Skip to content

GenerationUX/MinedDeep

Repository files navigation

MinedDeep

A structured, interlinked knowledge base built on the Karpathy LLM Wiki pattern, dedicated to the complete published intellectual output of Sir Demis Hassabis (b. 1976) — co-founder and CEO of Google DeepMind, 2024 Nobel Laureate in Chemistry.


What This Is

This wiki is a synthesis layer over a curated corpus of Hassabis's peer-reviewed papers, authored essays, formal lectures, long-form interviews, and significant public statements. It is maintained by Claude Code using the Karpathy/Farzaa wiki-generation pattern and the domain schema defined in CLAUDE.md.

The goal is not a biography. It is a navigable intellectual map: what Hassabis argued, when he argued it, how his positions developed, and where the record is incomplete.


Corpus Coverage

The corpus spans three phases:

Phase Years Character
PhD / postdoc 2007–2010 Cognitive neuroscience. Hippocampus, memory, imagination. UCL under Eleanor Maguire.
Early DeepMind 2010–2018 Deep reinforcement learning. AlphaGo. The neuroscience–AI bridge.
AlphaFold era and beyond 2018–present Protein structure. AI for science. Safety/governance. Nobel Prize. AGI theory.

~47 source documents in the initial corpus, including:

  • 30 peer-reviewed papers (Nature, Science, Neuron, PNAS)
  • 3 authored essays / reviews
  • 9 long-form interviews (with transcript access)
  • 4 blog / social posts
  • 1 doctoral thesis (UCL, 2009)

Architecture

raw/corpus/       Immutable. The source files. Claude reads; never writes here.
wiki/             LLM-maintained. Claude writes; human reads.
  _index.md       Page catalog by type.
  _log.md         Append-only operation log.
  _axis_index.json  Taxonomy axis lookup table.
  _corpus_log.json  Per-file read and absorption tracking.
  pages/          All wiki articles — flat, slug-named.
assets/           Diagrams, timelines, exported reference figures.

How to Use

Reading the wiki

Browse wiki/pages/ or use wiki/_index.md as the entry point.

Page types:

  • paper-- · lecture-- · interview-- · essay-- · post-- — individual source pages
  • theme-- · project-- · period-- — synthesis pages
  • collaborator-- · venue-- · claim-- — navigation pages
  • gap-- — documented gaps in corpus coverage

Extending the corpus

  1. Create a new file in raw/corpus/ named [slug]--Hassabis.md
  2. Include a ## §0 block with all 8 required taxonomy fields
  3. Include full text or best available excerpt of the source
  4. Run /wiki index [slug] in a Claude Code session
  5. Run /wiki absorb [slug] to generate the wiki page

Running Claude Code

Open Claude Code at the root of this directory. It reads CLAUDE.md automatically. Then run:

/wiki index all        — index all unprocessed corpus files
/wiki absorb [slug]    — generate a wiki page for one file
/wiki synthesise [axis]=[value]   — generate a synthesis page
/wiki status           — report current coverage

Source File Format

Every file in raw/corpus/ should follow this template:

---
# [slug]--Hassabis.md
---

## §0 Taxonomy & Identity

format: [controlled value]
domain: [controlled value]
era: [controlled value]
venue: [controlled value]
theme: [controlled value(s)]
project: [controlled value or none]
co-authors: [names]
impact: [controlled value(s)]

<!-- optional fields -->
claim: [if applicable]
influences: [if applicable]
contradicts: [if applicable]
open-questions: [if applicable]
notable-quote: [≤25 words, verbatim]

---

## §1 Full Text / Excerpt

[Full text, abstract, or best available excerpt of the source.]

Wiki scaffold version: 1.0
Upstream patterns: Karpathy LLM Wiki, Farzaa wiki-gen skill
Domain: Hassabis intellectual corpus
Scaffold date: 2026-05-13

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors