Anthropic releases 400k conversation study: understanding business is more important than knowing code, non-programmers have success rates on par with software engineers

Goldfinch Finance reports that the latest large-scale intelligent agent programming report from Anthropic indicates that with Claude Code assistance, programming skills themselves are becoming less relevant, and the key to success lies in management and domain knowledge. Analysis shows that Claude Code users are active for an average of up to 20 hours per week. Among the top ten major professions, the success rate of non-programmers using intelligent agents to write code has been reduced to within 7% of professional software engineers, with management roles even slightly surpassing professional programmers in success rate.
Large models have reshaped human-machine division of labor, with users leading about 70% of planning decisions and intelligent agents handling approximately 80% of specific execution decisions. Business experts who understand how to decompose and standardize delegated tasks can guide large models to perform more effectively; expert-level conversations can trigger 12 actions and produce 3,200 words per session, more than five times that of novice-level conversations. In contrast, novices are very likely to give up when encountering errors, with a frustration abandonment rate of up to 19%, while intermediate and expert users, even when facing difficulties, have abandonment rates of only 5% to 7%. Research proves that just moderate industry-specific knowledge in a particular field can enable seamless crossing of programming barriers through intelligent agents, resulting in a leap in personal productivity.
Developers' use of intelligent agents is shifting from debugging code to end-to-end autonomous creation. Over seven months, the proportion of debugging sessions used to fix broken code has halved from 33% to 19%, while the share of direct deployment, data analysis, and writing non-code documents has doubled. Users are also beginning to delegate more complex and high-value tasks, with the average estimated value of tasks increasing by about 25% over the seven months.
View Original
This page may contain third-party content, which is provided for information purposes only (not representations/warranties) and should not be considered as an endorsement of its views by Gate, nor as financial or professional advice. See Disclaimer for details.
  • Reward
  • Comment
  • Repost
  • Share
Comment
Add a comment
Add a comment
No comments
  • Pinned