Skip to content
Home  /  Neobiome Intelligence  /  Neobiome Knowledge
Neobiome Intelligence · Knowledge

What we read, how we checked it, and what we built

Every conclusion on this site rests on a body of reading. This page accounts for it: where it came from, what was done to it before it was trusted, and the searchable library it became.

Explore it yourself

The knowledge library

Every source behind this research is a page in a browsable, searchable library, with a graph view showing how the pieces connect. Like the Engine it is a working beta, and a public version is planned. The rest of this page explains how it was built.

Enter the knowledge library ↗ Opens in a new tab
Beta · in active development
Full-text searchGraph viewBacklinks
The idea behind it

A working implementation of an emerging pattern

This knowledge base did not start as an original idea about knowledge. It is a practical build of a pattern with a long lineage, taken further where the research needed it.

1945
Vannevar Bush
The Memex

A personal, curated store of documents with associative trails between them. The connections mattered as much as the documents.

2026
Andrej Karpathy
The LLM wiki

An LLM incrementally builds and maintains the wiki, so the connections stay current. You curate and question; it does the bookkeeping.

2026
This knowledge base
The pattern, made rigorous and public

The same idea, plus a verification regime it lacked, turned from a personal tool into a public research artefact.

What we added. Karpathy's pattern trusts the LLM to summarise faithfully and keeps the wiki private. This project could do neither. It added a verified-source-only rule, a typed-page schema with an independent verification pass, and it opened the result to the public as research evidence. The adaptation is where the contribution sits.
1 What we read

More than five hundred sources, filed by kind

The corpus is filed by what kind of thing each source is, not by topic. That sounds bureaucratic and turned out to matter: a district plan and a peer-reviewed paper carry very different weight, and the file type is the first signal of how far to trust a claim.

TypeWhat it holdsHow it is weighted
litAcademic and conference papersHighest. Cited directly
regDistrict plans, standards, regulationsAuthoritative for what is permitted
ds / rdDatasets and technical datasheetsPrimary for figures
crCompiled research, including AI synthesisNever cited alone; must resolve to a primary
urlWeb capturesSupporting, dated at capture
mmVideo and audio, with transcriptsSupporting
otEverything elseCase by case
How it was built

14 weeks, 544 source pages

Every source page in the corpus, by the week it was added. The line is the running total. Both are read out of the project's version history rather than a tally kept by hand, so the shape is what actually happened, including the quiet weeks.

035701051400150300450600retrieval and checking run in parallel from here544MayJunJulAug
sources added that week running total as at 17 August 2026

The corpus holds 544 source pages, of which 535 are published to the library above. The busiest week added 138, in the fortnight the method changed from retrieving sources one at a time to running retrieval and verification in parallel. Counting starts 18 May 2026.

2 How we processed it

From a file to a trusted claim

Following the pattern above, an LLM does the bookkeeping and a human does the curating. Every source becomes a typed page with structured metadata, linked to the concepts and claims it supports. Automated checks run over the whole corpus and refuse to pass if a page is missing required fields, or if a raw file exists with no page pointing at it.

Ingest

Filed, typed, duplicate-checked by content hash.

Verify

A separate pass, so the writer is not also the checker.

Link

Tied to the concepts and claims it supports.

Lint

Whole-corpus checks; no orphan files, no missing fields.

The rule we added to the pattern
It was not designed. It was forced by an error.

Karpathy's pattern trusts the LLM to summarise faithfully. Ours learned not to. Partway through, AI assistants were caught attributing figures to the right organisation but the wrong document. Plausible, well-formatted, and wrong. The response was a verified-source-only rule: a figure may be used only if the underlying primary document has been retrieved and the exact number confirmed in its text. Not the summary. The document.

On the use of AI This project was substantially operated with AI assistance, under the conditions of ethics approval MTF.8888.294. The verification regime is the part worth publishing. A method that caught its own failure and hardened against it is more trustworthy than one that reports none.
3 What we built

A knowledge centre of two kinds of page

It is a knowledge centre, not a document store. It is built from two kinds of page. Every ingested document gets its own page: a summary, its metadata, and the claims it supports, with the source file itself referenced rather than hosted. On top of those sit synthesis pages, which draw many sources together into a topic, renewable heating, food production, the self-sufficiency method, and connect the concepts that matter to the project.

Source page · one per document

Kerr (2024), community energy in Aotearoa

Type Masters thesis (lit)
Quality Medium
Source file Referenced, not hosted
Synthesis page · drawing sources together

Micro-hydro

Draws on Multiple source pages
Role The project's own synthesis
Feeds the Engine and the technology map
500+Source pages
60+Synthesis pages
10Domains
10SSI indicators
9Case studies
Sources are referenced, not hosted The written pages are ours to publish. Most of the source files behind them are not: much of the academic literature sits under publisher copyright. Every source is identified so it can be found through a library or publisher, openly-licensed ones are linked directly, and none of the files themselves are served from this site.
References

Sources for the method

Karpathy, A. (2026). LLM Wiki. gist.github.com/karpathy.

Bush, V. (1945). As We May Think. The Atlantic.

Enter the knowledge library ↗