Coding in the AI era · CodeWiki

All beginner intermediate advanced

show All Unread Read

01 · Working with agents

3 topics

Agent context management Select and refresh repository context so an agent reasons from relevant code, constraints, and current evidence. beginner 17 min Agent instructions and MCP Write AGENTS.md and CLAUDE.md guidance and expose scoped tools through MCP so agents follow repository-specific rules. intermediate 19 min Agent task planning Turn a coding goal into bounded steps, checkpoints, and stopping conditions an agent can execute and you can inspect. intermediate 19 min

02 · Reviewing AI code

3 topics

AI-assisted debugging Use an AI pair to form hypotheses, design discriminating checks, and follow runtime evidence to a root cause. intermediate 19 min Generated change review Review generated diffs across files for correctness, edge cases, maintainability, and unintended behavior. intermediate 18 min Unfamiliar code reading Trace entry points, data flow, side effects, and invariants to build a fast, testable model of unfamiliar code. beginner 18 min

03 · Specs and tests

3 topics

Executable specifications Express behavior, boundaries, and acceptance criteria in examples and checks an agent can execute. beginner 18 min LLM application evals Build representative cases, scoring rules, and regression thresholds for nondeterministic LLM features. advanced 18 min Tests as agent guardrails Design focused tests that constrain implementation choices, expose regressions, and make agent work auditable. intermediate 17 min

04 · Prompting for code

3 topics

Code prompt vocabulary Name interfaces, invariants, failure modes, and code structures precisely enough to request the intended change. beginner 19 min Constraint-rich prompts State scope, non-goals, compatibility limits, and acceptance commands in prompts that bound generated changes. intermediate 20 min Examples and counterexamples Use representative inputs, desired outputs, and counterexamples to disambiguate a coding request. intermediate 17 min

05 · Tooling

3 topics

AI coding agents How coding agents read code, call tools, react to execution results, and stay bounded by permissions and verification evidence. intermediate 16 min AI coding CLIs Bound AI coding tasks by directory, context, and permission, then verify delivery with reproducible commands, diffs, and exit codes. intermediate 18 min AI IDEs How AI IDEs assemble editor context, model suggestions, and local tools, and how to bound and verify generated changes. intermediate 20 min

06 · Judgement

3 topics

Agent use boundaries Decide when uncertainty, sensitivity, cost, or weak verification makes direct human work or simpler automation safer. beginner 16 min Code provenance and verification Check generated code with independent evidence and trace copied patterns, dependencies, and licenses before shipping. intermediate 17 min Generated code security Threat-model generated changes and check trust boundaries, secrets, dependencies, input handling, and authorization. advanced 20 min