AIEd Knowledge BaseAIEd KB
On this page

Research Article

Computer Science Achievement and Writing Skills Predict Vibe Coding Proficiency

created 2026-09-08 · updated 2026-09-08 · vibe coding, cs education, prompt engineering, writing education, computational thinking, higher ed, ai literacy, generative ai

In a preregistered cross-sectional study (N = 100 tertiary students), both written-communication proficiency (r = .29) and computer-science achievement (r = .39) significantly predicted GUI-oriented "vibe coding" performance, with CS achievement remaining a significant predictor after controlling for domain-general cognitive skills — a finding that links prose skill to LLM-driven development; in a joint model CS achievement contributed roughly twice the unique variance of writing skills, though both added independent predictive value, and prompt quality mediated the link from writing skill to vibe-coding success. Findings speak to tool and curriculum design — when to emphasize prompt-writing versus CS fundamentals.

Relevance to AI in Education: This study reframes writing as a primary programming skill in the era of LLM-native software creation, connecting directly to Prompt Engineering research, the construct of vibe coding, CS Education curriculum redesign, and the predictors of who succeeds when AI handles implementation. It provides controlled, preregistered evidence that complements the practice-oriented vibe-coding papers already in this knowledge base.

Key Findings

  • Both writing and CS achievement predict vibe coding. In a preregistered, cross-sectional study (N = 100 tertiary students), written-communication proficiency correlated r = .29 with vibe-coding performance and CS achievement r = .39; both were significant independent predictors in a joint model, where CS achievement contributed roughly twice the unique variance of writing skill.
  • CS achievement survives controlling for general cognitive ability. CS achievement remained a significant predictor (partial r = .281, p = .005) of vibe-coding performance after controlling for domain-general cognitive skills (ICAR16), indicating the relationship is not merely an artifact of general reasoning.
  • Writing works through prompt quality. Human-graded prompt quality mediated the association between writing skill and vibe-coding performance — response-process evidence that clear, structured prose translates into better natural-language prompts, supporting the construct validity of the measure and the importance of written communication in LLM-native development.
  • CS knowledge helps even in "no-code" mode (decomposition and algorithmic thinking are facets of Computational Thinking). Because the environment hid the generated source, CS achievement could only operate indirectly (problem decomposition, mental models of control flow); the authors argue this makes their CS estimate a lower bound for general LLM-augmented programming where code editing is also available.
  • The pattern extends beyond GUI surface tasks. In the data-centric, constraint-satisfying meal-planning subtask, both CS achievement (r = .320) and writing (r = .202) remained significant predictors, tentatively suggesting the findings are not limited to surface-level GUI construction.
  • Exploratory: prior LLM usage negatively correlated with writing and vibe-coding performance (though not with CS achievement) — the authors speculate LLMs may blunt expression, or weaker writers self-select into heavier LLM use; not causal.

Connected Concepts

Connected Articles

Citation

Thorgeirsson, S., Weidmann, T. B., & Su, Z. (2026). Computer Science Achievement and Writing Skills Predict Vibe Coding Proficiency. Proceedings of the 2026 CHI Conference on Human Factors in Computing Systems (CHI '26), Barcelona, Spain. 17 pages.

SourceHistoryEditChat

Embed this page

Copy the code below to embed a chromeless version of this page in a learning management system or other website. The embedded view hides the site header, navigation, and footer.