On April 4, 2026, I brought Way Back Home into the CompleteTech GitHub organization after working through the experience myself. I am dating this from the repository push into CompleteTech, not from the upstream initial commit history.
What I completed
Way Back Home is a Google Cloud AI workshop built around a rescue narrative. The participant starts as a stranded explorer and has to use AI systems to identify themselves, analyze their surroundings, locate coordinates, and coordinate an extraction.
That framing is useful because it makes the technical pieces feel operational. This is not a generic demo where an agent answers a prompt in isolation. It is a workshop environment where the participant has to move through a chain of evidence, tools, state, and infrastructure.
Why it mattered to me
When I posted the result on LinkedIn, I described the ending as being home, with the multi-agent system on Cloud Run acting as my eyes and ears until the end. That is the right way to read the project: the interesting part is not only the destination, but the way the system maintains enough context to get there.
The repository describes the workshop as a hands-on platform for building intelligent agents while rescuing a stranded space explorer. The stack behind that story includes Next.js, Three.js, FastAPI, Firestore, Firebase Storage, Cloud Run, Vertex AI, Gemini, Google ADK, MCP, Cloud Build, and Artifact Registry.
That combination lines up with the kind of work I keep coming back to at CompleteTech: agent systems that have to operate through real infrastructure, not just produce text.
The workflow
The workshop is structured as levels. Level 0 creates the explorer identity. Level 1 uses multi-agent crash-site analysis to locate the participant. Later levels expand toward SOS processing, coordination, rescue planning, and event-driven agent workflows.
Underneath the story, the architecture is straightforward in a good way. A backend API on Cloud Run tracks participants, stores evidence, confirms locations, and manages events. Firestore and Firebase Storage hold the operational data. The frontend renders a 3D mission view so progress is visible instead of buried in logs.
The lesson I took from it is simple: an agent workflow becomes much easier to trust when it has a visible state model. Identity, evidence, location, progress, and event state are all things a human can inspect.
What it demonstrates
Way Back Home is playful, but it points at a serious pattern. A multi-agent system needs orchestration, durable state, clear tool boundaries, and a way to show the operator what happened. The story makes the workshop approachable. The architecture makes it useful.
That is also why I connected it to CompleteTech publicly. The same principles apply outside the workshop: when AI is part of an operational workflow, the surrounding system needs to preserve evidence, expose progress, and make the handoff between agents and infrastructure legible.
Why I saved it
The Completech repository was created on April 4, 2026, with the main branch pushed shortly after the public LinkedIn post. I kept the upstream history intact, but the date I am using here is the CompleteTech-side repository event, because that is when I saved the work into our public body of evidence.
That distinction matters. The upstream project began earlier. My public artifact is the moment I completed the experience, posted the result, and preserved the repository under CompleteTech for future reference.
Sources
CompleteTech Way Back Home repository
Written by Tim Gregg, founder of CompleteTech LLC – Innovation at Every Integration.