GPT has been in business for four months, earning tens of dollars

Image source: Generated by Unbounded AI tool

“GPT has been in business for four months, and the results have been dismal. From March to now, it has been four months. It has invested five to six thousand dollars, with five to six thousand users, and the income is tens of dollars…”

Under the AIGC upsurge, some people sell courses, sell books, and make a lot of money as members, and some people use AI painting tools to open Taobao stores to customize avatars, make wallpapers, and receive orders. However, not everyone who sells water makes money.

A personal developer nicknamed “Shenhui” attracted a lot of attention after he shared his entrepreneurial experience on a certain platform two days ago. He began to develop a website based on GPT in March, realizing the functions of networking, drawing, modifying the theme, saving logs, and even charging phone charges. However, from March to now, ** has invested five or six thousand yuan, registered users five or six thousand, and earned only a few dozen yuan. **

In his opinion, the GPT website he made after spending a lot of time, staying up all night researching and writing codes is much more powerful than other shell GPTs, but due to lack of promotion experience in the C-end market, he failed to make the product popular. up, so the earnings are dismal.

In the interview, Fukae told Babbitt that he used to be a product manager, and the code was mainly self-taught, and many codes of the entire website were also written by GPT. Fukae’s website is called “oncetalk”, and it has rich functions. It provides different dialogue templates and extension plug-ins, including image generation, text dialogue, novel writing, copywriting, bookkeeping, role-playing, etc.

Compared with other similar websites, the main highlight of oncetalk is the extension plug-in. For example, the browser plug-in implements a browser mode similar to the official ChatGPT, which allows AI to call the search engine according to the situation, then analyze the search results, and obtain them in a loop and summarizing the content in the best matching links until you find something that meets your needs.

The author experimented a bit, and it is indeed more effective, and it is interesting to see the workflow of AI, but after in-depth use, I still find that many tasks are not completed well enough, and reading links will fail, which is far from the official GPT There are still many gaps in the browser mode.

In the plug-in for modifying the theme, you can let the AI modify the theme color of the website directly through dialogue. In addition, there are some picture generation, phone charge recharge (currently off the shelf), smart bookkeeping, etc.

Why can’t it fire up? Babbitt screened user messages and concluded the following reasons:

First, first-mover advantage. After ChatGPT became popular at the end of last year, many individuals quickly launched a shell website. Even if the function is very simple, as long as the pain points of domestic users who cannot access ChatGPT without scientific tools can be solved, many users can be gained.

Second, user pain points. For example, using GPT call credit, it looks very cool. However, there are many third-party recharge platforms, and most of them have discounts. Newly developed websites must first obtain user credit, and then compete with these mature third-party platforms.

Third, product positioning. The product features are rich and all-encompassing, but the core focus has not been found. You can turn to a smaller and more vertical track, find a core crowd, and play a selling point.

Fourth, programmer thinking. It is easy for programmers to inject too much energy into the back-end link when doing projects, while ignoring the importance of front-end traffic.

After working full-time for half a year, achieving such results will inevitably affect the mentality. However, the greatest impact on deep drawing is the ability of GPT. “I imagined a lot of scenarios when I was able to do it at the beginning, but in the process of doing it, I found that GPT is not so intelligent. I made an experimental AI social application before, and I want GPT to generate sql statements according to the needs and analyze the statements by myself. Security, and then execute, and then generate HTML from the query results to display to the user, the implementation is possible, but not ideal.” In addition, the limitations of the GPT context are also quite large, and more complex tasks cannot be completed.

Regarding future plans, Fukae told Babbitt that the entrance to the unified dialogue will probably return to the hands of major manufacturers, so I still focus on subdividing the field, first improving the experience of the website and plug-ins, and accumulating more experience, which may be done later B-terminal and G-terminal.

Reference link:

GPT has been in business for four months, and the results are dismal

Experimented with the ability to implement social applications through GPT‌