Notes
Research notes on spatial intelligence, agents, and artificial worlds, with methods and limitations from Walking Stick Labs.
AI Coding Agent Reliability
Failure modes, checkpoint metrics, and pilot findings from Iboga.
Spatial Intelligence in AI: Can an Agent Understand Its World?
Spatial IntelligenceWhat does spatial intelligence in AI mean? Explore how to test whether an AI agent can map a room, follow changing rules, and adapt when its world changes.
The Substrate Problem
MethodologyA behavioural claim about “LLM agents,” measured on a single model, is mostly a claim about that model.
Reflection Is Not Control
AgentsMaking an agent narrate its failures is an interpretability win, not a capability one.
Make the Model Commit
ReasoningFree-text reasoning is cheap and ungradeable. Make the model emit something a solver can falsify.
The Map Is Not the Prompt
SpatialA coordinate list can describe a world without making its connectivity easy for an agent to use.
Reading the Failure Logs
CraftWhat I actually look at when an agent fails.
Against the Fake Demo
PracticeWhy this site ships negative results and unfinished prototypes.
A valid generated world and a valid test of spatial understanding are separate achievements.