Cloud UX AI: The Organizational Blueprint
The Intent: Scaling Success
Following the success of the Gemini Enterprise Agent Builder, several major teams—including CoreUX, Corp Eng, and Google Material—sought to replicate our high-velocity collaborative environment. The intent of Cloud UX AI was to distill our successful “Jetski + GitHub Enterprise” architecture into a single, production-grade Blueprint Repository (go/cloud-ai-ux-repo).
The System: Scaffolding for Agentic Teams
The blueprint provides a “push-button” bootstrap environment that codifies the “New Way” of integrated functional prototyping. It includes:
- Unified Repository Structure: Standardized directory layouts for agents, tools, and UI components (using GM3).
- Pre-Configured CI/CD: The Cloud Build -> Cloud Run pipeline I developed for the Agent Builder, now generalized for any new organizational project.
- Embedded Prototype Tooling: Includes the Visual Inspector, Annotate (Draw), and Commenting systems as first-class citizens of the scaffold.
- Governance Standards: Built-in branch protection and PR-only invariants that ensure every new AI project starts with a hardened foundation.
UX AI CLI: Automated Onboarding Engine
As more teams began adopting the Cloud UX AI Blueprint, manual onboarding became a bottleneck. To solve this, I built the UX AI CLI (internally distributed as vibe)—an automated engine designed to handle the complex technical setup required to get a new team live on our collaborative platform.
The Function: Deterministic Onboarding
The CLI transforms the technical setup from a manual checklist into a deterministic execution flow. By automating the “physics” of onboarding, it moves the barrier to zero:
- Adaptive Vibe (Skill Level): Adjusts the toolkit features and guidance based on the user’s comfort level with IDEs and terminal workflows (e.g., Novice, Medium, Expert).
- Environment Bootstrapping: Automatically installs essential dependencies like Jetski and GGH (Google GitHub), and configures SSH Keys deterministically.
- Repository Initialization: Clones the blueprint and sets up the project organization within Google’s internal GitHub Enterprise.
- Infrastructure Configuration: Automates the creation of Google Cloud Projects and the necessary IAM roles for the CI/CD pipeline.
The Multiplier: Scaling the Architect
By automating these error-prone steps, the CLI acted as a multiplier for my own impact. It allowed me to scale the Staff Architect vision across multiple departments without being a manual bottleneck for every new team. It transformed a career’s worth of architectural knowledge into a repeatable organizational protocol, allowing any UX Engineer to bootstrap a high-fidelity agentic project in minutes.
Figure 2.0: UX AI CLI (vibe) — Automated Onboarding & Migration Protocol
The Multiplier: Cross-Departmental Velocity
By providing a standardized scaffolding, I eliminated the “zero-to-one” friction for other departments. Teams no longer need to spend weeks architecting their own deployment pipelines or onboarding protocols.
- Immediate Iteration: Teams fork the blueprint and focus on their specific agentic logic and UI variants from day one.
- Shared Skills: Skills published within the Cloud UX AI organization are immediately available to all blueprint-based projects, creating a cross-Google ecosystem of agentic capabilities.
The Human Factor: Collaborative Bootstrap
The blueprint is more than code; it is the physical embodiment of the Onboarding Protocol I developed. It includes the same Git-based rulesets and collaborative standards that allowed designers to transition into architects. By adopting the blueprint, teams also adopt a high-fidelity engineering culture where Review is the primary differentiator.
Technical Deep Dive: The Vibe Orchestrator
At the core of the blueprint is the Vibe Orchestrator (cloud-ai-ux), a specialized CLI tool designed to manage the lifecycle of agentic projects. It employs a Physicalization Engine that copies modular skills and workflows directly into target repositories, ensuring absolute portability and deployment readiness.
Adaptive Skill Registry
The orchestrator manages 11 Core Skills categorized by adaptive tiers (Novice, Medium, Expert), allowing teams to scale their technical complexity as they grow.
| Tier | Feature | Primary Skill | Logic & Impact |
|---|---|---|---|
| Novice | Scaffolding | gm3-vibe | Instant injection of Material 3 components and Vite-based HMR environments. |
| Novice | Visual Sync | vibe-apply | Bridges the browser-to-code gap by syncing real-time CSS/layout tweaks back to source files. |
| Novice | Zipline | deploy-zipline | Facilitates 1-click sharing of static snapshots via internal short-links. |
| Medium | Feedback | comments | Implements a Firebase-backed commenting system directly into the prototype UI. |
| Medium | Annotation | visual-drawing | Adds a high-fidelity drawing overlay for visual walkthroughs and screenshot captures. |
| Medium | Deployment | vibe-deploy | Automates the creation of Cloud Run services and custom Cloud Build triggers. |
| Expert | PR Reviewer | pr-reviewer | Automatically records a “Visual Diff” video of UI changes to accompany Pull Requests. |
| Expert | Reaper | vibe-reaper | A governance tool for the democratic cleanup of stale sandbox branches. |
| Expert | Identity | depot-auth | Automates SSH/Git machine registration for seamless GHE/internal access. |
| Expert | Auditor | ui-copy-auditor-vibe | Uses LLMs to audit prototype copy against Cloud AI branding guidelines. |
| Expert | Demo Rec | vhs | High-fidelity terminal recording for technical walkthroughs. |
Figure 2.2: Automated Deployment via Zipline + Jetski (Staff Orchestration)
The Vibe Registry: A Distributed Ecosystem
To address the bottleneck of maintaining and sharing foundational libraries like Google Material 3, I architected the Vibe Registry within the UXE organization on GitHub Enterprise. This Google-wide effort transforms the blueprint from a static template into a dynamic, bidirectional ecosystem.
- Distributed Ownership: Teams across the organization can contribute, own, and update their specific skills within the registry.
- Automated Synchronization: The system uses manifest-driven logic where consumer repositories are automatically updated whenever a skill owner commits a change to the registry.
- Breaking the Library Wall: This eliminates the friction of manual library maintenance, allowing organizations to maintain their local variants while remaining globally synchronized with the latest platform standards.