The research bench / 3 active experiments

Not everything is a polished product.

My laboratory for agent protocols, memory graphs, and experimental tools. The messy experiments here are where my production architectures get proven.

X-01
EXPERIMENTFieldoField-data capture experiment for physical work.Next.js
Postgres A-02
EXPERIMENTRavenAutonomous ops agent bench: memory, tools, and fallback.Hermes
OpenClaw
SQLite
Tailscale A-03
EXPERIMENTGraph MemoryRelationship memory between agents, tools, and knowledge.Postgres
pgvector
Hermes

Decision bench / how I choose

Tradeoffs, weighted in the open.

The weighted heuristic rules I use when choosing frameworks, model routing, and storage layers. A decided option is a fact; an open one is an experiment.

DATABASE

Postgres

Supabase

Firebase

AGENT FRAMEWORK

OpenClaw

LangChain

AutoGPT