Build and present a small end-to-end app in groups of 2-3 using AI coding tools. Keep the scope tight, ship something real, and focus your presentation on how you used the tools well.
Build a simple app with 1-2 clear user journeys.
Good examples:
- habit tracker
- mini task board
- event RSVP app
- micro SaaS landing page with signup
- simple capstone prototype
- Choose something small enough to finish.
- Prioritize one working flow over many half-built features.
- Use fake data or seeded data if that helps you move faster.
- Keep manual coding light, but do review what the tools generate.
You only need two tools:
- one primary coding agent
- one optional UI/app builder
Recommended order for this course:
| Rank | Tool | Best use | Access |
|---|---|---|---|
| 1 | Claude Code | Best default choice for repo-level coding, refactors, debugging, and task execution from the terminal. | Paid only. Included with Claude Pro/Max and Team/Enterprise seats, or available through Claude Console API usage. No official free tier. |
| 2 | Codex app / CLI / IDE extension | Strong for delegated coding tasks, repo understanding, and structured multi-step work. | Included with ChatGPT Plus, Pro, Business, Enterprise, and Edu. Codex is temporarily available to ChatGPT Free and Go users. |
| 3 | GitHub Copilot | Easiest option if your team wants to stay inside VS Code, JetBrains, or GitHub. | Free plan available with limited requests. Paid plans increase limits. Some students, teachers, and OSS maintainers can also get premium access for free. |
| 4 | Google Antigravity | Agent-first IDE from Google with editor, terminal, and browser control for end-to-end task execution. | Free plan available with weekly rate limits. Google AI Pro and Ultra subscribers get higher limits. |
| 5 | Cursor | Good all-in-one AI editor for fast multi-file product work. | Free Hobby plan available with limited requests. Paid plans expand usage. |
| 6 | Windsurf | Good alternative AI IDE if you want strong agent-style editing in an editor workflow. | Free plan available. Paid plans add more usage and team features. |
| 7 | v0 | Best UI-first builder for React/Next.js style prototypes. | Free plan available. Paid plans add more credits, collaboration, and privacy controls. |
| 8 | Lovable | Strong full-stack app builder for fast MVPs and demos. | Free plan available with daily and monthly credit limits. Paid plans add more credits and workspace controls. |
| 9 | Bolt.new | Useful for quick browser-based prototyping and fast iteration. | Free plan available with daily and monthly token limits. Paid plans raise limits and remove some restrictions. |
| 10 | Gemini CLI | Solid backup terminal agent, especially if your team already uses Google tools. | Free tier available with a personal Google account. Google AI Pro and Ultra subscribers get higher limits, and paid API usage is also available. |
Tools removed from the older version:
- long-tail IDE assistants that are not central to this project brief
- pricing-heavy comparisons that go stale quickly
- older/noisier tool lists that made choice harder instead of easier
- Form a team of 2-3 and assign roles: Driver, Prompt Lead, QA/Tester, Presenter.
- Pick your app idea and cut it down to 1-2 user journeys.
- Choose at least two tools and write down why they fit your workflow.
- Add project-wide instructions to
project/AGENTS.md. - Keep reusable prompt patterns and review checklists in
project/SKILLS.md. - Build in short loops: plan, prompt, review, test, adjust.
- Prepare a 5-minute demo with one clean story.
README.md is the human-facing entry point for a repository. It explains what the project is, why it exists, how to run it, and what someone should look at first.
Use README.md for:
- project overview
- setup and run instructions
- deliverables
- demo context
- links to the important project files
How to create it:
- start with the project name and one-sentence summary
- add setup and run instructions
- list the deliverables or main features
- link to other important files such as
project/AGENTS.md - keep it short enough that a new teammate can scan it quickly
How to use it:
- read it first when you open a repo
- update it when setup steps or project goals change
- use it as the source of truth for humans, not for detailed agent rules
Helpful docs:
AGENTS.md is the agent-facing instruction file. Think of it as a README for coding agents: it gives AI tools the project context, build goals, constraints, preferred stack, and decision rules they should follow while generating or editing the project.
In this repo, the main agent instructions live in project/AGENTS.md because that folder is intended to become the generated app workspace.
The project/ folder is included as a concrete example of how AGENTS.md and SKILLS.md can be used together to define and generate a specific project. In this case, it describes a small chatbot interface project that an agent can build from those files.
Use AGENTS.md for:
- what the agent should build
- project scope and non-goals
- architecture preferences
- commands, quality bar, and review rules
- special constraints that should not clutter the human README
How to create it:
- describe the project the agent should build or maintain
- state the preferred stack and any constraints
- list required features, non-goals, and quality expectations
- add decision rules so the agent can keep moving without constant clarification
- keep instructions concrete and specific to the repo
How to use it:
- place it where the agent will work so it is in scope for that project
- start the agent in that folder or point the agent to the file explicitly
- update it when the project scope, architecture, or priorities change
- use it to reduce repeated prompting and keep agent output aligned
Helpful docs:
- A repo with clear setup/run instructions, or a hosted preview link
- A working app that demonstrates the core user journey
project/AGENTS.mdwith repo-specific AI instructionsproject/SKILLS.mdwith reusable prompts, checklists, or handoff patterns- A 5-minute live demo covering the problem, user journey, tools chosen, example prompts, what worked, what failed, and what you would do next
- Does the core journey work end to end?
- Did the team choose tools intentionally?
- Did the prompts improve over time?
- Did the team collaborate clearly?
- Did the presentation show honest reflection, not just a polished demo?
- Pricing is omitted on purpose because it changes too often to be useful in a course brief.
- Official access references: Claude Code, OpenAI Codex, GitHub Copilot, Google Antigravity limits, Firebase Studio pricing, Cursor pricing, Windsurf pricing, v0 pricing, Lovable plans, Bolt pricing, Gemini CLI, Gemini CLI higher limits.