Vibe coding for POCs, viable coding for production

Why code integrity merges review, security, and quality.

Vibe coding puts developer flow first, but standardizing quality requires verifying code integrity whether it’s generated by humans or AI. In this episode, Itamar Friedman, CEO and Co-founder of Qodo AI, joins the We Love Open Source podcast to share why code review needs AI to handle thousands of lines across piled-up PRs, how blue teams generate code while red teams verify integrity, and why junior developers must use AI to learn, not just to fix feedback and push it back through prompts.

Itamar got interested in code quality 20 years ago using machine learning for hardware verification at a company sold to NVIDIA. Hardware specs are bytes in, bytes out. Software specs differ: User story in, user interface out. That’s where Large Language Models (LLMs) needed to catch up. His 2021-2022 realization that AI would generate most content, especially code, scared him. How do we avoid catastrophic moments where software stops working for fundamental services? Qodo launched three months before ChatGPT to address this.

Code integrity merges code review, security, and quality into one holistic solution. Red teams verify what gets built matches intent with high integrity. Blue teams handle code generation. The exciting future is shift left and shift right, pulling standards from PR discussions, Slack conversations, architecture documents to provide suggestions as you code. Mixing blue and red creates purple, hence Qodo’s purple branding.

Read more: Stop guessing, start measuring developer engagement

Junior developers face a critical choice in how they use AI. In one scenario, a senior gives feedback and the junior simply pushes it to AI to fix the code. This raises the question: Why do we need that junior at all? In the second scenario, the junior takes that same feedback but uses it as a learning opportunity. They ask questions of people and AI, then come back with understanding that goes beyond just prompting for fixes. The key is using AI to learn, not just to vibe code. That’s how you grow.

Vibe coding prioritizes flow. Agentic coding continues that flowish feeling. But Qodo advocates viable coding, keeping flows while making code reliable. Vibe coding works for Proof of concepts (POCs) and Minimum Viable Products (MVPs), alpha software can be done 2x to 50x faster. But beta and GA, especially heavy-duty software with personal information and transactions, requires viable coding.

Key takeaways

• Code integrity merges review, security, and quality: Red teams verify code meets intent while blue teams generate. Future is pulling standards from PRs and Slack to provide suggestions as you code.
• Junior developers must use AI to learn, not just fix feedback: Use AI to understand why code works by fetching similar open source projects and asking AI to explain differences, not just pushing senior feedback through prompts.
• Vibe coding for POCs, viable coding for production: Flow-focused vibe coding works for alpha MVPs built 2x-50x faster. Reliable viable coding is required for beta and GA software.

Itamar built Qodo before ChatGPT launched to address AI’s impact on code quality, proving that standardizing integrity matters as much as generating code.

More from We Love Open Source

• 15 open source backup solutions to protect your data
• 10 common Linux issues and the fast fixes you need
• Stop guessing, start measuring developer engagement
• How to engage with policy makers when you’re a developer
• 5 forces driving DevOps and AI in 2026