Module 5 — The toolbox: coding harnesses, cloud and hosting
A coding harness is where you talk to the AI that writes your code. Beginners should start in a browser app builder, graduate to an AI code editor, and add a terminal coding agent once the project gets serious. For hosting, Vercel + GitHub covers almost every first app.
Lesson 5.1 — Four kinds of coding harness
| Kind | What it feels like | Tools | Best for |
|---|---|---|---|
| App builders (browser) | Chat on one side, live app on the other; no installs | Lovable, v0 by Vercel, Figma Make | First prototypes, landing pages, front-end designs |
| AI code editors (IDEs) | A full code editor with an AI agent built in | Cursor, Windsurf, Google Antigravity | Day-to-day building once you own the code |
| Coding agents (terminal / extension) | You give a task; the agent plans, edits many files, runs commands and tests | Claude Code, OpenAI Codex, Cline, OpenCode, AdaL | Big multi-file features, refactors, scaffolding a whole repo |
| Personal agent runtimes | An always-on agent that lives in your chats, with memory and skills | OpenClaw, Hermes Agent (Nous Research) | Automations and assistants around your business, not app scaffolding |
Notes on each:
- Lovable — builds full-stack apps from chat and connects natively to Supabase and Stripe; exports to GitHub.
- v0 — Vercel's prompt-to-UI/app builder; deploys straight to Vercel.
- Figma Make — turns prompts and Figma designs into working prototypes; great for the design phase.
- Cursor / Windsurf — AI-first editors based on VS Code, with agent modes that edit across files.
- Antigravity — Google's agent-first IDE, where you manage several agents working in parallel.
- Claude Code — Anthropic's agentic coding tool in the terminal, IDE, desktop and web; strong at long, multi-step builds.
- Codex — OpenAI's coding agent, available as a CLI, IDE extension and cloud agent.
- Cline / OpenCode / AdaL — open-source or bring-your-own-model agents; flexible and cheaper at scale.
- OpenClaw / Hermes — open-source (MIT) agent runtimes that connect to messaging apps and learn skills over time.
- HyperAgent — several unrelated products share this name; confirm which one before teaching it.
Recommended path: prototype the look in v0, Lovable or Figma Make → export to GitHub → open the repo in Cursor or Antigravity → use Claude Code or Codex for big features.
Lesson 5.2 — Cloud platforms
The cloud is renting someone else's computers by the minute. Three giants offer hundreds of services; simpler clouds trade options for ease.
| Platform | Character | When to use |
|---|---|---|
| AWS | Largest, most services; steepest learning curve | Enterprise customers, Bedrock models, heavy infrastructure |
| Azure | Microsoft's cloud; deep Office/Entra integration | Selling to Microsoft-centric companies, Azure AI Foundry |
| Google Cloud (GCP) | Strong in data, AI and Kubernetes | Gemini/Vertex AI, BigQuery analytics |
| DigitalOcean | Simple, predictable pricing | Your own small servers (VPS) for bots and background jobs |
| Heroku | The original "git push and it runs" platform | Simple back-end apps and APIs |
Key terms: a compute instance is one rented virtual computer (AWS EC2, Google Compute Engine). A virtual private server (VPS) is the simpler, flat-price version (DigitalOcean Droplets). Serverless means you upload code and the platform runs it only when needed — you never manage a server. Vercel runs your back end serverlessly.
Lesson 5.3 — Hosting and deployment
| Tool | Role |
|---|---|
| GitHub | Stores your code and its full history; triggers deployments; GitHub Actions automates tests |
| Vercel | Hosts Next.js apps; every push to GitHub deploys automatically, with a preview URL per change |
| Netlify | Similar push-to-deploy hosting, strong for static and marketing sites |
| AWS (Amplify) / Azure (Static Web Apps) | Push-to-deploy hosting inside the big clouds, for enterprise requirements |
The default beginner stack: GitHub (code) → Vercel (deploy + host) → Supabase (database, auth, storage) → Stripe (payments) → Vercel AI SDK + AI Gateway (agents).
Exercise: start your Technical Stack Key Sheet — one row per building block from Module 4: tool, account link, plan/price, who owns the login, where the API keys are stored.
Check yourself:
- When would you move from an app builder to an AI code editor?
- What's the difference between a VPS and serverless?
- What happens when you push code to GitHub with Vercel connected?