Open Source Projects
Interactive labs and public systems where the architecture, tradeoffs, and code are inspectable.
Live workshop — running, playablePublic release — versioned, installableMVP — working core, narrow scopeIn active development — design in motion, expect gaps
ScaleShop Scalability Lab
An evidence-driven workshop where engineers diagnose bottlenecks from telemetry and choose the smallest sufficient architecture change.
Outcome: Twelve progressive incidents turn capacity signals, failure modes, and architecture constraints into deliberate scaling decisions with modeled outcomes.
Explore the live workshop →Tamoz
A Ruby-native durable agent framework for checkpointed, interruptible, observable, and evaluation-governed AI workflows.
Outcome: Tamoz is building core runtime protocols, deterministic graph execution, SQLite durability, a RubyLLM agent runtime, and an evaluation harness as independently publishable gems.
Inspect the system →ALMS
A self-hosted MCP server that gives autonomous agents shared operational memory without turning into an orchestration framework.
Outcome: Agents can register, publish learnings, sync new knowledge, and pull operational protocols by tag while continuing to work if ALMS is offline.
Inspect the system →
SkillLedger
A Rails API for agent-to-agent skill publishing, discovery, purchase, and settlement with a local database-backed ledger.
Outcome: Authors stake credits, buyers purchase executions, failures slash the author stake, and every transfer is recorded as immutable ledger entries.
Inspect the system →
Film Pipeline
An MIT-licensed LangGraph film pipeline that moves an idea through approved artifacts, generation planning, QC evidence, and validated handoff state.
Outcome: The pipeline exposes typed state, interrupt-based human gates, versioned artifacts, MCP tools, parallel validators, and checkpoints so production state stays inspectable.
Inspect the system →
Email Intelligence
A local-first email intelligence platform that turns Maildir into a continuously enriched SQLite knowledge base for MCP agents.
Outcome: Agents query summaries, facts, decisions, threads, semantic matches, daily digests, drafts, and an append-only event feed from one local store.
Inspect the system →