Independent Development Diary 19: Just Writing Randomly

I've been lazy a lot lately, so I haven't written for a long time.

The main reason is that there are too many things to do, and I have spent a lot of time learning Thai recently. My friend asked me why I want to learn a small language.

However, I feel that Thai is a patched language, with too many rules, and some rules are outdated. It does not have one space for each word like English, and is only separated by sentence spaces, so you need to segment the words yourself, and the rules for segmentation are all weird. Then the pronunciation depends on the word, and there are many special grammars that change the pronunciation. Just how to read a word requires learning two books of grammar, so you have to spend more time to understand the rules.

By the way, let’s talk about some interesting things that we have encountered recently. One is the AI coin speculation competition. Someone used the current mainstream domestic and foreign large models to do automated trading, and the entire process was broadcast live online. The result was that the domestic large model won. This result is not unexpected. After all, Deepseek is a quantitative person, but Qwan has come out. The only thing I can think of is to use Deepseek's data for training. Of course, it is just a guess, don't criticize.

Another relatively new thing is that Coinbase has launched an X402 protocol, which uses the HTTP status code 402 that has not been used before to make a micropayment protocol. Users do not need to spend extra gas. They can pre-sign through usdc and send it to a trusted third party for forwarding. The merchant will complete the service after receiving it.

This can replace most scenarios that require small payments. For example, for domestic students, subscribing to some large-model services often gets stuck in the payment process. If these areas are opened in the future, it will be much more convenient for everyone to pay. Of course, as a developer, I am more interested in the specific development process.

So I made two projects, one is digital product sales, that is, after users pay, they can see the content published by the publisher. During the docking process, if you want to use coinbase as the verification server, it has regional restrictions. Many regions do not support it, and the documentation will not tell you. It will only return a 403. It took me 2 days to troubleshoot this problem. I always thought that I did something wrong. The third-party verification server is, firstly, unstable, and secondly, of course, it is a trust issue.

Agentory.xyz Friends who are interested can also go up and play.

The overall experience is that you will get a normal transfer transaction, and then you can provide services to users. Like my digital goods sale above, you need to create an additional listener program to process the transaction, and then use your wallet to re-initiate a request to transfer the fee to the publisher. So it seems that it is more suitable to provide some non-contract services at present, such as buying a membership.

Another project involves contracts. Basically, you determine the payment status by verifying the passed header, and then use your wallet to call your contract. But this actually makes decentralization more centralized. In the past, users could directly call the contract, but now if they call it through your interface, the risk will actually be greater.

Another very important experience that I haven't verified yet is the payment during the MCP process. For example, if you call a certain MCP to create a picture, the traditional method is that we go to the corresponding official to pre-recharge, and then call the consumption. What can be done now is that only one signature can be used to complete the call each time, which greatly reduces the waste of funds, but it is not a good thing for the platform, so I currently feel that this advancement is difficult.

Because of this agreement, I have recently paid more attention to the entire base ecology. For example, farcaster, which used to be a decentralized social platform, has become like a decentralized WeChat. It currently has mini apps, similar to small programs. But no additional development is required. As long as your webpage writes a json statement, it can be opened in mini apps and can be used directly without review. So I integrated the x402 project that I had previously done with the contract, but the disadvantage is also obvious, that is, if you can't be on the list, people won't be able to find your application. At present, I only looked at the top 100 applications on the list.

You can see that my recent focus is on web3, but you see that the emergence of protocols like x402 is actually to open up the current new model of AI+Web3. In addition, the main AI really does not make money. Some of the independent developers I followed before are now engaged in self-media, posting fake content, and selling courses for 1,599. Only a very small number of them are making money by constantly iterating their own software. There is no way to fake and sell courses. I agree with traffic management, but by creating topic confrontation and black and white to get traffic, having traffic is right. Plagiarism and transfer are all reasonable, as long as there is income. To be honest, if I could do it I would have done it long ago.

I don’t have anything I particularly want to do at the moment. I’ve tried a lot this year, but none of them have shown much success. Of course, in the eyes of some people, I may just be someone who occasionally looks for a job and does nothing all day long.