420: AI for the Code-Writing Purist: How to Use AI Without Surrendering Your Keyboard

What It Covers

Arvid Kahl demonstrates how developers can leverage AI tools for code investigation, security review, optimization, and documentation without letting AI write production code, maintaining control while augmenting capabilities through strategic prompting and system configuration.

Key Questions Answered

• •Bug Investigation Partner: Configure AI with system prompts to investigate bugs by providing error messages, suspected modules, input/output data, and related code. AI builds internal theory of programs to trace issues without executing code, offering fresh perspective beyond static analysis tools.
• •SQL Query Optimization: Run queries with EXPLAIN ANALYZE output, provide database schema and existing indices to AI for optimization suggestions. This approach yields immediate speed improvements in 99% of cases by identifying wrong index ordering or missing indices without AI writing code directly.
• •Security Review Automation: Implement pre-commit hooks with security-focused system prompts expressing preferences and best practices. AI checks recent commits for vulnerabilities, semantic correctness, and adherence to standards, catching issues like misconfigured cache settings that could enable system abuse before deployment.
• •Documentation-Driven AI Reliability: Create comprehensive documentation blocks in each module describing purpose, data flow, performance requirements, and critical functions. After documenting 10 files manually, AI can replicate pattern across codebase with 95% accuracy, dramatically improving subsequent AI tool reliability and acquisition readiness.