The Open-Source Code of DeepSeek: A Discussion on the Spirit Behind It

Recently, DeepSeek has become a hot topic across various industries, to the point where even friends I haven't heard from in years have started asking me on WeChat: "What exactly is DeepSeek?" Most of the time, I just say: "It's a super powerful open-source model that can think and reason on its own. Open-source means they release all the code, and anyone can use it freely."

After I explain it this way, the next question usually is: "Why? Why would they give away something they've created for free?"

The answer lies in the "open-source spirit."

You might be curious, what exactly is the open-source spirit? To answer this, I recommend reading a classic book titled The Cathedral and the Bazaar.

Today, I'll use the ideas from The Cathedral and the Bazaar to explain what the open-source spirit is all about, and also discuss the impact that this surge of DeepSeek's open-source models might have on the technological development of society as a whole.

1. No Ivory Towers Here, Only Pure Garage Spirit

Before diving into sharing the content of this book, let me first share some information with you.

On February 21, 2025, at 12:00 PM, the DeepSeek team released a statement.

Such an action is incredibly bold.

You might be wondering, what's the difference between this and the previous release of the DeepSeek R1 model? Let me put it this way: the previous release was like a master craftsman sharing all of his knowledge selflessly with you. But now, it's not just about getting the craftsman's secret methods—you can also see how the craftsman works step by step. For many people, this process is even more valuable than the end result itself.

As expected, this announcement sparked a strong reaction from netizens around the world. The first comment read: "Real Open AI!" They intentionally left a space between "Open" and "AI," and this remark hit the nail on the head. It was not only a compliment to DeepSeek but also a subtle jab at OpenAI, led by Sam Altman.

2. The "Gears of Fate Begin to Turn" in Open Source

Now, let's get to the main topic.

Let's start with the title of the book. You might be curious why a book about technology culture has the words "Cathedral" and "Bazaar" in it. What do they mean?

In fact, the author uses "Cathedral" and "Bazaar" as metaphors to represent two different development models. The "Cathedral model" is like a grand, top-down plan: you set long-term goals and plans, which could span five or ten years. Then, a small group of leaders or senior experts take charge, with layers of management and collaboration, all working step-by-step until the final product is completed. Only when it is released does it make a big impact.

On the other hand, the "Bazaar model" is like a bustling open market, with vendors selling vegetables, fish, and other goods, and shoppers wandering around. There is no unified leadership, and no clear development goal or profit strategy. Everyone operates independently, in a somewhat chaotic fashion, yet it still maintains unique vitality, full of endless possibilities and energy.

A quick aside: on April 17, 2024, Baidu held its 2024 Baidu AI Developer Conference. At the event, founder Robin Li not only released their new model but also discussed open-source models. He said, "In the past, open-source was considered cheap, but in the context of large models, open-source is actually the most expensive. Therefore, open-source models will increasingly fall behind." However, less than a year later, DeepSeek burst onto the scene, and soon after, Baidu announced it would release its Wenxin model 4.5 series in the coming months, officially going open-source on June 30 of that year.

The point of sharing this example is to show that Baidu, by picking a specific date to announce their open-source plans, was positioning it as something to be heavily promoted. This suggests that Baidu wasn't fully committed to open-source in the true sense, but was rather adopting it as a defensive strategy. In theory, this isn't exactly the open-source spirit.

So, what is the open-source spirit?

In simple terms, open-source is when a tech person or team selflessly shares all of their code with the world. Developers everywhere can freely access it, learn from it, and even modify and apply it to their own projects. As I mentioned earlier, GitHub is the largest open-source community in the world.

Now, returning to The Cathedral and the Bazaar, the author, Eric S. Raymond (often referred to as ESR), is not only an experienced programmer but also a flagbearer of the open-source movement.

The book's core case study is the story of the open-source operating system Linux. While most ordinary users are more familiar with Microsoft’s Windows or Apple’s macOS, the reality is that many of the websites, smart appliances, and even routers we use every day run on Linux. Additionally, the core of Android phones is also based on Linux. So, in fact, most of us are Linux users without even realizing it.

Behind Linux, there is actually a very meaningful story.

In 1991, a Finnish university student named Linus Torvalds, in his spare time, decided to try out a UNIX-like operating system that was typically only found on large machines. He started re-writing the kernel on his home computer.

At first glance, this seemed like just a fun project by a student. But Linus didn’t stop there. He uploaded his source code to an internet community and invited other programmers to come and contribute.

To borrow a somewhat cliché phrase, "the gears of fate began to turn." This became a significant milestone in the open-source movement. As more and more programmers got involved, fixing bugs and improving the code, Linux gradually grew stronger. Because everyone could see the source code, modify it, and contribute their changes, over time, the Linux community evolved from a "student project" into one of the world’s most active and mature open-source operating systems.

3. Three Core Principles of Open Source That Left a Deep Impression on Me

Of course, besides telling the story of Linux, this book also delves into the core philosophy of open-source and explains why the open-source movement has been successful. Among the many insights, there are three underlying principles that left a deep impression on me:

1. Great Software Products Often Stem from the Developer’s Personal Needs

In the open-source world, whether it's the project initiator or the global contributors who submit bug reports and fix code without compensation, many people participate simply because they themselves need the tool.

It’s important to note that in the real world, many programmers aren't particularly passionate about the software they develop. They often view programming as just a job—getting paid a decent salary every month and completing tasks as required. But in the open-source community, the situation is completely different. Many experienced programmers, even those who are already millionaires, still volunteer to update and maintain open-source projects because they genuinely love the work and care about it. This is a perfect example of "passion-driven" work.

2. Release Early, Release Often, and Listen to User Feedback

In the open-source development model, releasing early and frequently is crucial. At first, many seasoned developers tend to think that early versions of software will have many issues. If released to users too soon, the poor experience might drive them away.

Traditional development models usually follow a "cathedral" approach. Teams of elite developers work on a project for six to eight months, testing it, before releasing a stable version.

However, Linux's innovation lies in its fast-paced release strategy. In the early stages, updates were often released multiple times a day. This allowed early testers and contributors to quickly identify and fix problems.

That’s why Linus Torvalds, the creator of Linux, once made a very interesting remark: “The more eyes you have, the easier it is to catch bugs.”

In a "cathedral" development model, bugs are often deeply hidden and hard to uncover, requiring months of work and multiple rounds of checks to identify.

But in the "bazaar" model, when thousands of developers are excitedly working on each new version and running the system on various devices, problems are discovered more quickly. Furthermore, the people who spot the bugs often fix them themselves to meet their own needs.

3. Free Open-Source Software Is Not About Moral Superiority

The author also emphasizes that we shouldn’t view open-source developers or teams as "altruistic" and praise their noble character. He points out that good open-source project initiators gain a lot in return, such as influence within their community or solving technical problems with the help of others. Like closed-source projects, open-source is simply a development choice and doesn’t mean it’s more noble than proprietary software.

Of course, the author is being somewhat modest here. In sharing his own open-source projects, he mentions that when you truly acknowledge that you owe others a lot, you'll find that the entire open-source community sees you as the creator of the software, and you'll receive a flood of praise emails. These positive feedbacks keep coming, telling you how your project has helped people. Also, when you adopt others' bug reports or modified code, the positive feedback for every contributor is equally strong.

So, I believe DeepSeek's decision to make their code repository and working details public next week is likely because they've received a lot of positive feedback in the past month. This feedback has motivated them, as part of the open-source community, to contribute every step of their exploration.

Just like how Linus Torvalds, the Finnish student, set the gears of open-source in motion, have you thought about what kind of impact DeepSeek’s move will have on the world? Could new gears already be silently turning?

4. The Tsunami Caused by DeepSeek Has Just Begun

Before diving into the specifics of DeepSeek's open-source initiative, I want to share a brief aside.

In December 2024, a Norwegian disaster series Escape from Volcano Island was released on Netflix. It tells the story of a fictional volcanic eruption on La Palma Island. Initially, smoke began to rise from cracks at the top of the volcano, which grew larger and eventually erupted. However, strangely, the eruption wasn’t the climax of the disaster. The volcanic ash, mixed with rocks, slid into the sea and initially resembled a mudslide. But as the waves spread, they eventually formed a massive tsunami that swept across neighboring islands.

Why am I mentioning this underrated series with a low rating? Because when I saw the volcanic eruption, I thought that was the biggest disaster. I never imagined that a tsunami would follow. Looking back, DeepSeek's situation is similar—the progress so far has been like the eruption, but the real massive impact on various industries is only just beginning.

For example, DeepSeek’s decision to open-source their code repository is far more valuable to me than the previous release of their R1 model.

Why is that? Simply put, R1 was just the result of their internal training—an output they shared for everyone to use, allowing people to develop their own products based on it. That alone was impressive. But now, they’re not just releasing the results; they’re also sharing the details of their daily work. What does this mean? If their training method doesn’t work, teams around the world can immediately change direction. If it works, other teams will quickly be inspired, sparking even greater innovation. This embodies the "release early, release often" mentality, and DeepSeek is taking it to the extreme—they’re open-sourcing their code daily, continuously driving innovation.

At the end of last year, we were still talking about how the AI field lacked a killer app. Well, DeepSeek is now the first killer app.

In my opinion, the tsunami caused by DeepSeek will have two main waves.

First Wave: A Massive Boost to Personalized User Experiences

Now, many independent developers or ordinary users can leverage DeepSeek’s model to accomplish tasks that were previously unimaginable. Just like the AI image model Stable Diffusion (SD), which was open-sourced in August 2022, it quickly gave rise to various illustration-generating applications. What's even more impressive is that on GitHub, there are now over 1,000 projects related to SD. This demonstrates how the success of one open-source project can trigger the birth of many more.

Many ordinary users who never interacted with AI tools before can now use DeepSeek to write copy or research information, significantly boosting productivity. As more people use it, we will see even more AI-based products emerge, specifically designed to solve personalized needs. This is true technological inclusivity.

Second Wave: Lowering the Bar for AI Adoption, Especially for Small and Medium Enterprises

In the past, if an enterprise wanted to build an internal system, they usually had to outsource, go through bids, and face various barriers. If they wanted to build an AI application, it would be even more complicated—limited by talent, technology, and resources. But now, thanks to DeepSeek’s open-source model, individuals within companies who have ideas can quickly build their own enterprise applications after some research.

Moreover, powerful open-source models like R1 have made it so that for all non-AI companies, the underlying model is no longer a bottleneck nor the core competitive edge, since everyone can access it. This means companies will focus more on their core data, unique industry insights, and the ability to implement AI in real-world scenarios. These will be the new competitive advantages.

It’s not just Baidu, Tencent, or Alibaba embracing DeepSeek—teams like ours, which already have software development capabilities, are doing the same. Even my family’s smart lock brand, which I manage, released a WeChat article announcing that their smart locks are now powered by DeepSeek.

At that moment, I had a flash of inspiration as a product manager. I started to imagine: What could happen if a smart lock had reasoning abilities? Could it decide whether to open based on your mood as you stand at the door? Or, if installed at the door of a certain person, could it detect a visitor and alert the spouse if there was an affair?

Of course, I’m joking, but it clearly illustrates one point: integrating AI into products is no longer a difficult task.

Returning to DeepSeek, the core of their success is their commitment to open-source.

By releasing the R1 model as open-source, DeepSeek allowed global developers to download it, try it out, and be amazed by its capabilities. Not only that, but they also made the R1 model's papers and reports available, attracting attention from technical experts who examined their innovations in reinforcement learning. These technical breakthroughs have allowed Silicon Valley's tech community to view DeepSeek—and China’s AI products—on equal footing.

Speaking of DeepSeek's R1 model and its reasoning ability: The first AI model with reasoning capabilities was OpenAI’s o1. While o1 could reason, OpenAI didn’t showcase the reasoning process, considering it not the focal point. DeepSeek, on the other hand, boldly displayed its reasoning process step by step in text. For me, this process was more moving than the final output itself.

Now, DeepSeek is about to release their code repository, letting developers worldwide see what their team does every day.

The message they released feels like a declaration for the open-source world. The final sentence really lights up the entire conversation: “No ivory towers - just pure garage-energy and community-driven innovation.”

I sent this quote to DeepSeek and asked if they could write the conclusion for my article. They replied, “The frescoes of the cathedral will eventually fade, but the ledgers of the bazaar remain ever fresh.”

Source:https://www.dedao.cn/share/course/article?id=0mPqglk6GzZwKr5p2nXMLBEO3ba2AR