Part 3 of 7 · how to code with AI
Prompt Patterns
The reusable structures of effective prompts.
7 lessons·~90 seconds each
Constraints, framing, multi-turn refinement, the critique loops, the three-implementations technique. The actual craft of vibe coding.
3.1
Constraints first. Instructions second. Always.
An LLM follows constraints far more reliably than it follows instructions. Lead with what it must NOT do.
Read
3.2
Negative prompting.
Draft pendingTell it what NOT to do. The defaults are the trap.
Read
3.3
The hostile reviewer critique loop.
Draft pendingMake the engine roast its own output before you accept it.
Read
3.4
Ask the engine to break it before it builds it.
Get the LLM to enumerate failure modes BEFORE it writes code. Half the bugs surface in the analysis, not the implementation.
Read
3.5
The three-implementations technique.
Draft pendingAsk for three. Pick one. The third is usually the one.
Read
3.6
Paste structure, not just intent.
Draft pendingShow the engine the shape it should produce.
Read
3.7
The 'show me the diff' reframe.
Draft pendingDiffs force precision. Prose hides drift.
Read