LOCKEDIN / FORWARD-DEPLOYED ENGINEERING 101
Understand it.
Then build it.
You can be an experienced engineer and still be new to agents. Start with what each thing is and how it helps a person do useful work. Explore 64 original introductions across software, agents, cloud, security and delivery. Start here without a model key; each topic connects to deeper instruction and practice.
A place to start.
Choose a short reading route, or look up a concept below.
- 101UnderstandPlain language, visual walkthrough, vocabulary
- 102Build with guidanceOriginal lessons, examples and knowledge checks
- 103Practise and explainRunnable labs, failure cases and handover
Coding agent · OpenAI
What is OpenAI Codex?
Codex is OpenAI’s coding agent: software that can help read a repository, make changes and run development tools. The CLI is its terminal interface. It is not a programming language, and the application it edits does not automatically become an AI application.
An FDE needs to turn an unfamiliar repository into a reviewed, reproducible change. The assistant helps with implementation; you still need a clear contract and evidence.
UNDERSTAND THE SYSTEM
A coding agent works inside an engineering loop
Codex is OpenAI's coding agent. Follow a request through repository context, permitted actions and evidence you can review. The application you build is a separate program.
Read the connections in plain English
- 01You define the change → Repository + instructions
Keep the request small enough to inspect.
- 02Repository + instructions → Codex
Give relevant files, the behavior contract and project instructions.
- 03Codex → Permitted local tools
A proposed command still runs under environment controls.
- 04Permitted local tools → Observed result
The test runner produces evidence about this copy of the code.
- 05Observed result → Codex
Use the observed failure to revise the implementation.
- 06Observed result → Review and hand over
Check behavior, scope and evidence before accepting it.
Check your understanding: Is adding an OpenAI API call necessary to learn Codex?
No. Codex assists software development. The project it edits can be a deterministic offline application with no model calls. Model access for the coding assistant and runtime model access for your application are separate.
CONNECT IT TO WHAT YOU KNOW
Start with something familiar.
Think of pairing with a colleague who can inspect files and run tools, while you define the behavior and review the result.
The coding assistant and the deployed application are separate. A passing test is an observation from the runner, not a claim in the chat.
Words worth knowing
- Coding agent
- A system that coordinates model proposals and development tools.
- CLI
- A command-line program used through a terminal.
- AGENTS.md
- Repository guidance; it does not grant technical permissions.
- Sandbox
- An enforced execution boundary, distinct from instructions and approvals.
Useful foundations
Open a definition and diagram here. Your lesson stays in place.
Interfaces to programsWhat are a terminal, a shell and a CLI?
QUICK REFERENCE / 101
What are a terminal, a shell and a CLI?
A terminal is the window where you type and see text. A shell interprets your commands. A command-line interface (CLI) is how a program accepts those commands. An AI coding agent can use a terminal, but the terminal itself does not reason or decide what to do.
You need to understand what will run, in which folder, and with whose authority before approving a coding agent's action.
Version control and reviewable changesWhat is Git?
QUICK REFERENCE / 101
What is Git?
Git is a distributed version-control tool. It records selected snapshots of files so you can compare changes, recover earlier versions and collaborate. It runs on your computer; a hosting service such as GitHub can hold another copy of the repository. Git is separate from your editor, coding agent and deployment system.
When you or a coding agent change an application, an FDE must explain what changed and which version was checked. Saving a file, selecting it for a commit and recording that commit are distinct actions. Understanding them keeps unrelated work out of a handoff.
Software behavior and evidenceWhat is an automated test?
QUICK REFERENCE / 101
What is an automated test?
An automated test is a program that exercises behavior and checks an expectation. You prepare an input, call the code and compare the actual result with the required result. A fixture is the prepared input or state. An assertion is the check written as code. A test runner selects and executes tests and reports what passed or failed.
An FDE reviewing a coding agent's change needs to understand what its test command actually established. A green report can describe one selected example, an incomplete check or the wrong expectation. Read the comparison and its scope before accepting the claim.
ONE EXAMPLE / FICTIONAL CASE
Follow the connections.
The fictional ticket project has a cross-tenant defect. Codex can help locate and repair it while the application itself remains offline and deterministic.
UNDERSTAND THE SYSTEM
A coding agent works inside an engineering loop
Codex is OpenAI's coding agent. Follow a request through repository context, permitted actions and evidence you can review. The application you build is a separate program.
See what moves and why.
Read the connections in plain English
- 01You define the change → Repository + instructions
Keep the request small enough to inspect.
- 02Repository + instructions → Codex
Give relevant files, the behavior contract and project instructions.
- 03Codex → Permitted local tools
A proposed command still runs under environment controls.
- 04Permitted local tools → Observed result
The test runner produces evidence about this copy of the code.
- 05Observed result → Codex
Use the observed failure to revise the implementation.
- 06Observed result → Review and hand over
Check behavior, scope and evidence before accepting it.
Check your understanding: Is adding an OpenAI API call necessary to learn Codex?
No. Codex assists software development. The project it edits can be a deterministic offline application with no model calls. Model access for the coding assistant and runtime model access for your application are separate.
Work through it, step by step.
Explain
Read the repository and identify the expected behavior.
What you have now: A baseline you can reproduce.
Bound
Provide relevant files, project instructions and a small task.
What you have now: A plan with an explicit scope.
Observe
Inspect the diff and run the actual tests.
What you have now: Expected versus actual results.
Transfer
Name the exact version and receiving action.
What you have now: A change another engineer can check.
FROM RECOGNITION TO REASONING
Explain the visual, predict a case and compare it with the linked workshop.
- 01
Name each component in plain English.
- 02
Predict the outcome before revealing the explanation.
- 03
Run the linked exercise and retain actual evidence with its limits.
A labeled explanation and a workshop observation you can reproduce.
Compare your reasoning when you are ready
The coding assistant and the deployed application are separate. A passing test is an observation from the runner, not a claim in the chat.
Keep notes in your own working copy. Moving through these cards does not record completion.
TURN THE IDEA INTO USEFUL WORK
You can now go deeper.
The coding assistant and the deployed application are separate. A passing test is an observation from the runner, not a claim in the chat.
Academy lessons use your existing training access. Free guides stay open.