Qodo just released a report on 2025 state of AI code quality. As a pretext, it is important to note that there has been only 609 developers involved in this survey, and unfortunately there’s no information about these participants. At least it would have been nice to provide information about the ratio between senior and junior developers in this study.

Some of the findings that I find interesting:

When we asked developers experiencing “context pain” what they most want from their AI tools, one answer stood out: richer contextual understanding.

Moreover:

Manually selecting context for every prompt — files, functions, folders — might have worked in early tools, but it doesn t scale. It’s tedious, error-prone, and leads to frustration when results still miss the mark.

• 54% of developers who manually select context say the AI still misses relevance
• That frustration drops to 33% when tools choose context autonomously
• And falls even further to 16% when context is persistently stored and reused across sessions

This pretty much resonates with my own findings.