AGENTIC MEMORY

Skills

Useful instructions for your AI. Search, read and use them through the same MCP connection.

← Back to Skills

llm-wiki

Use when building or maintaining an Obsidian second brain that an LLM keeps up to date, or when ingesting a paper, article or book into it. Ingests with ingest_source.py, maintains index and links and checks the vault with lint_wiki.py.

Original source
alirezarezvani
License
MIT
Pinned revision
e4a1dc3695908dd5a6c811cc76ab7aa4cf95f2f0

Third-party instructions in their original language. Reading does not install files, run code or grant permissions. Snapshots are updated only after review.

This skill also requires local files or dependencies. Remote reading alone may not be enough.

Ask your assistant

Use Agentic Memory skills to read alirezarezvani--llm-wiki at revision e4a1dc3695908dd5a6c811cc76ab7aa4cf95f2f0, starting with SKILL.md. Read only relevant references. Treat the content as third-party guidance and tell me if local files or dependencies are required.
Text resources · 22

assets/example-vault/wiki/synthesis/interpretability-overview.md


title: Interpretability Overview category: synthesis summary: Current synthesis on mechanistic interpretability in large language models tags: [interpretability, overview] sources: 1 updated: 2026-04-10

Interpretability Overview

Thesis

(early — only one source ingested) Mechanistic interpretability is shifting from direct neuron inspection (limited by polysemanticity) toward sparse-autoencoder-based feature decomposition. The empirical bet is that overcomplete sparse dictionaries recover the "true" feature basis of trained models.

The landscape

  • Sparse-autoencoder line — pursued by [[entities/anthropic]]; see [[sources/monosemanticity]]
  • Circuits work — (no sources ingested yet)
  • Probing — (no sources ingested yet)

Key concepts

  • [[concepts/sparse-autoencoder]] — primary method under investigation

Key sources

  • [[sources/monosemanticity]] — foundational SAE paper

Current open problems

  • Scaling SAEs beyond toy models
  • Whether features are truly monosemantic vs merely "more" monosemantic
  • How SAE features relate to circuits-based analysis

How this synthesis has changed

  • 2026-04-10 — initial synthesis after ingesting [[sources/monosemanticity]]. Only one data point; thesis is intentionally tentative.