Don't babysit your AI Agents to keep them on track with Lukas Holzer

About this episode

Andi Grabner discusses the challenges and risks associated with AI coding agents with Lukas Holzer, focusing on quality, context, and evolving engineering roles.

AI coding agents are fast—but speed alone doesn’t guarantee quality. In this episode, Andi Grabner talks with Lukas Holzer (Straion) about why large context files and “almost right” AI code create new risks for engineering teams. You will learn about the "Lost in the Middle Syndrom" and why many organizations are not getting the promised 10x engineering boost right now! Andi and Lukas also explore rule adherence, dynamic context generation, enterprise readiness for AI-first development, and how software engineering roles are evolving in the age of AI. Tune in to learn more ... Links we discussed LinkedIn Profile: https://www.linkedin.com/in/lukas-holzer/ Straion Website: https://straion.com/ 90 Percent Rule Blog: https://straion.com/blog/90-percent-rule-adherence-straion-coding-agents/ 1million tokens Blog: https://straion.com/blog/1m-tokens-wont-save-your-engineering-standards/

People in this episode

Host: Andi Grabner

Guest: Lukas Holzer

Topics covered

• AI coding agents
• engineering risks
• dynamic context generation
• enterprise readiness
• software engineering roles
• Lost in the Middle Syndrome

Keywords

• AI agents
• engineering boost
• context files
• rule adherence
• software development
• coding standards

Mentioned in this episode

Organizations: Straion, LinkedIn

More episodes of PurePerformance

• Beyond the Hype: Open Source, Observability, and Finding Your AI Breakthrough · June 8, 2026 · 34 min
• 8 Factor Producers to Scale Platform Engineering in an AI-First world with Abby Bangser · May 25, 2026 · 35 min
• Observability in the AI‑Native Era with Hilliary Lipsig and Rob Rati · May 11, 2026 · 52 min
• From Bowling Lanes to AI Lanes: Chris LaBrado on MDCD and the AI Interface Era · April 13, 2026 · 54 min
• AI-Ready Codebases: Engineering Discipline for Agentic AI with Adam Tornhill · March 30, 2026 · 53 min
• AI‑Native: Building Faster Than We Can Spec with Wolfgang Heider & Benedict Evert · March 16, 2026 · 44 min