Recent trends in software engineering and artificial intelligence indicate a shift from an emphasis on scale and speed to a focus on effectively managing context. This change follows years of industry belief that progress in AI was primarily driven by these two factors.
In February 2025, the term “vibe coding” was introduced by Andrej Karpathy, generating significant discussion within the tech community, including at Thoughtworks. Some experts expressed skepticism regarding its implementation. During an episode of the Thoughtworks technology podcast in April, concerns were raised about potential issues arising from vibe coding and its influence on software development practices.
The introduction of vibe-based coding has led to increased instances of complacency, particularly regarding AI-generated code. The latest Technology Radar has highlighted this concern, noting that as users began to demand more from AI models, including larger prompts, the reliability of these models started to decline.
This situation has spurred heightened interest in the importance of context in engineering. Collaborating with coding assistants like Claude Code and Augment Code has underscored the necessity of providing relevant context to ensure the outputs remain consistent and reliable. Such measures are expected to enhance software quality while potentially increasing productivity by minimizing rewrites.
When utilized effectively, generative AI can aid in understanding legacy codebases and even operate without full access to source code, provided that appropriate context is established. Interestingly, insights from using generative AI for forward engineering suggest that abstraction from the details of legacy code can enhance effectiveness, as it opens a wider solution space and capitalizes on the creative capabilities of AI models.
The emergence of agents and agentic systems necessitates a reevaluation of how context is perceived in the industry. These systems demand considerable human involvement to navigate complex and dynamic environments, moving away from a solely vibe-driven approach.
Source: https://www.technologyreview.com/2025/11/05/1127477/from-vibe-coding-to-context-engineering-2025-in-software-development/