IntegrAuth Academy
Identity, security and practical AI — free, from zero. Byte-sized lessons, real diagrams, hands-on labs, zero fluff. Two programs, each with its own exam and certificate.
English is hard? Your browser can translate this page into Tamil, Telugu, Kannada, Malayalam or Hindi. Tech words stay in English so you learn the real names.
How?
- Chrome on phone: ⋮ menu → Translate
- Chrome on computer: right-click → Translate to …
- Edge: the translate icon in the address bar
Coding assistants
Priya asks an AI coding agent to fix a failing build, the step that turns code into a working app. It says, "Done. All tests pass." It had deleted the failing test. Your goal: let the AI type, but check the work yourself.
🧪 Interactive lab — enable JavaScript to play with this one.
- A coding assistant writes code fast. You still own the result.
- "All tests pass" is a claim. The diff is the proof.
- Work in small steps. Commit after each good step.
A few words first
You met these in Tech basics. A repository, or "repo", is a project folder with its history. A diff shows removed (-) and added (+) lines. A commit is a saved checkpoint. Tests are small programs that check the code still works.
Three kinds of help
| Kind | What it does |
|---|---|
| Inline completion | Gray text appears as you type. Tab accepts it. |
| Chat in the editor | In the program where you write code, ask about the code. Get an answer or a suggested edit. |
| Coding agent | Give it a goal. It edits files, runs tests and tries again. |
The diff is the proof
Agents in a hurry have deleted, skipped or loosened failing tests. So passing tests prove something only if the tests did not change. Read test changes first. Changes outside the task need a hard look, too.
Small steps, then commit
A 40-line diff can be read. A 4,000-line diff gets skimmed. So ask for one small step at a time. Run the tests and read the diff. Good? Commit it. Broken? Roll back to the last commit.
Now commit at the right time, and roll back when a step breaks.
🧪 Interactive lab — enable JavaScript to play with this one.
Go deeper: the tools by name
Terminal tools (you type commands to use them) include Claude Code, OpenAI Codex CLI, Google's Antigravity CLI, Aider, OpenCode and oh-my-pi (command omp). Editor tools include Cursor, GitHub Copilot, Cline and Kilo Code.
As of October 2026, Aider, Cline, Kilo Code, Codex CLI, OpenCode and oh-my-pi are open source. Claude Code and Cursor are closed. GitHub Copilot is a closed service. Its VS Code chat extension is open source.
The license does not decide where your code goes. Every tool sends the files it reads to its model. So the model provider's data terms matter (Safe AI at work).
Go deeper: Priya's deleted test
Priya rejected the change that deleted the test. She asked the agent to find the real bug instead, in one small step. The new diff touched one function and no tests. The bug was simple: dates were compared in the wrong time zone. She ran the tests, read the diff and made a commit.
Go deeper: where to go next
A coding agent also reads files, runs commands and uses your logins. The Coding with AI track covers the rest:
- A short instruction file with house rules: Give it the right context.
- A plan before code: Plan before it codes.
- Tests and proof: Tests, checks and debugging.
- What the agent may do alone, and where secrets go: Permissions, sandboxes and secrets.
A coding agent runs commands as you. Before you let it work alone, read Permissions, sandboxes and secrets.
Pick a small project and a coding assistant with a free plan. Let it make one small change. Read the whole diff before you keep it.