header-langage
简体中文
繁體中文
English
Tiếng Việt
한국어
日本語
ภาษาไทย
Türkçe
Scan to Download the APP

Anthropic Releases 400k Session Study: Business Acumen as Important as Code Knowledge, Non-Programmers Equally Successful as Software Engineers

According to Perceive Beating monitoring, the latest large-scale intelligence agent programming report released by Anthropic points out that, with the assistance of Claude Code, programming skills themselves are becoming irrelevant, and the key to success lies in management and domain knowledge. The analysis shows that Claude Code users are, on average, active up to 20 hours per week. Among the top ten major professions, the success gap in code writing between non-programmers using intelligent agents and professional software engineers has narrowed to within 7%, with the success rate of managerial positions even slightly surpassing that of professional programmers.

The large-scale model has reshaped the division of labor between humans and machines, with users driving about 70% of the strategic direction decisions, while the intelligent agents take on around 80% of the specific execution decisions. Business experts who understand how to break down and standardize delegated tasks can guide the large model to achieve greater efficiency. A single expert-level session can trigger 12 actions and 3200 words output, surpassing novice-level sessions by more than 5 times. In contrast, novices are extremely prone to giving up when encountering errors, with a dropout rate of up to 19% in novice sessions. Even when facing trouble, intermediate and expert-level users have a dropout rate of only 5% to 7%. Research has shown that with just moderate industry-specific knowledge, one can seamlessly overcome the programming barrier with the help of intelligent agents, leading to a significant leap in personal productivity.

Developers' use of intelligent agents is transitioning from debugging code to end-to-end autonomous creation. Within 7 months, the proportion of debugging sessions used to fix broken code has halved from 33% to 19%, while the proportion of direct deployment, data analysis, and writing non-code documents has doubled. Users are also starting to delegate more complex and high-value work, with the average estimated value of tasks increasing by about 25% over 7 months.

举报 Correction/Report
Correction/Report
Submit
Add Library
Visible to myself only
Public
Save
Choose Library
Add Library
Cancel
Finish