Interview Baye: The man who fought against the “Macau Casino”

Developers are the backbone of the mobile Internet ecosystem. They are tough and courageous. They adhere to the right values and believe that innovation can create value.

AppSo is also the same, we hope that those developers with valuable quality, and their products with flash points, can be seen by more people.

AppSo invites great developers every week to talk to them about interesting and moving stories in the process of developing products.

A few days ago, there was a paragraph that was widely circulated:

Close iMessage You won’t lose anything, but with iMessage open, you can get the whole Macau.

Spam SMS is annoying, and it’s the most crazy of “Macau casinos, online licensing”. In a sense, Baye, who developed Panda Eating SMS, may be the man who is at the forefront of fighting Macau casinos.

“Panda Eat SMS” is a very interesting product that has been on the opposite side of spam since its inception. The height of the magic is one foot, the height of the road is one foot, and the form of spam messages is changing. “Panda eating text messages” has also evolved through algorithms. Today, we went to Baye and asked him to talk to us about his fight against the Macau casino.

Get to know Baye

Network ID Baye, real name Zou Zhenlu, 80 (89), has been engaged in program development since graduation, first worked in the company, but not used to it, most of the time after working remotely, started full-time independence about two years ago. Develop and decide to make your own product.

▲ Yan value developer, article reading 10w+ only appeared

In one sentence, let’s introduce Panda Eating SMS.

“Panda Eat SMS” is an iOS app that uses machine learning to block spam messages locally.

Why would you think of developing such a product?

I was also deeply harassed by spam messages. Apple’s app after the iOS 11 release of the spam filtering interface did not meet my needs, so I decided to do it myself.

Which “blacklists” are more difficult to block?

Spam messages are generally divided into two categories by source: port SMS and peer-to-peer SMS. The former is very easy to intercept, the features are obvious, at least readable. Many of the text messages in the latter are unreadable (such as Mars), and the features often change and are more difficult to intercept. Currently only through the “user report → learn new features” cycle process interception.

What is your long-term plan for Panda Eating SMS?

In fact, it is limited to the iOS system’s permission restrictions, and there is not much room for innovation in the function of the SMS itself. In the long run, it is the continuous acquisition and learning of the latest SMS trends, so that the app always maintains a high degree of accuracy. In addition, I am also investigating the feasibility of launching the Android version.

What is the biggest challenge in the development process?

Because I may be the first in China to try to block spam on the phone with CoreML, I didn’t know if the idea was technically feasible and could meet the minimum availability requirements before development. It was not until I spent two weeks setting up the initial sample library by manual labeling and verifying that the product was feasible.

Commercial feasibility is easy to verify. The biggest challenge in the long-term maintenance that follows is how to get the model with the highest accuracy.

What is the role of the machine learning model update in the recently updated version?

The routines for most text messages are fixed, such as promotional items, which are all a few sentences. There are also a lot of text messages that like innovation, such as online games, online licensing and so on. This requires constant “collecting new samples → learning new models”It is then released to end users to protect them from the harassment of new routines.

Develop a product. Under what circumstances can you determine the right direction?

In general, it depends on the market and demand research, but I generally skip this process. I will quickly make a minimum available product, and then let the potential users try it out. According to feedback and my own thinking, I will judge the direction. right.

How do you build a good user feedback mechanism?

The most important thing about a machine learning product is continuous evolution. It can be said that the demand for data is endless. I have a built-in user feedback channel in the app that makes it easy to submit new samples. Through this mechanism all users can actively participate to help the pandas become better.

How about the feedback from users on this “Panda Eating SMS”?

User feedback is still very good, and many early adopters are also very enthusiastic to help me expand the sample library, thank you very much.

Specially unexpected, one day I received an invitation from the “12321 Network Bad and Spam Reporting Center”, and they were very interested in “Panda Eating SMS”. I went to their unit to exchange a lot of spam related questions and learned the policies and governance methods of national spam management.

Will you keep updating?

It will continue to be updated continuously, and “Panda Eat SMS” will probably stay at least no less than one month.

How do you make users like your product?

First of all, the product should be good. Secondly, the independent developer is relatively easy to establish contact with the user directly. The rapid response to the demand and feedback can enhance the user’s goodwill towards the developer.

Developing the two apps “No Phase” and “Follow-up”, what is your original intention?

Developing “No Phase” is because “App, applet auto-generation” is a very common technical topic among developers, because most applications have the same architecture.

Including includean also has a small program development platform that knows the cloud, but such services are used by the webmaster. I was thinking about why I didn’t give the end user a means to turn the website into an app, so I did “No Phase.”

Doing “follow-up” is because I will use the news to understand the operating rules and “exceptions” of the world. But a lot of representative news has no tail and can’t form a complete cognition. So I did an app that would automatically track the consequences of the news, and hope to see some events more comprehensively.

How do you judge your strengths and integrate with the industry?

Is my career a favorite thing to do? Should it be a strength? I really like and enjoy programming very much, and my products are mostly from a technical point of view. It may be easier for a technician to make a product to make something different.

What do you think about the opportunities and challenges of the App market?

Shrinking down, I think CoreML or machine learning may bring many new opportunities, such as how to improve efficiency in image processing and information retrieval, and how they can be applied to commercial products. of.

If you join Apple’s “Design Development Accelerator“, what support would you like to receive?

I personally feel that promotion is often the hardest for developers, and it would be great if Apple could help in this area.

What are your work habits?

Although the current design and prototyping tools are already very convenient, I am still used to doing technical architecture and UI design on paper notebooks, which can be deduced more quickly.

Can you share your daily workflow with us?

My daily changes are great, an efficient day might be: wake up naturally → write code → eat → write code → sleep.

Sometimes is: Naturally awake → How is this done? → Eat → Can this be done in the end? → Sleep.

What are the ways to get inspiration in peacetime?

My current inspiration comes mostly from learning new technologies, and then I will think about where the technology can be applied and how it will fall. I will also pay attention to the release of some new products and get inspiration from the product dimension.

What is your favorite hardware?

HomePod should be the most commonly used. I usually listen to some music at random to make background sounds during development. Of course, random listening is also forced, because Apple Music really can’t learn what music I like to listen to.

What do you like most about development?

See some interesting videos of knowledge, including science, politics, and documentaries. Occasionally play games.

Can you talk about your favorite game?

My current favorite game should be regarded as the “Sid Meier’s Civilization” series, you can feel the calendarThe thickness of history is too time consuming.