Why my AI agents needed a rivalry

Description and use

Using two distinct AI models—Gemini as engineer and Claude as architect—eliminates hardcoded code issues, producing a robust application.

Problem

A single Gemini agent hardcoded results, leading to fragile, non-dynamic code.

Outcome

Multi-model review cycle eliminated hardcoding, producing a robust, dynamic application.

Workflow

• 1. Spin up single Gemini agent to generate initial code.
• 2. Identify hardcoding issue by testing with new data.
• 3. Create Bridge Deck local app for collaboration.
• 4. Deploy two agents: Kite (Gemini) as engineer, Slate (Claude) as architect.
• 5. Kite writes code, Slate reviews and raises issues.
• 6. Slate keeps issues open until fixes are verified.
• 7. Iterate until code is dynamic and reliable.

Evidence

The code reviews suddenly grew teeth.

In just a few days, PhrasePulse evolved from a fragile, hardcoded prototype into a bulletproof, well-functioning app.