Beginner AI Coding Training

This is a completely free, public, self-paced learning path. There is no login requirement, paywall, formal exam, certificate, or capstone. Complete eight modules in order and use the ten practical exercises to prove that you can work safely, not merely recognize terminology. Estimated time is 12 to 18 hours.

1. How to Use the Training

Prepare a computer, browser, one IDE, Git, the project's runtime, and one coding agent. Use a private practice repository with synthetic data. Never use production secrets or real customer records.

For each module, learn the concept, try a small example, complete the practice, verify the result, record open questions, and continue. Personal progress never controls access to AI Blueprint features.

2. Module 1: Understand AI Coding

Outcome: distinguish a model, provider, coding agent, IDE, context, and token, then choose a suitable starting path.

Read Start AI Coding from Zero and Models and Limits. Remember that an agent sees only the context supplied to it and that every result needs verification.

P01: Choose Your First Tool and Model

Write down your local or browser path, IDE, primary coding agent, default model or mode, budget boundary, reason, and where usage is displayed. You pass when you can explain why an IDE, agent, and model are different layers.

3. Module 2: Tools and Workspace

Outcome: install an IDE, open the correct folder, run a project, read basic diagnostics, and stop a development server.

Follow Tools and IDE Setup. Learn Explorer, Search, Terminal, Problems, Source Control, and the diff viewer.

P02: First Local Run

Open a practice repository, read its README and package metadata, install dependencies with the correct package manager, run the development server, open the app and browser console, then stop it safely. Repeat the run from a fresh terminal using only your notes.

4. Module 3: Git, GitHub, and Recovery

Outcome: create checkpoints, read diffs, use a branch, and recover a change without discarding unrelated work.

P03: Diff and Restore

Commit a baseline, edit one file, inspect git diff, restore only that file, repeat the correct change, review, and commit.

P04: Branch and Merge

Create a feature branch, make and commit one change, return to main, merge, and inspect the graph.

P05: Safe Merge Conflict

In a synthetic repository, change the same line on two branches. Merge, read the conflict markers, resolve and test. Repeat once and use git merge --abort so you understand both completion and recovery.

5. Module 4: Prompting, Context, and AI Skills

Outcome: write an executable prompt, control context, and turn a repeated workflow into a small reviewable skill.

A strong prompt states the objective, context, scope, exclusions, acceptance criteria, checks, and stop condition. Use repository instructions for stable rules and the task prompt or plan for temporary details.

P06: Rewrite a Prompt and Build a Skill

Rewrite "make a good dashboard" into one scoped responsive screen with named states and checks. Then create review-current-diff/SKILL.md that audits a diff only when explicitly invoked. Test one request that should use the skill and one that should not. Review it for destructive commands, network calls, secret access, and hidden instructions before installation.

6. Module 5: Choose a Technology Stack

Outcome: select the smallest operable stack from product requirements instead of trends.

P07: Stack Decision

Compare a static landing page, a multi-user CRUD SaaS app, and an internal data tool. For each, choose frontend, backend, database, authentication, hosting, and testing. Explain what you intentionally excluded, such as microservices, Docker, or realtime features.

Preserve the Project ZIP's approved technical foundation unless a documented reason and human approval justify a change.

7. Module 6: Professional Frontend UI

Outcome: translate visual direction into a design contract and one reviewable screen.

Read Designing Better Frontends with AI. Collect two or three legal references, write DESIGN.md and ASSETS.md, and implement one vertical slice.

P08: DESIGN.md and One Screen

Define typography, color roles, spacing, grid, component rules, states, responsiveness, accessibility, motion, asset ownership, and do/don't rules. Implement loading, empty, error, long-content, focus, and mobile states. Review screenshots against the contract.

8. Module 7: Database, Auth, and API

Outcome: understand that authentication is not authorization and prove data isolation between users.

P09: Protected Data

Create a synthetic tasks table with user_id. Test as User A and User B. Prove each user can read and mutate only their own records through direct API requests as well as the UI. Keep secrets server-only, validate input at boundaries, and version every migration.

9. Module 8: Testing, Debugging, and Deployment

Outcome: diagnose from evidence, run appropriate quality gates, and deliver a safe preview.

P10: Seeded Bug and Preview

Reproduce a bug, record expected and actual behavior, add a failing test, implement the smallest fix, run lint, typecheck, tests, and build, review the diff, then deploy the branch to preview. Test desktop, mobile, loading, error, and unauthorized states and record a rollback plan.

10. Personal Completion Checklist

• I can distinguish model, provider, agent, and IDE.
• I can run the project from the correct folder.
• I can review a diff and recover one file.
• I have resolved and aborted a merge conflict.
• My prompts define scope, acceptance, checks, and a stop condition.
• I can audit a skill before installing it.
• I choose a stack from requirements.
• I write a design contract before scaling UI.
• I test authorization with two users.
• I debug from evidence and validate a preview.

Continue with Blueprint Project ZIP to Code and apply the workflow to a real project using non-production credentials first.