Summary of Key Points
This article focuses on the competition between two approaches in the AI field: Cui Tianyi from DeepSeek draws inspiration from the idealistic spirit of Blizzard’s peak era, adopting a “people’s war” strategy that combines an open-source ecosystem with a sense of mission to attract developers worldwide for its Harness framework; whereas OpenAI, once an idealistic organization, has evolved into a “precision-driven machine” focused on commercial interests. The article contrasts the fundamental nature of these two approaches—whether to recruit top talent with high salaries for closed models (elitism) or to engage developers through a sense of achievement and the desire to make history (a people’s war)—and suggests that DeepSeek’s approach may be closer to the ultimate goal of AGI (Artificial General Intelligence).
I. Cui Tianyi’s “Blizzard-style” Recruitment: “Silicon-based Statues” Are More Inspiring than Money
Cui Tianyi’s method of recruiting beta testers for Harness is particularly appealing: he posted on X, and developers eagerly applied (800 people from 18 different fields). He even directly invited interested individuals to submit their resumes in his tweets, almost like a boss pursuing potential employees. How does he manage to attract so many people? The article explains this by citing Blizzard’s practice of honoring deceased employees and players as NPCs in games (for example, the “Temple of the Soul of Warriors” dedicated to a 19-year-old artist), allowing their names to live on in the game. Developers seek more than just money; they want to feel respected, achieve self-fulfillment, and even be remembered in history—much like having a “permanent statue” built in the digital world. Cui Tianyi uses this spirit to inspire developers, making them feel that their contributions will be recognized by their peers, which is more compelling than an annual salary of $500,000.
II. The “Plugin Magic” of the Harness Framework: Why Is It Called an AGI Accelerator?
The core of DeepSeek’s Harness framework is its plugin-based design. It is not a fixed tool but an open platform where any useful feature can be developed as a plugin and even given a permanent name (similar to the VS Code plugin market, where developers can create plugins for free with their names permanently displayed). The advantage of this approach is rapid evolution: as developers from around the world contribute plugins, the ecosystem grows like a snowball. Liang Wenfeng, the founder of DeepSeek, emphasizes that everything is aimed at achieving AGI. Even if Harness is still in its early stages, it doesn’t matter—just like a racecar that doesn’t stop to pick up hitchhikers for short-term profits but focuses on reaching the finish line (AGI). The plugin-based system allows for rapid iteration, far faster than closed models.
III. OpenAI’s “Evolution into a Dragon”: From “Benefiting Mankind” to “Frenzied Money-making”
OpenAI started as a non-profit organization with the goal of making AGI beneficial to all humanity in 2015. However, it adopted a profit-limiting model in 2019 for funding purposes. In 2023, Sam Altman was fired for secretly pushing commercialization efforts, and by 2024, OpenAI’s valuation had reached $80 billion with revenue approaching $10 billion. It has now transformed into a “precision-driven business machine” that recruits top talent, uses closed models, and sells APIs and enterprise services. Although ChatGPT has seen rapid user growth, it has lost its initial appeal; developers no longer join for idealistic reasons but for money. Similar to Blizzard, which became profit-driven after being acquired and abandoned by its fans, OpenAI has also forgotten its original purpose.
IV. The Competition Between the Two Approaches: “Luxury Apartments” or “Drawing a Circle?”
The article compares the two approaches as follows:
- OpenAI’s elitism offers a “300-square-meter luxury apartment by the Huangpu River” (current high salaries and stable income), but the focus is on making money, deviating from the priority of achieving AGI.
- DeepSeek’s people’s war approach provides the potential for “drawing a 50-square-kilometer circle by the South Sea” (future opportunities and self-fulfillment), with commerce being a secondary aspect of the ecosystem, with the primary goal being the advancement of AGI.
The author believes that the latter approach is more likely to lead to the realization of AGI, as the ecosystem evolves faster than closed teams. This is similar to how VS Code’s plugin-based ecosystem outperformed many paid editors.
V. China’s AI “Latecomer”: Can DeepSeek Overcome OpenAI?
China already has competitive AI products like GLM and Kimi, and DeepSeek is seen as a latecomer in the AI arena that is steadily developing. This competition represents a clash between idealism and commercialism. Just as OpenAI once was a visionary organization but has become a commercial entity, can DeepSeek use its people’s war strategy to overcome it? The author finds this outcome promising, as the essence of the internet spirit lies in openness and collaboration, which is precisely what DeepSeek is testing.
In essence, this article questions the future of the AI field: will it be dominated by commercial machines or return to the open and collaborative internet spirit? DeepSeek’s efforts may provide an answer to this question.