QCon London AI Coding State of the Game: More Capable, More Expensive, More Dangerous Coding Agents

"Böckeler highlighted that context engineering is probably the most significant development of the year, referring to the curated information a model or agent reads to improve its results."

"Anthropic has since broken down this 'monolithic' file into smaller skills, resulting in a more granular approach to coding capabilities, enabling a pragmatic approach known as 'lazy loading.'"

"Böckeler emphasised that we are moving closer to 'hands-off' coding, as these coding agents can now run unsupervised for up to 20 minutes."

The AI coding space has evolved significantly, moving from vibe coding to the use of autonomous coding agents. Major concerns include worsening security and rising costs. Context engineering has emerged as a key development, allowing for a more granular approach to coding capabilities through lazy loading. This method improves organization and efficiency. Coding agents are now capable of running unsupervised for extended periods, integrating directly with CI/CD pipelines. Some practitioners utilize multiple local sessions to enhance productivity.

Read at InfoQ