Why DeepSeek Is Gaining Traction Beyond ChatGPT: Insights from the Global Developers Conference

At the Global Developers Pioneer Conference, the speaker was most impressed by DeepSeek.

A friend asked why DeepSeek seemed only slightly better than ChatGPT yet was so popular.

The answer lies in the classic "timing, geography, and people" (天时，地理，人和) factors.

Timing

DeepSeek launched just before Chinese New Year, sparking patriotic enthusiasm and heightened interest in domestic AI; the holiday gave users ample time to try and share the product, creating strong word‑of‑mouth momentum.

Geography

China has a vast user base for large‑language models, but prior domestic models were weak. DeepSeek’s lower training compute, performance comparable to top models, and cost‑effectiveness—costing only a few percent of foreign alternatives and free for individuals—quickly attracted users.

People

The product’s strong reinforcement‑learning techniques inspired the developer community, suggesting that AI could eventually surpass human capabilities in areas like autonomous driving; it also runs on domestic GPUs with modest requirements and is partially open‑source, while U.S. restrictions add urgency.

The speaker notes that AI’s impact across programming, education, and office work is prompting schools to introduce AI literacy courses.

By 2025, large‑language models may achieve a "Qingbei moment," scoring at top university admission levels.

By 2027, humanoid robots could match human athletes in capability.

By 2027, internet data for training will be exhausted, making algorithmic innovation essential.

Sam Altman’s advice includes several coding assistants: CSDN InsCode (Chinese), SenseTime Xiaowanxiong (Chinese AI), GitHub Copilot (Microsoft), and Cursor.

The article also recommends the book "揭秘大模型：从原理到实战" for those interested in large‑model fundamentals.