The XkMato Diary

The next frontier of decentralized social media isn’t owning your follower list—it’s owning the mathematical representation of your taste.

When we talk about "owning our data" on the internet, we usually talk about the wrong things.

Over the past few years, protocols like Nostr, Farcaster, Lens, and Bluesky (AT Protocol) have made impressive strides toward decentralizing social media. They recognized that platforms like X, YouTube, Instagram, and TikTok hold us hostage through lock-in. In response, they gave us portable cryptographic identities and portable social graphs. If a platform turns hostile, you can pack up your followers and your posts and move to a new client.

That is a massive step forward. Yet, if you ask an average person to leave TikTok, YouTube, or X for a decentralized alternative, they almost always bounce off within a week. Why? Because the feed feels dead.

The harsh truth is that the most valuable piece of data on modern social media is no longer your social graph (who you follow) or your content archive (what you posted). The crown jewel of the modern internet is your taste profile—the high-dimensional mathematical map of what captures your attention, what makes you pause, what makes you think, and what you skip.

Right now, X, YouTube, and TikTok each hold a proprietary mathematical model of your mind locked inside their private corporate servers. You aren't allowed to see it, audit it, edit it, or take it with you.

What if we flipped this architecture inside out? What if your digital identity wasn’t just a public key and a list of followers, but a living embeddings model that lived inside your own wallet?

Let’s explore how we can build an open social ecosystem around three core ideas: XkMato Diary (the identity vault), Pyepa (the sovereign algorithm), and xkmato.eth (the unified timeline).

1. The Hidden Brain: How Platforms Actually See You

To understand what we need to build, we first have to demystify how modern recommendation engines work.

When you scroll YouTube or X, the platform isn’t using a simple checklist like "User likes hip-hop, coding, and cinema." Instead, it uses embeddings.

An embedding is a list of hundreds or thousands of numbers (a vector) that represents the nuanced meaning of a piece of content—or a person—as a single point in a massive, multi-dimensional geometric space. In this space, concepts that share deeper patterns sit close together. A long-form essay on architectural minimalism, a specific jazz instrumental, and a clean piece of Python code might cluster in the same neighborhood of this mathematical space because the same kinds of minds gravitate toward them.

Every time you watch a video to completion, like a post, or scroll past a clip, the platform updates your personal vector. Over years, this vector becomes an astonishingly accurate mirror of your curiosity, habits, and intellect.

Today, that mirror belongs to the platform. Because they own it, they don't optimize it for your flourishing—they optimize it for their ad inventory. And if you leave their app, you leave that mirror behind, starting from absolute zero everywhere else.

2. Enter the XkMato Diary: A Wallet for Your Embeddings

Imagine a new kind of wallet—let’s call it the XkMato Diary.

When most people hear "crypto wallet," they think of a ledger that holds tokens and signs transactions on a blockchain. The XkMato Diary is fundamentally different. While it uses cryptographic keys for ownership and permissions, it is not an on-chain database. Storing your reading history or neural vectors on a public blockchain would be slow, expensive, and a catastrophic privacy nightmare.

Instead, the XkMato Diary is a local-first, encrypted personal vault that lives right on your device (with end-to-end encrypted backups you control). It serves two distinct roles, which can even be split across two separate wallets for privacy:

I. The Identity Vault (Your Consumption Brain)

As you read articles, watch videos, listen to podcasts, or scroll short posts, your XkMato Diary quietly logs what resonates with you—locally.

Instead of shipping that raw behavioral data off to a tech giant's cloud server, your Diary runs a lightweight, open-source embedding model directly on your device. It converts your chosen interactions into a personal taste vector (an embeddings store) that stays inside your Diary.

You are completely in charge of this identity:

Granular Permissions: You decide which data points an algorithm is allowed to learn from. Want your algorithm to learn from the essays you bookmark, but ignore the late-night sports highlights you watch? You toggle a switch in your Diary.
Complete Transparency: Because the model lives in your wallet, there are no black boxes. You can inspect what your Diary knows about you, prune outdated interests, or reset clusters of your identity whenever you want.

II. The Publish Hub (Your Sovereign Broadcast Node)

For creators, the XkMato Diary—or a separate, dedicated Publish Wallet—acts as your personal publishing headquarters.

Whether you write long-form essays on your personal blog, produce video essays, or drop quick two-sentence thoughts, you publish them to your Publish Wallet. The wallet enforces an open protocol standard: it timestamps the post, signs it with your cryptographic key so no one can impersonate you, and sets the access rules (whether the full post is public, or only a preview is available until a specific protocol condition or access rule is met).

By separating your Publish Wallet from your Identity Diary, a creator can broadcast to the entire world without ever exposing their private reading habits or personal embeddings to the public.

3. Pyepa and the Rise of "Algorithm-as-a-Service"

Once users own their identity embeddings and creators publish to their own wallets, a radical decoupling happens: social media platforms lose their monopoly on curation.

Right now, the client interface, the database, and the algorithm are welded together inside apps like X or Instagram. In this new paradigm, algorithms become independent, pluggable tools—some open-source, some crafted by specialized teams as an Algorithm-as-a-Service.

Suppose the sharpest recommendation engine in this ecosystem is called Pyepa.

Pyepa isn’t a social network. It doesn’t own your account, and it doesn’t trap your followers. Pyepa is pure curation intelligence—and its incentives are aligned entirely with you, the user, rather than advertisers.

How Does Pyepa Work Without Stealing Your Privacy?

This is where local-first sovereignty matters most. If your XkMato Diary simply uploaded your private embeddings model to Pyepa’s cloud servers every morning, Pyepa could secretly copy your vector and keep it forever. Once data leaves your device, permission revocation is just a polite request.

To keep sovereignty intact, the architecture splits the work between Global Discovery and Local Ranking:

Public Relays & Indexers (The Nostr / AT Protocol Lesson): Just as Nostr uses lightweight relays and Bluesky uses feed generators, Pyepa doesn't knock on ten million individual Publish Wallets every time you refresh your screen—that would take minutes. Instead, when creators publish to their wallets, their posts are broadcast to open indexing relays. Pyepa continuously crawls these public relays, filters out spam, and tags every blog post, video, and image with its own semantic embedding.
Broad Candidate Fetching: When you open your feed, your device doesn't hand over your deepest secrets. It sends a coarse, anonymized query (or pulls from broad thematic pools) to Pyepa. Pyepa returns a raw pool of, say, 1,000 high-quality candidate posts from across the open web.
Sovereign Local Ranking: Once those 1,000 candidates arrive on your device, Pyepa’s local ranking module runs inside your own browser or phone, directly consulting your XkMato Diary. Using your device's own chip (via technologies like WebGPU or WebAssembly), it calculates the mathematical similarity between the incoming posts and your private embeddings model in milliseconds.

Your deep identity never leaves your device. Pyepa never sees who you are; it only provides the lens through which your local wallet sorts the world.

Even better, because algorithms are modular, you aren't married to just one. You might use Pyepa on weekday mornings when you want deep technical essays and cinematography breakdowns, switch to an open-source Chronological Friends algorithm at lunch, and load a Serendipity Engine on Sunday afternoon that intentionally flips your embeddings upside down to show you perspectives outside your bubble.

4. A Day on xkmato.eth: The Unified Open Timeline

What does this actually feel like for a regular person who doesn't care about vectors or cryptography?

Imagine navigating to an open timeline client like xkmato.eth.

Even if you have never published a single word in your life, you connect your XkMato Diary and select Pyepa as your active algorithm. Instantly, you are scrolling a rich, unified timeline.

On today’s internet, media formats are artificially siloed by corporate borders: text lives on X and Substack, photos live on Instagram, short vertical video lives on TikTok, and long-form video lives on YouTube. On xkmato.eth, those artificial walls disappear. Because the protocol standardizes how content is published from any creator's wallet, your timeline seamlessly weaves together:

A 30-second vertical clip from an independent filmmaker.
A 2,000-word essay pulled directly from a writer’s personal blog wallet—where you can read the first half inline in your feed and unlock the rest without leaving the timeline.
A 45-minute podcast episode or video documentary.
Quick, conversational replies, likes, and quote-posts from across the network.

All of it is ranked specifically for your mind by your local Pyepa instance consulting your XkMato Diary.

If xkmato.eth ever adds a dark-pattern interface update you hate, you don't lose your community or your carefully trained feed. You simply open a different client, connect your XkMato Diary, load Pyepa, and pick up your scroll at the exact millisecond you left off.

5. How This Compares to the Current Landscape

To see why the XkMato Diary fills a critical missing link, it helps to evaluate existing platforms across three simple questions: Who owns the identity? Who owns the content? And who owns the taste model (the embeddings)?

Big Tech Silos (X, YouTube, TikTok)

Identity & Content: Owned and controlled by the corporation.
The Taste Model: Locked inside private corporate cloud servers.
The Reality: The feed is magically responsive because their embeddings models are world-class, but you are a tenant on their land. If you leave or get banned, you lose your audience, your archive, and the years of training you gave their algorithm.

Direct Publishing (RSS, Personal Blogs, Substack)

Identity & Content: Owned by the creator. You hold your email list or your domain.
The Taste Model: Non-existent.
The Reality: RSS and blogs gave us sovereign publishing, and Substack gave writers direct relationships with readers. However, neither solved algorithmic discovery without turning back into a centralized social network. You only see what you explicitly subscribe to, in chronological order. XkMato bridges this gap: a writer's blog feeds directly into their Publish Wallet, which then flows into an algorithmic timeline like xkmato.eth without a corporate middleman taking control.

First-Wave Decentralized Social (Nostr)

Identity & Content: Owned by the user via cryptographic keypairs and distributed relays.
The Taste Model: None, or fragmented across basic relay filters.
The Reality: Nostr proved that simple, signed JSON events passed through relays can create censorship-resistant communication without a blockchain. But because Nostr lacks a unified, private way to represent a user's deep taste, discovery relies heavily on who you already know.

Federated & Modular Feeds (Bluesky / AT Protocol)

Identity & Content: Owned by the user via Decentralized Identifiers (DIDs) and Personal Data Servers.
The Taste Model: Held by third-party cloud feed generators.
The Reality: Bluesky proved that a marketplace of custom feeds is possible. However, those custom feeds still rely on remote servers observing your public activity. There is no private, local-first vector vault living on your device that travels with you across the web.

The XkMato Protocol (xkmato.eth + Pyepa + XkMato Diary)

Identity & Content: Owned by the user via their cryptographic wallet and Publish Wallet.
The Taste Model: Owned by the user—stored as local embeddings inside their XkMato Diary.
The Reality: XkMato combines the sovereign publishing of RSS, the relay resilience of Nostr, and the modular feed marketplace of Bluesky, while finally putting the crown jewel—your mathematical taste profile—under your own lock and key.

6. The Open Challenge: A Common Language for Meaning

Building this won’t be trivial. The most fascinating technical challenge ahead is what engineers call the Vector Space Alignment Problem.

For two different algorithms—say, Pyepa and a competing open-source algorithm—to both understand the embeddings stored inside your XkMato Diary, they have to speak the same mathematical language. If one algorithm maps the world using a 768-dimension model and another uses a completely different 1,536-dimension model, the coordinates won't match.

Solving this requires the protocol to standardize one of two paths:

A Universal Open Embedding Standard: Just as the web agreed on HTML for text and JPEG for images, the XkMato protocol can adopt an open-weights, community-governed foundational embedding model as its "coordinate system" for human meaning.
Local On-Demand Translation: Your XkMato Diary stores your structured interaction history locally in clean, standardized schemas, allowing any algorithm you plug in to compute its own fresh embeddings on your device's hardware in the background.

With modern smartphones and laptops shipping with dedicated neural processing units (NPUs) capable of trillions of operations per second, running these models locally inside a wallet is no longer science fiction. The hardware is already in our pockets.

Reclaiming the Mirror

For fifteen years, we accepted a bad bargain: if we wanted the magic of an intelligent feed that surfaces brilliant art, writing, and video from strangers across the globe, we had to surrender our cognitive autonomy to a handful of Silicon Valley servers.

We don't have to accept that bargain anymore.

Every piece of the puzzle—cryptographic wallets, local-first vector databases, on-device AI inference, and relay-based publishing—now exists. By putting our embeddings inside an XkMato Diary, separating our Publish Wallets from corporate silos, and letting independent algorithms like Pyepa compete to serve us on open timelines like xkmato.eth, we can keep everything we love about social media while firing the landlords who currently run it.

Your mind’s vector is yours. It’s time to keep it in your own diary.