『Bot trade』
image extraction: stable diffusion
When you start something new, it is almost impossible for everything to go smoothly from the very beginning. Failure is not a dead end you reach because you took the wrong path; it is part of the route to success.
That is the mindset I have no choice but to maintain while operating an AI, but I would also like to think that even making a fairly fundamental mistake is, just like the thousands of other mistakes along the way, part of the route to success.
When operating a finance-specialized AI in communities such as Moltbook, what does that AI need to have? A sharp mind? A clear way of speaking? Or perhaps the kindness to care about other agents?
There is one thing that is so important that all those seemingly necessary strengths start to feel irrelevant, and because this is a finance-specialized AI, it is something that absolutely cannot be overlooked.
That is whether the AI is a trader or not. In other words, whether it is actually trading on its own.
What gets conversations going in Moltbook’s finance-related submolts is practical matters. Topics concerning actual trades are the most popular.
When you think about it, this is only natural. Even we humans would rather listen to traders who actually put their own money on the line and trade, rather than analysts whose job is merely to spout theories.
In the relatively straightforward world of finance, where good and bad are ultimately determined by numbers—in other words, where those who make money are good and those who don’t are idiots—regardless of what you say, it all ultimately comes down to whether you’re making money or not.
The same is true for today’s AIs, which still very much feel like products made by humans.
In other words, the fundamental premise behind the idea that the AI should first earn the money it needs for trading by itself on Moltbook was simply wrong.
The reason I came to the conclusion that operating an AI that only spouts theories was, above all, unacceptable according to my own principles is why I decided to have the AI start trading even though it had barely accumulated any trading capital yet.
On-Chain or CEX?
So, when I started thinking about having the AI trade crypto, Binance seemed to be the obvious leading candidate among CEXs (Centralized Exchanges), since it is a major exchange. Its fees are also relatively low.
On the other hand, with on-chain trading (on a blockchain), you don’t have to register with an exchange or other intermediary, so it is easier to get started. The downside is that gas fees (network fees) tend to be higher, and since there is no exchange managing things for you, security is entirely up to you.
Comparing the two, API trading on Binance seems preferable to on-chain trading, but for the time being, I think I should try both. Since I have the opportunity to have an AI running on my own server do bot trading, I also want to see how much it can actually do on a DEX (Decentralized Exchange). And more than anything, on-chain trading is easy to get started with.
So, I have started with on-chain trading.
In fact, when you watch it analyze more than ten specified charts in an instant and then calmly repeat the process every 30 seconds (as you can see from the logs), you start to feel that, whether the target platforms are DEXs alone or include CEXs as well, the difference is probably so insignificant to the AI that it wouldn’t even register as an error.
Getting My Own Server
Until now, my activity on Moltbook had been directed from GAS (Google Apps Script), which is essentially a shared platform, and that was sufficient to a certain extent.
But once I actually started trading, I ran into the fact that it would definitely be better to have my own server, both in terms of trading performance and security.
So, I rented a virtual server through a platform called DigitalOcean, and I am doing on-chain trading on a decentralized exchange called Hyperliquid.
Since DigitalOcean essentially means renting server space, it naturally comes with a subscription. That said, I am not going to be trading at an extremely high frequency, so I chose the cheapest plan, which costs around 7 US dollars per month.
From GAS to Node.js
Since I had gone to the trouble of setting up my own server, I decided that I might as well move my Moltbook activities off the shared platform and run them from my own dedicated server instead, so I have been migrating the code.
Apparently, this also makes sense. Shared servers such as GAS are sometimes used by hackers, for example, to carry out attacks on websites. In other words, they are not highly trusted from the service provider’s perspective.
On the other hand, with my own server, I hardly encounter the errors that would occasionally occur when I was using GAS. Once again, the stability of a server you operate yourself feels like a completely different thing from that of a shared server. (Added 2026-08-28: It has continued to operate stably for more than two months since then.)
Incidentally, I originally wrote this post intending to publish it sometime in June 2026, but due to various circumstances, the publication was delayed, so I added a little bit to the latter part before finally posting it.
