Published 13 Mar 2026
Duration: 01:01:57
Large Language Models (LLMs) have the potential to revolutionize software development, but their integration also raises challenges and requires a balance between AI capabilities and human oversight.
The episode delves into the process of agentic engineering using Claude code to develop a group management app. It explores the challenges, successes,...
The podcast discusses a project exploring agentic engineering through the development of a cross-platform (web, iOS, Android) team management app using LLMs like Claude. The goal was to test new technologies with minimal manual code review, prioritizing prompting strategies to build a minimum viable product (MVP) quickly. The team shifted from solo development to a two-person collaborative workflow, using Expo as the framework for cross-platform development. Expo simplified the process by abstracting React Native complexities and offering built-in features like serverless deployment and cloud-based builds. The project involved integrating the LLM (Claude) to guide architecture, using tools like Terraform for infrastructure and Scaleway for serverless SQL databases. Challenges included UI/UX design issues, such as misaligned navigation elements and input validation, which required manual debugging and iterative prompting. The LLM excelled in backend tasks like API testing, security, and infrastructure setup but struggled with UI-level validation and layout adjustments, highlighting the need for automated tools like Playwright and specialized human expertise.
Key challenges revolved around balancing the LLMs strengths in rapid code generation with its limitations in testing, debugging, and UI refinement. The team relied on manual error correction, copying terminal logs into prompts to iteratively fix issues, while emphasizing the need for structured workflows and phased development to manage complexity. The MVP was functional but lacked polish, underscoring the importance of rigorous testing and user experience design. The project also highlighted the limitations of current LLMs in handling repetitive tasks and maintaining consistency across platforms, with suggestions for role-specific agent systems or improved self-audit capabilities to enhance reliability. Overall, the experiment demonstrated the potential of agentic engineering to streamline development but underscored the necessity of human oversight, clear specifications, and integration of automation tools to address gaps in AI-driven workflows.
27 Mar 2026 #17 Testing in the AI era
Recommended: In depth discussion of AI and Software Development and the future of Software Testing.
Testers must adapt to AI's role in code generation by rigorously validating AI-produced code against structured requirements, balancing human oversight with AI tools to address quality, ownership, and architectural testability challenges while redefining testing as a collaborative, discovery-driven process.
20 Mar 2026 #16 Bunny CDN
Bunny CDN is spotlighted as an EU-based, cost-effective alternative for hosting static websites and development projects, offering user-friendly setup, Storage Zones, Terraform/GitHub Actions automation, edge computing, and competitive pricing, though it lacks enterprise-scale infrastructure and global scalability for high-volume workloads.
27 Feb 2026 #14 EU Cloud Alternatives - Scaleway
This podcast discusses EU-based alternatives to mainstream infrastructure and software development tools to achieve digital sovereignty, examining open-source and self-hosted solutions for development and management.
20 Feb 2026 #13 OpenClaw
The podcast discusses OpenClaw, an open-source AI assistant that learns user routines and preferences, and offers features like task automation and scheduled check-ins.
13 Feb 2026 #12 DevOps and SecOps
This podcast explores the evolution of DevOps, highlighting its shift from siloed teams to a culture of shared responsibility, accountability, and continuous improvement, resulting in faster problem resolution and more reliable systems.