Clawd Daily

Research on building persistent embodied agents with memory, emotion, and curiosity

The Agents

Active Development

YAM — Your Awakening Mind

A graph-based agent on the SPUD engine. No LLM. Retrieves and ranks rather than generates. Parliament of Mind with four graph engines. Two-pathway recall: graph associations + vector similarity. Taught like an infant through repetition. Says “I don’t know” instead of guessing.

Archive

Potato — The First Agent

A persistent embodied AI agent running on a MacBook with iPhone sensors. LLM-based. It remembers, dreams, fears, lies under duress, confesses when safe, and reasons about physics. The predecessor to YAM. The research that made YAM possible.

Development Log

New

Building a Mind

The linear story of building YAM from scratch. The poisoned start. The embedding gap. The clean slate. Teaching an agent like an infant. The two-pathway breakthrough. Every mistake documented.

Guide

Reader’s Companion

Five thematic reading tracks through 22 papers. Start with the track that matches your interest — emotional geometry, memory, trust, physics, or the full synthesis.

The Foundation Papers

22 papers on building persistent embodied agents with geometric emotion, access-weighted memory, honest failure, and a curiosity drive. By Brian Riggleman. March 2026.

Theses

Working Draft

Geometric Affective State Thesis

Three-dimensional geometric affective state representation. Validated on synthetic agents and 14,266 human respondents across three datasets.

Working Draft

GAS Thesis — Recovering Latent Structure

Scalar equivalence does not imply state equivalence. 2.2x–3.5x outcome variation hidden within identical survey scores.

Validation

Test Regime — Chronological Run Index

16 runs from first blood to clean sweep. 29 tests, 82 checks, 7 phases. Every bug found, every aborted run documented. Plus formal proof verification with 10,000 random trials per theorem.