What does prompt actually stand for?
If we imagine ChatGPT as an employee with maximum execution power and breadth of knowledge, then Prompt is the instruction you give. The more detailed the instruction, the richer the context, and the clearer the output definition, the better the results you get.
Because of the strong correlation between Prompt and output results, the market highly recognizes the value of Prompt. Some speculators even advertise bundles of Prompts, while others trade them on marketplaces like Promptbase. The most popular approach is to wrap high-quality prompts in a UI and package them as SaaS products (such as Jasper and CopyAI). Except for companies with their own unique data or large models, these companies essentially rent out Prompts to gain value.
We believe that Prompt is not a commodity to be rented out or sold, but rather a content that will be shared and discussed, much like short videos and code.
After the explosion of ChatGPT, the market's awareness of Prompt has rapidly increased, and influencers in various fields have shared a large number of free and useful Prompts. As the creation threshold itself is low, more and more people are participating in Prompt creation, and the supply of Prompts is growing explosively. Jasper's users have discovered that they can use free Prompts on ChatGPT found online, and even if the effect is not as good as Jasper's, they have all gone to use ChatGPT. In the future, the market will not be willing to pay for Prompts because users have too many free options. Low-threshold and highly diverse Prompts will become a new type of content, shared, consumed, and iterated upon.
3. The Next Generation of (Natural) Programming Language
What excites us most about Prompt is that it greatly reduces the threshold for users to manipulate data in bulk, turning anyone into a software engineer who can build solutions for the information age.
Taking software as an example, code creators are programmers, while end-users are terminal users. Programmers don't know what users want, so a product manager is needed to understand users, translate their needs into product functionality, and then implement it for programmers. This process involves a lot of communication and information loss, making it very inefficient. The reason for this process is that end-users don't have the time and skills to write code and build software.
The advent of Prompt completely changes this process. From now on, you only need to describe the problem and the corresponding steps in natural language to build a personalized solution, eliminating the steps of having a dedicated team to find and iterate on PMF products because users are the best product managers themselves.
Compared to code, Prompt has much lower creation thresholds, stronger capabilities, more use cases, and therefore, a larger volume.
4. The Vehicle of Product Imagination
In the past, people needed imagination, professional skills, and time to create something. For example, if you wanted to write a story, you had to first conceive and imagine the content of the story, and then use writing skills to implement it. Or if you wanted to create an app, you had to first conceive of its functionality and interface, and then use design and programming skills to implement it. This could require one person or a team. The problem is that the cycle to develop these skills is long, and the costs are high, which is why there is division of labor.
However, the emergence of generative models has made anyone possess industry expert-level skills. To implement any idea, you only need to describe it with a prompt, and it can be quickly implemented and iterated upon. Prompt has become the vehicle of imagination.
Before GPT appeared, there was no point in accumulating imagination, because only the final product would be used or viewed. Therefore, what we saw were platforms that accumulated finished products, such as the App Store, YouTube, and Spotify. Now, imagination is the finished product.
The ultimate form of Prompt engineering is the exploration of large model usage scenarios. Currently, we face a very real problem that people do not know what AI can be used for. For example, before hearing about HustleGPT (using ChatGPT as a boss to help you make money), we had no idea that ChatGPT could be used in this way. Even when we see the various Prompts uploaded by everyone on FlowGPT every day, we are shocked by their imagination. We had never thought of such usage scenarios before.