AI for Technical & Development Professionals

Generation, explanation, tests, docs, review support — and what stays with the engineer: architecture, security, accountability.

In this lesson:

• Name the five places AI works in an engineering workflow
• Identify the three things that stay with the engineer
• Adopt the stance for the rest of the course

The review standard, read every line, and the tells of plausible-but-wrong code.

In this lesson:

• Apply the same review standard to generated code as to a stranger's PR
• Recognize the common patterns of plausible-but-wrong generated code
• Refuse to merge code you can't explain

The brief for a coding task — purpose, constraints, conventions, examples — and the repository context the tool needs.

In this lesson:

• Write a coding brief that produces code in your conventions on the first try
• Give the assistant the right repository context without over-sharing
• Maintain a project-level context file the whole team uses

What never goes in a prompt, license and provenance questions about generated code, and approved tools.

In this lesson:

• Keep secrets, customer data, and restricted code out of AI tools
• Handle license and provenance questions about generated code
• Apply the approved-tools policy for coding assistants

Date-stamped research on productivity, task fit, and the over-reliance trap for skills.

In this lesson:

• Summarize what controlled studies have found about AI and developer productivity
• Match AI use to the tasks where it measurably helps
• Guard against the perception gap and skill erosion

Small, specified tasks; interfaces first; the compile-test-review loop.

In this lesson:

• Break work into small, specified generation tasks
• Design the interface before generating the implementation
• Run the compile-test-review loop on every generated change

Coverage of edges, not just the happy path; tests that would fail; reviewing generated tests as carefully as code.

In this lesson:

• Brief the assistant for tests that cover edges and failure modes
• Verify that generated tests can actually fail
• Review generated tests with the same rigor as generated code

The debugging brief — code, error, expected, actual, minimal reproduction — and hypotheses, not conclusions.

In this lesson:

• Write a debugging brief that gets useful hypotheses
• Treat the assistant's diagnosis as a hypothesis to test, never a conclusion
• Recognize when the assistant is pattern-matching instead of reasoning

READMEs, API docs, and runbooks from the code — kept in sync, with the accuracy pass.

In this lesson:

• Generate documentation from code that reflects what the code actually does
• Keep docs in sync with code through the review process
• Run the accuracy pass on every generated doc before it's published

The assistant as a first-pass reviewer — what it catches, what it misses — and the human review that is not optional.

In this lesson:

• Use the assistant as a first-pass reviewer on every PR
• Know which classes of problems AI review catches and which it misses
• Keep the human review as the decision on every merge

Least privilege, sandboxes, and human approval before merge, deploy, or delete.

In this lesson:

• Scope a coding agent's task and permissions before it starts
• Run agents in sandboxes with least privilege
• Place human checkpoints before every irreversible action

Injection, secrets, dependencies, unsafe defaults — the security checklist and scanning tools.

In this lesson:

• Recognize the security weaknesses generated code commonly introduces
• Run the security checklist on every generated change
• Use scanning tools as a floor, not a ceiling

Comments, issues, and inputs that carry instructions — and defensive patterns for agent workflows.

In this lesson:

• Explain how prompt injection reaches coding assistants and agents
• Identify the untrusted content sources in an engineering workflow
• Apply defensive patterns that limit what injected instructions can do

Conventions, disclosure in PRs, shared prompts, the test set for prompts, and the review policy.

In this lesson:

• Establish team conventions for how AI tools are used in the codebase
• Disclose AI assistance in PRs in a useful, non-performative way
• Maintain shared prompts with a test set and a review policy

Brief, generate, review, test, protect — the checklist and your next feature; the finale.

In this lesson:

• Recall the five-part practice of the AI-enabled engineer
• Apply the course checklist to a real feature
• Complete AI for Technical & Development Teams