No. 00 A second brain that maintains itself

Your knowledge stops evaporating the moment you stop paying attention to it.

Knownest is an AI librarian for your personal wiki. It files what you read, links it to what you already know, and tells you when your understanding has gone stale — so your notes don't rot.

Onboarding a small first cohort · 2026

No. 01 Recognition

You built a second brain. Then you watched it rot.

Notion, Obsidian, Roam — you set them up with real intent. A few months later it's a graveyard of saved-but-never-revisited pages you don't trust enough to rely on.

FILE 01 · UNTOUCHED

The clipped article you never re-opened

Saved in a burst of intent. Filed under “read later,” which both of you know means never.

Last opened · 8 mo
FILE 02 · ORPHANED

The note that links to nothing

A real insight, stranded on its own page. Nothing points to it, so you never find it again.

Backlinks · 0
FILE 03 · CONTRADICTED

The claim your newer reading disproved

Two pages now say opposite things. You don't remember which one you still believe.

Conflict · unresolved

The pages didn't fail. The upkeep did — and upkeep was always the part you were never going to keep doing by hand.

No. 02 The reframe

The problem was never capture. It was never search. It was maintenance.

A decade of tools bet on faster capture. RAG bet on better search. Both skipped the work that actually keeps a knowledge base alive.

Capture was never the bottleneck

You already saved everything. The graveyard is proof you can capture. Saving was easy; it's the saved pile that rots.

Search just reads the rot back to you

Ask a stale, contradictory base a question and you get a confident, stale, contradictory answer. Retrieval doesn't reconcile anything.

Maintenance is the work humans abandon

Reconciling each new source against everything you already know, and pruning what's gone stale. Nobody sustains it by hand — and it's exactly what an LLM now does well.

No. 03 The librarian at work

An AI that does the upkeep, in the open.

Four things it does on its own — and one promise that makes letting it run feel safe.

PILLAR 01

Files & links on ingest

Add a source and the agent reads it, files it on the right pages, rewrites those pages to absorb the new claim, and lays cross-references to what you already know — in one pass.

PILLAR 02

Audits its own health

It continuously surfaces contradictions, stale claims, orphaned pages and missing links — with citations — without being asked. The half of the job you never do.

PILLAR 03

Versions every change

Each edit is diffed, attributable and undoable. The substrate is a full-history record, not a mutable blob — so nothing it writes is ever a one-way door.

PILLAR 04

Hands back structure

What you get is a living, interlinked wiki that thickens with every interaction — not a stream of answers that vanish the moment you close the tab.

No. 04 The objection, answered

Let it run. Trust nothing it can't show you.

Yes — an AI writes to your knowledge base on its own. That only works because every change it makes is visible, attributable, and reversible.

CHANGE #4,212 · Model Collapse agent · 2.1s ago
Training on synthetic output erodes the tails of the distribution.
The long-run effect is still debated.
+It compounds across generations and is hard to reverse. [cit. 2406.04]
Attributed Diffed ↩ Undo this change

Every write is a record you can inspect, not a silent mutation you have to hope was right.

A.

Diffed

See exactly what changed, line by line, before and after.

B.

Attributed

Every claim traces to the source that justified it.

C.

Undoable

One click reverts any edit. No change is ever permanent or hidden.

No. 05 The payoff

The longer it runs, the more your base is worth.

Every pass deepens a structure tuned to your sources and how you think. What you learn finally adds up — and walking away would mean abandoning months of integration no export can rebuild.

Start a base that compounds
WEEK 1
MO 1
MO 3
MO 6
MO 12

↑ Reconciled links & trusted pages per topic

Early access · 2026 No. 06 Request access

Your knowledge, kept alive.

We're onboarding a small first cohort of researchers, analysts and writers. Leave your email and we'll reach out.