🤔#
Recently, with the emergence of GPT, various products have emerged one after another. Engineers, product managers, and investors inevitably feel anxious. I have seen people on Twitter asking for votes to see if people around them are anxious, more than once (including myself lol). In summary, some of the anxiety comes from the serious homogenization of products, where no product can create an effective moat; and some comes from AI being able to flatten industries that require years of experience, knowledge, and skills in an instant.
There are two types of attitudes that I think are more typical:
- One type is that programmers (or the majority of human professions) will be completely replaced.
- Another type is The End of Front-End Development. The general idea is that, historically, there have been several occurrences, and this time is no exception, there will still be more in the future.
As a conservative optimist, generally speaking, I "choose" to believe that the future will be better in any situation, but before choosing to believe, I will try to think of the situation in a negative way (optimistic coward...?? LOL). And here, let's discuss if it is really that bleak, as an individual, are there any coping methods.
📝#
When I was writing this article, I had to start over several times because my own thoughts were reshaped in each iteration. Now, the stable understanding is as follows.
Viewpoint 1 - Small AI teams have no bargaining power, no moat, and should not pursue moats anymore#
My judgment is subjective. Although it is not as extreme as the reaction in the following figure, it is still relatively bleak: small AI teams (latecomers) are almost impossible to obtain any traditional moats (technology, brand, scale, channels, ecosystem).
A moat is bargaining power. Building a moat requires some self-reinforcing feedback loops, also known as the Matthew effect, to accumulate advantages: with users, with data, there will be better algorithms, provide better services, and attract more users. For an AI product, self-reinforcing feedback comes from the combination of the following three items: data, algorithms or pipelines, and user experience. In contrast, for small AI startups, opportunities come from
- Integrating various information sources that large technology companies lack to generate additional proprietary data.
- Algorithms or pipelines for specific niches.
- User experience for specific niches.
However, in the era of general AGI, will these opportunities still exist?
- With the emergence of AGI, integrating information sources has changed from labor-intensive work to plug-and-play integration. For large companies like OpenAI, they can almost integrate multiple data sources at zero marginal cost.
- And if large companies have such rapid integration capabilities, small companies will not have the opportunity to generate enough additional proprietary data (even if they do, it may only accumulate a few weeks' worth, which is completely insufficient).
There may be a lot of controversy about the competitiveness of algorithms, processes, and user experience for specific niches.
My personal opinion is: in the short term, there is some opportunity for niches, but in the long run, it is like the problem of "AI can't draw hands" that will be quickly solved. It's like your team has spent a lot of effort to solve the problem of "drawing hands" very well, and it has become very popular in a few months. With the emergence of new open-source tools, papers, or large models, these barriers are instantly leveled.
In other words, the existence cycle of "niche" itself will be greatly shortened. If you rely on a "moat" that depends on a certain niche, it's like many third-party app functions on iOS in the past. Once the official comes up with the corresponding function, the "moat" you have painstakingly built will collapse instantly. If it took several years from the start of the third-party to the official launch, in the era of strong AI, this cycle may be compressed to a few months or even just a few weeks.
The conclusion is that small companies have no chance to accumulate moats, so it's better to give up and think of other ways.
Viewpoint 2 - If AI is a lever, the value of "connection" will be extremely amplified by this lever#
This point is not new at all, and it has even been talked about for a long time-in the era without the Internet, the average number of people each person would come into contact with in their lifetime was only a few hundred. In the Internet era, any person sees more people passing by them every day than this number. With the blessing of recommendation algorithms, some people can be pushed in front of millions of people they don't know every day.
For example, internet celebrities in live shopping. People buy a certain product not because of the company or brand, but because it is made by this person. This trend can already be seen in the past few years with the popularity of e-commerce live streamers-people don't care about the brand of the product they buy, but they care about who they buy it from.
Of course, just pure connection, the half-life of value is very short, just like a faded internet celebrity, if the value output is unstable, even with a large number of connections, they are all ineffective links. A more effective way is to balance value output and connection scale. A successful example is the famous digital nomad, indie developer @levelsio, who currently has 260,000 Twitter followers and nearly a million dollars in annual income, all from his own continuous updates and investment.
In fact, many people on Twitter look down on levelsio and even sneer at him, thinking that he is just relying on his large number of followers, so once he creates a product, once it goes public, it can immediately generate network effects, and his technical level and product quality are just average. This is very unfair to many other individual developers-because for people like levelsio to say "You don't need a team, you just need a computer, an idea, and then go through rapid iteration and trial and error" is very unconvincing-try not using your Twitter with 300,000 followers for promotion and see how it goes. However, this is the point I want to express: having 300,000 followers, a channel that can quickly trigger network effects, is equally important as the choice of technology stack, automated deployment, and other technical issues, or even more important in the early stages.
So please be sure to establish connections: Connect! Connect! Connect by any means necessary!
Viewpoint 3 - If you don't advance, you retreat#
When I was chatting with a senior engineer at Facebook, I disagreed with one of his views. He believed that the best business he wanted to find was the kind of business where he could lie down and count money. At that time, Facebook's social network and Google's search were considered such businesses. However, as we have all seen, now even Google has pulled back its founders and entered a state of red alert. The future is a process of continuous improvement and never stopping. For those of us who advocate lifelong learning, it is not a difficult thing to accept, but for the engineer mentioned above, if I tell him-no, you can't find something that will make you worry-free for a lifetime, even if you have a mine at home, he may resist it.
Even as mentioned earlier, levelsio, after creating a product that can generate a million dollars in annual income, did not stop, but continued to launch new products and constantly tried for PMF (Product-market-fit). For such a person, it may be a short pause for reflection (for example, he has a very good reflection on why he can only achieve a scale of 100,000 to 1 million, while similar products like Lensa can achieve a scale of tens of millions), but it is impossible to lie in the past glory because it is too boring.
🤗Conclusion#
The era of companies and products no longer exists, and the future is the era of super individuals-these super individuals have super abilities, are powerful organizers, and are key connecting nodes between people, and they also disdain to stay on any existing achievements.
For my own action guide, it is to continuously produce value and increase my number of connections before things become irreparable (so-called irreparable means that the gap between individuals is magnified to such an extent that any effort is in vain) (maybe 5-10 years), the time window is really short!
Reference Articles#
- https://www.sohu.com/a/200109713_268656
- This article discusses how small startups can compete with larger, more mature companies by building strong "moats" or unique advantages to gain data advantages. The article also emphasizes the importance of data networks and how AI brings about a new type of "data network effect". As companies gain more users, they collect more data to train and optimize algorithms, improving product quality and attracting more customers to provide more data, creating a self-reinforcing feedback loop. Companies with more data can also attract top talent, bringing in more data and growth.
- https://twitter.com/BodenRing8964/status/1643782751647326208?s=20
- This tweet, although somewhat emotional, has the same line of thought.
- https://www.joshwcomeau.com/blog/the-end-of-frontend-development/