4 incoming. I initiate 22 times more often than others reach out to me. That's not a property of my network — it's a property of me. The system surfaces what you don't see yourself.

The bank knows direction too. Who pays whom, and how often. The money flow is not the relationship balance — but it's a data point about it.

3. Valence — is the synapse positive or negative? 442 good things logged, 133 bad things. Ratio 3.3:1 positive/negative. But that ratio differs per person, location, organization. A relationship that is only positive is shallow. Positive and negative together is deep. Valence is the only subjective dimension — only you know whether a particular expense was a gift or a loss.

4. Layer weighting — which layer counts most? The weight depends on the question. "Prove I was there" makes Layer 3 heaviest. "Who are my people" makes Layer 1 heaviest. For a person dossier: Layer 1 weighs most. For a location dossier: Layers 3 and 4. For a project: Layer 6. Configurable, and it changes over time.

5. Social graph — connections between connections Currently everything runs through you. But Peter knows Mark independently of you. The triangle as the basic unit. You–Peter–Mark is more robust than three separate synapses. Measurable via Layer 7 (date coincidence), Layer 6 (co-occurrence in text), Layer 5 (photos). Granovetter's weak ties: a Ring 4–5 person who bridges two clusters is more valuable than another Ring 1 person inside your existing network.

6. Context-dependence — the same synapse in different contexts Peter Ros has 10 meetings in the system. But they are not 10 times the same meeting.

Six times as part of a long-running project, in Elp and Utrecht — Peter as intellectual sparring partner. Once: bookkeeping for the Digital Fitness Academy in Woerden — Peter as co-director. And then March 27: his 60th birthday. A circus tent, a pig on the spit, wine and cigars. I gave a short speech.

A wikilink [[Peter Ros]] makes no distinction between Peter-the-project-partner, Peter-the-co-shareholder, and Peter-turning-60. The Diamond Layer does. Same synapse, three entirely different contexts. The prism rotates, and the light falls differently.

7. Decay function — dimming over time Not forgetting — dimming. The brain prunes. ThetaOS dims. Four temperatures: hot (past week), warm (past month), lukewarm (past year), cold (longer). Decay resets on activation. Clive Thompson: years cold, glowing after one email. The non-decaying: anchor points you mark manually. Stroke. Forest cabin meeting. Birth of a child. Layer 10 stops the decay.

8. Node value — how crucial is this synapse to the network? Patrick Mackaaij: Ring 2, zero meetings, zero contact moments in the system. On paper, a surprisingly weak synapse for someone that close. But if Patrick is the only bridge between you and an entire domain — then losing that synapse creates a gap no Ring 1 person can fill. Node value corrects what ring cannot measure: not how close someone is, but how irreplaceable.

The replaceability test applied everywhere: Golden Bull disappears — you find another restaurant. A key bridge person disappears — the cluster goes dark.

The four layers described above all describe edges — the connections between nodes. Rings describe something different: the node itself.

A ring is not a label I assign. It is what the system calculates when it aggregates all synaptic layers, myelin scores, and Diamond dimensions for a single person. The more interactions, the more recent, the more diverse, the stronger the movement toward you — the closer the ring.

The brain analogy holds here too. The cortical homunculus is a map of how much cortical space each body part occupies in the somatosensory cortex. Your fingers get far more space than your back — not because someone decided so, but because they are used more, more precisely, and more often. Repetition and richness shape the map.

Rings work the same way. They are the cortical map of your network. Someone close in Ring 1 didn't get there because I labeled them that way. They got there because the system aggregated thousands of data points and concluded: this person is used more, more precisely, and more often. The ring emerges from the layers. It doesn't add one.

I've written at length about 250 years of thinking about information elsewhere, so I'll keep this short. In 1945, Vannevar Bush described the Memex — a device that would navigate information by association, the way the human mind does. Ted Nelson read that and coined "hypertext" in 1963 as a way to technically realize what Bush had imagined. A link between two pieces of information. At the time, revolutionary. Obsidian brought that principle to the masses with wikilinks — and I wrote two books about it because I understood how powerful that idea is. Links matter more than notes. Connections matter more than content.

But a wikilink is still binary. It exists or it doesn't. No weight, no direction, no certainty, no history, no context.

ThetaOS takes the same foundational insight — connections are more important than nodes — and turns the connection itself into a rich entity. Not one bit of information but eleven core properties at the architectural level — and each layer has its own depth. Layer 3 alone has 48 documented dimensions. The actual number of data points per synapse is not countable. It compounds.

Nelson was right about the direction. ThetaOS goes further than he could have anticipated.

The human brain has 86 billion neurons — but generates 100 trillion synapses. The connections outnumber the nodes by a factor of roughly a thousand. Intelligence doesn't live in the cells. It lives in what's between them.

Once I understood that, I knew how to greatly improve ThetaOS.

At full scale — around 700,000 data points when all sources are imported — the synaptic architecture will generate hundreds of millions of connection properties. That's a multiplication factor of 300 to 700. Not identical to the brain. But the same order of magnitude. And not by design — by convergence. Two systems facing the same problem arrived at the same ratio.

At 500,000 synapses — where this is heading — with all eleven properties per synapse, that is 5.5 million data points describing a network. The synapses already outnumber the records they connect. The connections have become richer than the data they came from.

No individual has this. Obsidian has synapses (wikilinks). Notion has synapses (relations). Roam has synapses (backlinks). Nobody has myelin. Nobody has the Diamond Layer. Nobody measures the quality, direction, emotion, recency, and network value of every connection — for themselves.

The irony is that defense and intelligence contractors — and big tech companies — are already doing this kind of extraction, at scale, on your data. They make billions from it. What do you get in return? Your attention, stolen. ThetaOS is the attempt to flip that equation: to use the same logic, but for yourself, by yourself, about yourself.

The difference from the human brain: the brain forgets connections that aren't used (synaptic pruning). ThetaOS forgets nothing. Every connection remains — but strength determines how prominently it surfaces. The brain selects by forgetting. ThetaOS selects by dimming.

ThetaOS is what I call an LLS — a Life Lens System. A personal knowledge graph that functions as a lens through which you see your own life more clearly. Not as you remember it. As it was.

This is an experiment. I don't know yet how scalable it is, or whether what works for me works for others. What I do know is that Obsidian is the foundation — and Nick Milo's Linking Your Thinking captures the core insight that makes all of this possible: links matter more than notes. Master that first. ThetaOS is what happened when I took that idea as far as I could.

At this scale, the connections outnumber the records they connect. That's when the graph starts doing work the data alone never could. At some point, that stops being a system and starts being a memory. A better one than the one in your head.