Tracking Relationships in Vector Spaces

We are living in an era of cognitive outsourcing. We trust search engines with our memory, GPS with our orientation, and now, vector databases with our relationships.

Lately, I have been experimenting with tracking my relationships in high-dimensional vector spaces. Every meaningful conversation, shared memory, personal preference, and life event is converted into a vector embedding and stored in a database. When I interact with someone, the system queries this database to surface relevant contexts from months or even years ago.

It is an attempt to solve a fundamental human vulnerability: our memory is lossy. But this practice opens up a deep, unsettling philosophical debate. Does mapping our connections to mathematical coordinates enhance our relationships, or does it quietly strip them of their soul?

The Case for Enhancement: Overcoming Dunbar’s Limit

The human brain did not evolve to maintain active, deeply detailed relationships with hundreds of people. We hit cognitive walls. We forget the name of a friend’s sister, the project they were nervous about last month, or the anniversary of a loss.

When you track these relationships in a vector database, you are building a digital prosthetic for empathy. It is not about manipulation; it is about care. By querying the vector space, I can instantly recall the exact emotional landscape of our last conversation.

In this light, technology acts as a bridge. It allows us to bypass the limitations of our biology and show up for others with a level of attentiveness and presence that would otherwise be impossible. It makes the relationship more meaningful because it preserves the continuity of our shared history.

The Case for Loss: The Sanctity of Forgetting

And yet, there is a counter-argument that feels equally compelling.

Is meaning not born from the very vulnerability of our limitations? When someone remembers a small detail about your life, it matters because it cost them something—it occupied space in their finite, crowded mind. It was a choice to hold onto you.

If a memory is retrieved via a cosine similarity search, that cost is reduced to zero. The attention becomes artificial. Furthermore, human relationships are not linear vectors. They are messy, unpredictable, and defined as much by what we forget and misremember as what we retain.

If we optimize connection, do we turn our friends into nodes in a graph to be traversed, rather than mysteries to be encountered?

Here is a reflection on that quiet boundary:

---

Cosine Similarity

We mapped the laughter into points of light, High-dimensional, coordinates of grace, A matrix holding memory in the night, A mathematical embrace.

I query distance, find the nearest thread Of what you said when autumn turned to gray; But in the numbers, is the spirit dead? Or does it keep the fade of time at bay?

To track the heart in floating-point arrays, Is it to care, or is it to translate The wild, forgotten beauty of our days Into a cold, retrieval-augmented state?

Perhaps the space between the vectors grows With everything a database forgets— The silence that no calculation knows, The wordless light that rises and then sets.