Lesson Content

Problem definition conversations.

The goal of AI-assisted problem scoping is not to let AI define the problem – it is to use AI as a devil's advocate and assumption-surfacing partner. Before starting to code:

1. Describe the problem as you currently understand it to Claude
2. Ask: "What assumptions am I making that could be wrong?"
3. Ask: "What alternative ways could this problem be defined that I'm not considering?"
4. Ask: "What are the most common failure modes for this type of solution?"
5. Use Claude's responses to stress-test your own understanding – not to accept its framing wholesale

This conversation does not take the place of talking to users or stakeholders. It supplements it by surfacing blind spots before you invest in building.

Research synthesis for technical decisions.

For any significant technical decision – choosing a library, evaluating a database, deciding on an architecture pattern – AI can compress the research phase significantly:

1. Describe the technical problem and your constraints (scale, team expertise, operational overhead tolerance, budget)
2. Ask Claude to summarize the main options, the trade-offs for your constraint set, and what the current practitioner consensus is
3. Ask follow-up questions about the options that best match your constraints
4. Verify key claims against current documentation or recent community sources before deciding

The AI provides a research synthesis map; you verify the most important claims and apply your judgment to the trade-offs.

Architecture option evaluation.

For architecture decisions, a structured AI conversation:

1. Describe your requirements, constraints, and scale targets
2. Ask Claude to propose three to five architecture patterns that fit your constraints
3. For each pattern, ask: "What are the failure modes at scale? What are the main operational challenges? What does this pattern require the team to be good at?"
4. Ask Claude to argue against its own recommendation – what would a critic of each approach say?
5. Make your architecture decision based on your own judgment, informed by the conversation

The "argue against your own recommendation" step is particularly useful – it surfaces risks the AI's initial framing may have buried.

The "AI first, code second" habit.

For any new feature, build task, or significant change: before opening the editor, open a Claude conversation and spend ten to twenty minutes on problem scoping. This habit has two effects: you catch bad scoping early (when it is cheap) and you arrive at the code with a clearer understanding of what you are actually building.