Can Open Source Topple ChatGPT? CSDN Founder Jiang Tao Breaks Down DeepSeek's Game-Changing Strategy

One Paper That Reshaped the Industry

In the AI world, commercial leaders tend to guard their core secrets closely — and not only do they avoid sharing them, they may actively mislead competitors to protect their technical moat. CSDN founder Jiang Tao recently offered a sharp observation on exactly this dynamic, pointing out that it's the very strategy employed by closed-source models like ChatGPT.

"The person at the top has no interest in sharing their secrets with you. Not only will they not tell you — they'll steer you in the wrong direction," Jiang said, describing the business logic of closed-source giants. Under this information asymmetry, challengers are forced to pay a heavy price, feeling their way through trial and error to figure out what works.

The Closed-Source Moat: A Business Logic Built on Information Asymmetry

At its core, the closed-source moat is a commercial barrier built on information asymmetry. Take OpenAI's GPT series: starting with GPT-3, full technical details were no longer disclosed, and GPT-4 has kept model architecture, parameter scale, training data, and other key information strictly confidential. This strategy reflects a clear commercial rationale — once a core technical path is made public, competitors can dramatically cut R&D costs by replicating or improving upon it, eroding first-mover advantage. Closed-source models also enable the API monetization model, where users must pay to access capabilities via an interface rather than deploying the model themselves, generating a steady revenue stream. Historically, Google's search algorithm and Amazon's recommendation system followed similar strategies. But in the era of large AI models, this approach faces an unprecedented challenge: as compute costs continue to fall and open-source community collaboration grows more efficient, the information walls of closed-source development are being systematically dismantled.

Then DeepSeek arrived and changed the rules of the game. The day DeepSeek publicly released its technical papers and code in full, Jiang Tao believes the industry's prevailing "myth" was shattered in an instant.

How DeepSeek Cracked the Closed-Source "Secret"

Jiang Tao used a vivid analogy to describe what DeepSeek did: "It's like scientific research — if every scientist locks up their work, science cannot progress."

According to Jiang, when American experts got their hands on DeepSeek's published papers and code and put them to the test, the reaction was immediate: "Oh my God — so that's how it works." That moment of revelation speaks volumes: DeepSeek didn't just claim a technical breakthrough — it laid out the complete implementation path and methodology for everyone to see.

DeepSeek's Technical Papers and Open-Source Practice

DeepSeek's open-source strategy is remarkably systematic and thorough — a genuine rarity in the AI industry. Between 2024 and 2025, DeepSeek released comprehensive technical reports for the DeepSeek-V2, DeepSeek-V3, and DeepSeek-R1 model series, with detailed disclosures covering core innovations such as Mixture of Experts (MoE) architecture, Multi-head Latent Attention (MLA), and the GRPO reinforcement learning training pipeline. Model weights were made available on Hugging Face. DeepSeek-R1's reasoning capabilities matched OpenAI o1 on multiple benchmarks — at a fraction of the training cost, reportedly around $6 million, orders of magnitude cheaper than comparable closed-source models. Once this cost-efficiency roadmap went public, Silicon Valley took notice immediately. Engineers at top research institutions and tech companies reproduced and verified the results firsthand, confirming the "aha moment" effect Jiang described. Technical transparency itself became DeepSeek's most powerful source of credibility.

This radical transparency is, at its core, a direct challenge to the closed-source business model. Where closed-source models built their lead on information barriers, DeepSeek put the "recipe" on the table for everyone — don't believe we pulled it off? Take it home and try it yourself. The dish you cook will taste exactly like we said it would.

Open Source: The Most Powerful Weapon for Challengers

Jiang Tao emphasized the core strategic value of open source for any challenger looking to take on an established giant.

Why Open Source Earns Trust

When a newcomer claims their technology outperforms the industry leader, they're usually met with skepticism — "nobody believes you." Open source solves exactly this trust problem. "Just say: take it home and try it yourself. I've already put it on the table — go cook the dish and it'll come out just like I said."

This "seeing is believing" approach lets technical capability be independently reproduced and verified, rapidly establishing credibility across the industry. Open source is no longer just an idealistic form of knowledge sharing — it's a shrewd competitive strategy capable of reshaping market dynamics.

Open Source Accelerates the Entire Industry

From a broader perspective, the significance of open source breaking down information barriers extends far beyond any single company's competitive position. The open-source software movement traces its roots to Richard Stallman's GNU Project in 1983 and Linus Torvalds' creation of the Linux kernel in 1991. Over the decades, the open-source model gave rise to foundational technologies like Apache, MySQL, and Python — the backbone of the modern internet — powerfully demonstrating the vitality of collaborative, shared development.

In the AI era, the significance of open source has undergone a qualitative shift. Traditional software open source centered on sharing code; open-sourcing a large AI model means sharing not just code, but also model weights (trained parameters), datasets, technical papers, and training recipes — dramatically raising both the bar and the value of open-source contributions. The emergence of open-source large models like Meta's LLaMA series, Mistral, and Falcon has created a vibrant open-source AI ecosystem. What makes DeepSeek special is that it is, to date, the open-source project that has come closest to — and in some areas surpassed — top closed-source models in performance, sparking genuine industry-wide debate: Does open source now have what it takes to challenge closed source across the board?

Just as scientific progress depends on the public sharing of knowledge, open-sourcing AI technology allows researchers worldwide to stand on each other's shoulders and collectively push the boundaries of what's possible. In a real sense, DeepSeek's open-source release helped "the whole world figure out" the core techniques that had previously been monopolized by a handful of giants.

Open Source as a National Technology Strategy

Jiang Tao goes further, elevating open source to the level of national strategy. He argues that China has made open source a strategic priority because it represents the best position to occupy at this stage of technological competition.

He candidly acknowledges that in the AI field, "we're still in catching-up mode — the US has a slight edge in innovative talent." That's an honest self-assessment. But he also points to a clear path forward: "As long as we keep embracing open source."

The Strategic Context Behind China's Open-Source AI Push

Elevating open source to national strategy reflects a profound set of real-world circumstances. With compute chips subject to export controls — the US has restricted exports of high-end GPUs to China, including Nvidia's A100 and H100 — China's AI industry faces structural hardware constraints. Open source is precisely the key path to breaking through at the level of algorithmic efficiency and model optimization. DeepSeek is the proof: by pushing algorithmic innovation to the extreme, it trained high-performance models under constrained compute conditions — a path that embodies the open-source ethos. At the policy level, agencies including the Ministry of Industry and Information Technology (MIIT) and the Ministry of Science and Technology have introduced measures in recent years to support open-source infrastructure, and domestic platforms like ModelScope and PaddlePaddle have emerged as significant players. From a global competition standpoint, for a technology follower, the significance of engaging with open-source ecosystems goes beyond technology acquisition — it's about contributing high-quality projects to participate in setting global technical standards, gradually shifting from rule-taker to rule-shaper. That is the deeper strategic meaning behind what Jiang Tao calls "the best position to be in."

For a party in the catching-up position, open source carries natural strategic advantages:

• Close the gap faster: Absorb and integrate the latest global technical advances through the open-source ecosystem, avoiding redundant reinvention.
• Build technical influence: Contributing strong open-source projects earns standing in the global developer community, gradually shifting from rule-taker to rule-participant.
• Harness ecosystem power: Open source aggregates the collective intelligence of developers worldwide, creating a positive cycle of collaborative innovation.
• Overcome hardware constraints: With compute chip access limited, algorithmic open-source innovation enables a leapfrog — DeepSeek itself is the clearest proof of concept.

Conclusion: The AI Competitive Landscape Has Changed

The core insight in Jiang Tao's analysis is this: by going open source, DeepSeek changed the rules of a game that had been dominated by closed-source giants. It demonstrated that in a frontier field like AI, transparency and sharing are not just ethical choices — they are competitive weapons capable of destabilizing the established order.

When technical "secrets" are no longer a moat, and when challengers can use open source to earn trust and accelerate their catch-up, the fundamental logic of competition across the entire industry shifts. For China's AI industry, still in pursuit mode, embracing open source may well be the most promising path to leapfrogging the competition.

Key Takeaways