About · A Personal Atlas

From known
to unknown

Unknown Nodes — handing the compass back to the reader: a knowledge graph woven from real Wikipedia links, leading from what you know toward what you don't.

v 0.1 · A Creative Work 2026 · 06 Solo Maker Project
01 / Manifesto

Manifesto

In the age of the recommendation algorithm, hand the compass back to the reader.

Over the past decade, "what to look at next" has increasingly been decided for us by recommendation algorithms. The longer we scroll, the more we get used to being fed — and the easier it becomes to forget the question we actually wanted to ask. Unknown Nodes is not another recommender. It is a tool that runs the other way: it puts the steering wheel — "where do I go next" — back in the reader's hands, and answers using the plainest material there is — hyperlinks that genuinely exist on Wikipedia.

01

Restore agency

Every step of the path is the reader's own choice. The AI never decides what you look at next.

02

Real links, zero hallucination

Every term on the path comes from a real internal Wikipedia link — not something a model made up.

03

From known to unknown

No need to master a new field's jargon first. You migrate there step by step, from concepts you already know.

02 / Problem

The problem it solves

Unknown Nodes focuses on one very specific spot on the learning curve: from 0 to 0.5. This stretch isn't fully covered by beginner tutorials, and it isn't something a search engine settles with a single article — it's the process of "building a coordinate system inside an unfamiliar field."

What "0 → 0.5" actually means
0

Completely unfamiliar

You don't even know what to search for. The terms are isolated nouns with no connections.

¼

Can name things

You start telling a few key terms apart, but don't yet know how they relate to one another.

½

A sense of structure

Nouns join into lines, lines into a plane — and you can start asking questions with direction.

Terms the author has been searching a lot lately include MCP RAG SaaS MaaS RaaS. Each new search is like lighting up one more star on an unfinished star map — and what Unknown Nodes sets out to do is weave those scattered dots into a visible constellation.

03 / Origin

Why build this

① A term is the key into an unfamiliar field

"What is xxx?" is the standard opening move for making sense of an unfamiliar field. But search engines usually hand back isolated definition pages. Search enough, and you notice on your own: "Oh — RAG, LLMs, vector databases, and embeddings are all one cluster." That process of connecting dots into lines, and lines into a plane, is itself the joy of learning — yet no tool serves it directly today.

② The cognitive load of cross-field leaps

Leaping straight into an unfamiliar field triggers a flinch: too many terms, no idea where to start reading. But if you can begin from a concept the reader already knows and weave a path across, with every step resting on the familiar, the experience is entirely different. In cognitive psychology this is called scaffolding.

The familiar is a foundation. You can't learn suspended in mid-air, but you can lay bricks toward the unknown from one you already know.
04 / Product

What it is, roughly

An interactive web page. The reader enters two Wikipedia terms: a known term (input_01) and a term they want to explore (input_02). The system calls Wikipedia's real link graph, renders related terms around each, and the reader picks step by step, walking the path from input_01 to input_02 themselves.

unknown-nodes.online/explore
I know
React
I want
MCP
Weave
Figure 1 · Unknown Nodes graph
Ink = neighbors of the known term · warm = the path you wove · blue = neighbors of the unknown term
The path walked so far
React API JSON LLM Tool Use MCP

The core promise

Every term is a node and connection that genuinely exists on Wikipedia — not something the AI conjured up. Each choice corresponds to a real internal link — meaning the path you walk is verifiable and keeps reading onward, rather than an explanation a model "rounded off" for you.

Four-step interaction

Enter two anchors

The reader enters a known term and a target term; the system checks both are real Wikipedia entries.

Render the neighbors

It calls the Wikipedia link graph and shows each anchor's most-linked outgoing terms as choices.

Weave step by step

The reader picks a term as the next step; the system expands its neighbors, looping until it reaches the unknown.

Leave a constellation

When done, the reader has a private path of real links — one they can save, share, and walk again.

05 / Audience

Who it's for, and when

Who

When

In the margins of the day — on the subway, over lunch, before sleep — you want to understand some unfamiliar term you keep bumping into, but you're not about to sit down with a whole intro book or a ten-thousand-word essay.

06 / Status Quo

Without it, how do people cope today?

A

Searching short-video apps

Search the unfamiliar term on RED / Douyin and get flooded with short videos — information reprocessed by creators, mostly unchecked for facts.

B

Chatting with an LLM

Ask Doubao / DeepSeek / Kimi for a walkthrough. But getting plain, layperson language takes careful prompt editing — costly in time, and never free of hallucination.

Figure 2 · Five axes: existing options vs. Unknown Nodes

The three don't replace each other — LLMs are good at explaining, short videos at atmosphere, and what Unknown Nodes adds is something else: making "which term do I look at next" a deliberate choice by the reader again, rather than something pushed to them by someone else.

07 / Why It Matters

Value & meaning

The attention ledger of the algorithm age

As of March 2025, the three big short-video / content platforms are consuming users' time budgets at a staggering scale[1]:

1.00B
Douyin monthly active users
46.54h
Douyin monthly hours per user
573M
Kuaishou monthly active users
235M
RED (Xiaohongshu) monthly active users
Figure 3 · Monthly hours per user across the three platforms (2025-03)
Source: Sina Finance, 2025-05[1]. Douyin users average over 1.5 hours of short video per day.
The recommendation algorithm brings platforms endless opportunity — while stripping users of their attention and their agency.

Unknown Nodes doesn't expect to fight that curve. But it wants to carve out, alongside the algorithmic feed, a small test plot where the reader picks their own road — retraining the awareness of "which direction do I actively explore next." That sliver of awareness may be the seed of resistance to being fed.

08 / In One Line

In one line

Unknown Nodes (织 · 点) = Wikipedia's real node network + the reader's curiosity + one visible compass.
09 / At a Glance

At a glance

DimensionWhat Unknown Nodes answers
ManifestoRestore the reader's agency; hand the compass back to them.
ProblemThe fast map-building phase of a new field, from 0 to 0.5.
Core interactionEnter a known term + a target term → weave a path of real links.
Data sourceOnly real Wikipedia hyperlinks; no AI fabrication.
AudienceCurious people who read across fields often.
OccasionSpare moments, to grasp some stray unfamiliar term.
DifferentiatorDoesn't explain (leave that to LLMs), doesn't push (leave that to algorithms) — it only weaves paths.
North-star metricComplete paths woven per user / month.

— Made with curiosity, on a quiet afternoon. —