The ability to assess code proficiency is essential for understanding how developers comprehend and apply programming concepts effectively. With the rapid adoption of AI-assisted code generation tools, such as large language models (LLMs), this issue has become increasingly urgent. While these tools can produce functional code snippets on demand, developers must still comprehend, verify, and modify the generated code to ensure correctness and efficiency. The gap between code generation and code understanding exposes a critical challenge. Therefore, a systematic and objective way to measure code proficiency is vital not only for human learning and assessment but also for ensuring responsible and effective use of AI-generated code. Despite prior efforts to measure programming proficiency, such as the \textit{pycefr}, which adapts the Common European Framework of Reference for Languages (CEFR) to Python remain subjective. They rely on expert opinion or informal surveys, lacking empirical validation and a reproducible grounding in how developers actually comprehend programming constructs. This study addresses that gap by introducing an automated, data-driven framework to determine code proficiency levels using introductory programming textbooks as empirical ground truth. Our framework leverages the pedagogical sequencing of textbooks to derive proficiency levels for 138 Python code constructs. We propose two automated methods: (1) Übersequence, a consensus-based sequence that statistically models the progression of code constructs across textbooks, and (2) Clustering, a network-based grouping of constructs using the Louvain community detection algorithm. These methods reveal both the linear order of increasing conceptual difficulty and the relational grouping of constructs typically introduced together. Our results show strong consistency among textbooks, demonstrating significant positive correlation in construct sequencing and ensuring that textbooks are suitable as an empirical basis for measuring code proficiency. The clustering analysis yields six stable construct groups. Early groups represent basic constructs, while later ones capture advanced concepts. This work offers a robust alternative to subjective expert ratings and provides a methodological foundation to calibrate proficiency in the era of AI-generated code. Enabling future systems to assess, align, and personalize developer learning and code comprehension. By integrating empirical pedagogy with automated analysis, this framework advances the pursuit of reliable, scalable, and AI-aware models of code proficiency.