SDLC & AI tools
How AI assistants and agents help write code and build systems across the software lifecycle — and the controls that keep the output trustworthy.
Questions on using AI to write code and build systems, across the whole software development lifecycle rather than just autocomplete: turning vague requirements into a spec, prompting and steering an agent, implementing, testing and reviewing generated code, keeping secrets and proprietary code safe, and delivering and operating what ships. The emphasis is on what each phase gains, what can go wrong, and which controls (tests, review, permissions) make the result trustworthy. Tool-agnostic; examples lean on agentic CLI tools such as Claude Code and Codex.
Subtopics
- AI across the SDLC — overviewMid / SeniorA phase-by-phase map of where AI helps, the levels of autonomy, and the shared failure modes.6 concepts
- Requirements and design with AIMid / SeniorUsing LLMs to refine requirements, explore design options and record decisions without outsourcing the judgment.7 concepts
- Prompting and workflows for developersMid / SeniorHow to brief a coding agent: context, plan-then-implement loops, verification criteria, session hygiene and parallel work.9 concepts
- AI-assisted implementationMid / SeniorChoosing between completion, chat and agents, working in small verified steps, and handling hallucinations and legacy code.10 concepts
- Testing and code review with AIMid / SeniorGenerating meaningful tests, using AI as a first-pass reviewer, and avoiding tests that only restate the code.9 concepts
- Security, secrets and data leakageMid / SeniorKeeping secrets and proprietary code out of AI tools, constraining agent permissions, and securing generated code.9 concepts
- Delivery, operations and governanceMid / SeniorAI in CI/CD, infrastructure as code, incident response and documentation, plus the policies for adopting it.9 concepts