Summary of "OR: Сколько кода сегодня пишет AI - выпуск 24 #RubyRussia 2025"

Summary of the Video “OR: Сколько кода сегодня пишет AI - выпуск 24 #RubyRussia 2025”

The podcast discusses the current and future role of artificial intelligence (AI) in software development, focusing on how AI tools are transforming coding practices, QA processes, product integration, and business workflows. The conversation critically examines the hype versus reality of AI as a “silver bullet” in development.

Key Technological Concepts & Analysis

AI as a Development Tool

AI is often hyped as a 10x productivity booster, but in reality, it is a complex tool that requires skilled users.
AI cannot replace human subjectivity; it remains a tool that demands a higher level of expertise to use effectively.
Modern AI coding assistants (e.g., GitHub Copilot, ChatGPT, Claude) are sophisticated “smart autocompletes” or multi-agent systems that decompose tasks but still produce inconsistent outputs.
AI models have limitations such as hallucinations, inability to perform accurate calculations, and contextual shifts that require continuous quality control and testing.
Example: Repeated prompts to generate a Bash script resulted in multiple different valid solutions, illustrating variability and unpredictability.

Quality Control & Review

AI-generated code increases the volume of pull requests, raising the mental burden on developers to review and ensure quality.
Human oversight remains critical; automated QA scripts generated by AI still require manual review and testing.
The distinction between “wipe coding” (blindly accepting AI-generated code) and “engineering-assisted coding” (using AI with professional discipline) is emphasized.
Statistics suggest about 30% of Python code on GitHub is AI-generated, but quality depends on human review and modification.

AI in QA Automation

AI tools are used to write test scripts, aiming to reduce QA bottlenecks without eliminating QA roles entirely.
Automated testing reduces risk since errors in test scripts are less costly than production bugs.
Human reviewers are necessary to ensure generated tests are correct and meaningful.

AI Integration in Products and Business Processes

AI is increasingly embedded in products, often for knowledge aggregation, customer support, and internal communication (e.g., internal Slack bots that answer questions using company knowledge bases).
AI can simplify onboarding and automate routine tasks, but raises privacy and compliance concerns, especially when handling sensitive data (e.g., email access for identity management).
AI-driven process automation can radically change workflows, such as accelerating microelectronics production troubleshooting by quickly matching cases from historical data.
Sales and customer interaction are evolving with AI-driven chatbots and lead qualification, reducing costs and increasing responsiveness outside normal working hours.
The rise of AI changes user expectations for speed and quality of interaction.

Future Outlook on Programming and AI

The development industry will experience waves of hype and adjustment (“sine wave” or hype cycle).
Programming may shift focus from writing code to architectural design and orchestration, as AI handles more of the coding.
Platforms like Lavable and Supabase offer pre-architected solutions that leverage AI, potentially popularizing standardized architectures.
Human roles may evolve towards interaction design and emotional intelligence in AI-human communication.
Legacy systems and languages (e.g., COBOL) will persist due to inertia and business needs.
Standardization and reduction in the diversity of languages and tools are expected as AI-driven development grows.
AI will not fully replace developers but transform their roles; “AI operators” or specialists who guide AI tools may emerge.

Ethical and Practical Considerations

Privacy, data security, and compliance are major challenges in integrating AI into business processes.
The need for human control and responsibility remains paramount.
Blind reliance on AI without understanding or reviewing output is risky.
Developers must maintain engineering discipline and adapt to new workflows involving AI.

Product Features & Use Cases Discussed

GitHub Copilot and AI-assisted coding tools
- Widely used by developers.
- Support multi-shot prompting and context-aware code generation.
- Require human review and integration into industrial architectures.
Internal AI Bots (e.g., Slugbot at Get)
- Automate answering questions in company Slack channels.
- Aggregate internal knowledge and documentation.
- Improve onboarding and knowledge retention.
AI in QA Automation
- Writing test scripts.
- Reducing QA team size or workload.
- Maintaining quality through human reviews.
AI in Business Processes
- AI-driven onboarding and identity verification.
- AI-assisted sales lead qualification and customer interaction.
- Process acceleration in manufacturing through AI case matching.

Guides, Tutorials, or Recommendations

Use AI tools as assistants, not replacements; maintain engineering rigor.
Implement continuous quality control and testing for AI-generated code.
Employ human reviewers for all AI outputs, especially in QA and production code.
Understand AI model limitations (hallucinations, variability).
Digitize and analyze business processes before automating with AI.
Prepare for evolving roles in development, focusing on architecture and human interaction design.
Stay updated on AI trends and integrate them thoughtfully to optimize workflows.

Main Speakers / Sources

Alexander (Host) – Moderator of the podcast.
Artem Kuznetsov – Speaker on AI’s impact on development, critical of hype, emphasizes engineering challenges.
Stas German – Expert in complex architectures and AI integration, shares practical experience with AI in QA automation and internal tools at Get.

Overall, the podcast provides a nuanced, experience-based analysis of AI in software development, balancing optimism with caution, and emphasizing the ongoing need for skilled human involvement.