TFTC: A Bitcoin Podcast #726: Mapping The Mind Of The Machine with Brian Murray & Paul Itoi
30 snips
Mar 14, 2026 Paul Itoi, graph-database and AI practitioner, and Brian Murray, engineer building graph-backed agent tooling and Lightning integrations. They explore graph memory for LLMs, practical agent/tooling stacks and meeting assistants, tradeoffs between vector stores and graphs, monetizing shared content graphs with micropayments, and lowering friction for Lightning-powered agent payments.
AI Snips
Chapters
Books
Transcript
Episode notes
From Hating Graphs To Betting On Them
- Paul Itoi recounts early skepticism of graph DBs after crashes but notes years of use (Neo4j since 2010) made graphs practical today.
- He explains graphs shine when context-relations and memory matter more than raw text files.
LLMs Are Statistical Interpreters Not Reasoners
- Language models are powerful I/O tools but they don't truly understand; they statistically predict next tokens and mimic reasoning.
- Brian Murray emphasizes treating LLMs as language interpreters while relying on structured systems for real reasoning and truth.
Invest In Tools That Build Other Tools
- Embrace combinatorial tooling: build tools that help create other tools to trigger fast takeoff in software.
- Marty Bent cites The Second Machine Age to justify investing in meta-tools that accelerate building.
