No Slides.No Hype.Just Raw
AI Engineering.
The underground community, hands-on courses, and execution stack for devs who build with autonomous agents, custom MCPs, and zero limits. Ship code with Claude Code, Cursor, OpenCode & Way of Teams.
Show up. Screen-share. Ship it live.
No pre-recorded webinars. Real builders, real terminals, real bugs — streamed from the underground garage at 2 AM. Sessions run on the ai-geeks & freaks Discord server.
The 2 AM Bug Fix
Live screen-share office hours. Bring a stuck agent, a half-broken prompt, or a failing eval — we debug it on stream until it ships.
Geek & Freak Showcases
10-minute lightning talks. Members present custom skills, Jido agent workflows, and apps built in OpenChamber. No slides — live terminals only.
Factory Workshop: MCP Deep Dive
Deep-dive technical masterclass. Build a custom MCP server end-to-end, wire it into the agent swarm, and deploy to a hosted workspace.
Spec sheets, not brochures. Outcome per module.
Every course ships with real infrastructure. Learn in class, then keep the workspace running while you build your own thing.
View Full CatalogueThe 30-Min Feature Run
Take a fuzzy feature request from idea to merged PR in one focused 30-minute loop — prompt, plan, code, review, ship.
Multi-Agent Systems & MCP
Architect a swarm of specialist agents that collaborate over a custom MCP server — and orchestrate them with Way of Teams.
Shipping a Custom Skill in 90m
Design, build, test, and publish a production-grade ai-harness skill — bundled, documented, and deployed to the registry.
Learn in class. Coordinate in Way of Teams. Build in OpenChamber. Ship in the Software Factory.
This isn't a place to just talk. The community plugs straight into the toolchain — every meetup, course, and shipped commit flows through one execution pipeline. Way of Teams coordinates the work; the Software Factory executes it.
Community
Discord, meetups & the live stage. Where the 2 AM bug fixes happen and skills get shared.
Curriculum
Academy courses mapped to the real toolchain — Claude Code, Cursor, MCP, Jido, evals.
Way of Teams
The AI Engineering platform that coordinates the work — tickets, standups, kanban, and every agent under one roof.
OpenChamber
Your coding tools — Claude Code, Cursor, OpenCode — running in cloud-hosted workspaces, always on and reachable from any browser, even your phone.
Software Factory
The one factory — the swarm that plans, codes, reviews, and ships your work through a gated pipeline.
Community
Discord, meetups & the live stage. Where the 2 AM bug fixes happen and skills get shared.
Curriculum
Academy courses mapped to the real toolchain — Claude Code, Cursor, MCP, Jido, evals.
Way of Teams
The AI Engineering platform that coordinates the work — tickets, standups, kanban, and every agent under one roof.
OpenChamber
Your coding tools — Claude Code, Cursor, OpenCode — running in cloud-hosted workspaces, always on and reachable from any browser, even your phone.
Software Factory
The one factory — the swarm that plans, codes, reviews, and ships your work through a gated pipeline.
One-line installer for the ai-harness CLI. Pipes you straight into the hosted workspace queue.
deno run -A https://raw.githubusercontent.com/Way-Of/aiharness/main/install.ts --install-cliA real factory, not a metaphor. Agents do the work. You launch, converge & observe.
Behind every shipped thing here runs the Super Simple Software Factory — a swarm of specialist agents (planner, coder, reviewer, eval) driven through a strict phase pipeline with hard gates. Success must be earned: every phase defaults to fail until it clears review. Runs are traced, learned from, and replayed in live visualizers.
- [01]planneridle
- [02]coderrunning · +412 LOC
- [03]reviewerrunning · 3 notes
- [04]mcp-bridgerunning · 8 tools
- [05]jidopending · skill
- [06]evalpending · gate
Hard gates
Every phase must clear review + evals before the run may continue.
Full traces
Every run recorded & replayed — learn what shipped and why.
The one rule: you launch, converge & observe — you never hand-implement. The factory's agents own the work.
Read the full deep-diveThe platform that coordinates the pipeline. Orchestrate multi-agent teams,MCP servers, and shipped code at scale.
Way of Teams is the AI Engineering platform that coordinates the factory. Kanban, tickets, standups, skills, notifications, AI chat — all backed by Ash Framework, Jido agents, and Oban workflows. No separate Node runtime. No trigger.dev. Pure Elixir. This site runs on it.
Kanban Boards
7-phase boards with drag-and-drop, 16 templates, real-time sync across tabs.
Multi-Tenancy
Company › Department › Team › Developer hierarchy with tenant isolation.
GitHub OAuth
Native OAuth2 via AshAuthentication — primary verified emails, zero config friction.
AI Chat + MCP
wo-agent SDK with SSE streaming, 6 tool extensions for tickets/standups/skills.
Oban Workflows
Staleness sweeps, daily snapshots, weekly releases, email/Telegram digests — all Elixir-native.
Jido Agents
TicketManager, SkillAutoUpdate, 3-layer Orchestrator (auto/notify/escalate).
Backlog
3- Spec: MCP Server
- RFC: Agent Orchestration
- Idea: Skill Registry UI
In Progress
3- Build: Custom Skill CLI
- Fix: WebSocket Reconnect
- Refactor: Kanban Drag
Done
3- Ship: GitHub OAuth
- Deploy: Chat Server
- Docs: API Reference
Community feeds Academy. Academy fuels Factory. Factory ships Code.
Shipped code feeds Community.
This isn't a linear funnel. It's a flywheel. Every shipped commit generates new meetups, new course material, new factory runs. The loop compounds — each revolution raises the baseline for the next.
Academy
Courses mapped to the real toolchain — Claude Code, Cursor, MCP, Jido.
Browse Courses →OpenChamber
Hosted workspaces pre-wired with the agent swarm. Spin up in minutes.
Request Workspace →Software Factory
Agent swarm executes — planner, coder, reviewer, eval. Hard gates, full traces.
Read Deep-Dive →Six principles. Zero compromise. This is how we ship.
These aren't values on a wall. They're constraints baked into the toolchain. The factory enforces them. The academy teaches them. The community lives them.
Launch, Don't Implement
You define the outcome. The factory's agents own the work — planner, coder, reviewer, eval. You converge & observe.
See the factory →Hard Gates, No Theater
Every phase defaults to fail. Review + evals must clear before the run continues. Success is earned, not assumed.
Read the gates →Learn From Every Run
Full traces recorded & replayed. Failed runs teach more than successes. The eval dashboard turns every run into training data.
Explore traces →Real Infra, Not Slides
Hosted workspaces (OpenChamber), MCP servers, Jido agents, Oban workflows — all running on real PostgreSQL, real Elixir, real OTel.
Request workspace →Outcome Per Module
Academy courses ship with spec sheets, not brochures. 30-min feature run. Multi-agent MCP. Custom skill in 90m. Measurable outcomes.
Browse courses →The Loop Compounds
Shipped code → new meetups → new curriculum → new factory runs → more shipped code. Each revolution raises the baseline.
Watch the flywheel →Want the full technical breakdown?
Read the Software Factory Deep-DiveCustom-built extensions. Serious tools. No dead links.
We build on the real, open-source AI toolchain. These are the engines we extend, configure, and run — with links that actually go somewhere.
ai-harness
Our AI engineering harness — 51+ battle-tested skills, 14 agents, and workflows wired into 7 AI coding tools. Install once, use everywhere. MIT / Apache-2.0.
- ▸one installer wires skills + agents + commands into every AI tool
- ▸runs on OpenCode, Claude Code, Pi, Codex, Gemini, Wo Coder, Antigravity
OpenCode
The open source AI coding agent we drive from. Terminal, desktop, or IDE — 75+ LLM providers through models.dev, plan mode, LSP, and your existing Copilot/ChatGPT subs.
- ▸bring your own model or reuse existing subscriptions
- ▸Plan mode, /undo, /share, multi-session — free and open source
Pi
The minimal, extensible coding agent by Mario Zechner. Adapt Pi to your workflow — not the other way around — via extensions, skills, prompt templates, and packages.
- ▸15+ LLM providers · tree-saved sessions · MIT open source
- ▸install: curl -fsSL https://pi.dev/install.sh | sh
Real commits. Real builders. Zero vapor.
Live stream of projects shipped through ai-harness and hosted workspaces. Every entry is a real git commit — refresh and watch the counter climb.
View Full Archive- claude-code92%
- cursor74%
- opencode51%
- jido+mcp38%
Stop watching.
Start shipping.
Join the underground community running the AI Software Factory. Get meetup invites, course access, and a hosted workspace seat — before the cohort closes.