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