GitHub use cases

Prototyping & DesignCoding & Engineering

Use AI to close the gap between design and code by bringing production context into design work and turning design intent into more implementation-ready code changes.

Outcome: Shows how product teams can reduce handoff loss between designers and engineers. Instead of treating Figma and the codebase as separate worlds, the workflow helps design reflect what is actually in production and helps code changes preserve the intent of the design.

Coding & EngineeringKnowledge Management

Uses Mermaid diagrams and Claude Code automation to make codebases easier for AI to understand, review, and improve, turning architecture context and quality checks into reusable development artifacts.

Outcome: Shows how developers can make AI coding tools better by giving them visual maps of the system and automated guardrails. Instead of relying on long prompts, the workflow turns architecture, task flow, and code-quality checks into files and hooks the AI can reuse.

Research & AnalysisCoding & Engineering

Uses error analysis and targeted evals to improve AI products by inspecting failures, grouping them into patterns, and turning those patterns into tests that guide future product changes.

Outcome: Shows a practical way to improve AI quality without guessing. Instead of only tweaking prompts, teams can study real failures, classify them, build focused evals, and use those evals to decide whether the product is actually getting better.

Coding & Engineering

Uses a model-vs-model development workflow where one AI agent builds the first version and another reviews the output for bugs, edge cases, production readiness, and missed requirements.

Outcome: Shows how AI-assisted coding can become more reliable when creation and critique are separated. One model produces the work, another challenges it, and the human decides which feedback is worth acting on.

Coding & EngineeringBusiness Operations+1

Uses AI agents to make enterprise knowledge and engineering operations more actionable, including centralizing internal docs for agents, identifying tech-debt opportunities, and improving hiring or process workflows.

Outcome: Shows how enterprise AI adoption often starts by making the company legible to agents. When docs, decisions, code context, and process data are easier for AI to understand, agents can help with engineering cleanup, operational analysis, and better handoffs across teams.

Coding & Engineering

Uses advanced Codex workflows to run multiple engineering tasks in parallel, keep work organized with Plans.md, and route generated changes through GitHub review before merging.

Outcome: Shows how AI coding tools become more useful when paired with planning files, isolated worktrees, and normal code-review habits. Instead of one linear chat, the workflow turns Codex into a structured collaborator that can explore several tasks without tangling the codebase.

Coding & EngineeringPrototyping & Design

Uses a dual-agent app-building workflow where one AI agent acts as the builder and another acts as the reviewer, helping a non-traditional developer move faster while keeping a second layer of critique on the work.

Outcome: Shows how AI-assisted building can become more reliable when the work is split into roles: one agent creates the first version, another inspects it, and the human decides what to accept.

Customer Support

Uses Claude Code across Galileo’s 15 code repositories to answer complex customer deployment questions, then turns valuable Slack support conversations into public knowledge-base articles with Pylon.

Outcome: Gives customer-facing teams a way to answer deep technical questions without constantly interrupting engineers, while turning one-off support conversations into reusable public documentation.

Coding & Engineering

Claude Code’s founding engineers explain how Anthropic uses it internally: planning before coding, standardizing team settings, using stop hooks to finish tests, deploying subagents for review, and splitting large migrations into parallel work.

Outcome: Shows how AI coding tools become more reliable when teams add structure around them: plans before edits, shared permissions, automated checks, specialized review agents, and clear task boundaries.

Knowledge ManagementPersonal Productivity

Uses Claude Code inside Obsidian as a second brain, giving AI access to roughly 1,500 notes so it can retrieve old research, organize ideas, support deep work, and help him resume thinking after interruptions.

Outcome: Turns a personal note archive into an active research partner. Instead of manually searching old notes or rebuilding context after a break, Noah can ask Claude Code to find relevant material, organize it into project folders, ask clarifying questions, and help him pick up where he left off.

Coding & EngineeringPrototyping & Design

Uses HTML as a richer collaboration layer with Claude Code, turning AI-generated plans, brainstorms, editing interfaces, and design systems into visual artifacts that are easier for humans to read, adjust, and share.

Outcome: Makes AI-assisted development more visible and easier to steer. Instead of reviewing giant Markdown plans, humans can inspect interactive webpages, edit disposable micro-apps, and reuse living design-system files as context for future work.

Company AI AdoptionBusiness Operations

Built a company-wide AI adoption system at Delight.ai with an internal automation marketplace, gamified quests, AI usage tiers, secure app templates, and personal knowledge workflows like The Gardener.

Outcome: Turns AI adoption from a top-down mandate into an internal product. Employees can request automations, team up to build them, earn rewards, ship on secure templates, and see a clear path from beginner to advanced AI use.

Coding & EngineeringBusiness Operations

Uses Notion AI, Codex, GitHub, Slack, Honeycomb, and Whisper to make engineering work more structured: preparing standups, turning specs/comments into PRs, and giving engineers better context before they start building.

Outcome: Shows how AI can improve engineering velocity without skipping the parts that make software teams work well: shared context, written specs, reviewable changes, observability, and human judgment.

Coding & EngineeringBusiness Operations

Built internal Claude Code workflows that improved Intercom’s engineering process: PR-description quality hooks, AI telemetry dashboards, self-improving flaky-test agents, and agent-friendly installation paths for product integration.

Outcome: Shows how AI coding becomes more valuable when a team builds supporting systems around it: better PR metadata, better visibility into agent behavior, automated follow-up on flaky tests, and product setup flows designed for coding agents as well as humans.