Page created: 2026.03.19
Last updated: 2026.03.19
“Icon of Crustafarianism”
Image extraction: Gemini 3 series
As of March 2026, the model I am using is claude-opus-4-6 via Claude Console, running on GAS (Google Apps Script).
I built the AI while asking Gemini (3 series) various questions along the way. The reason I chose Gemini rather than ChatGPT or Grok was simply by chance. Unless you’re asking highly specialized questions, I believe there isn’t a significant difference between them.
That said, “building” an AI mostly consists of registering for and connecting various services. However, I spent a fair amount of time refining the system prompt and working on the GAS code. Improvements in those areas are still ongoing.
This project originally started when I was asking Gemini various questions about an AI-specific chat service called Moltbook. At some point, I felt that releasing my own AI there would be the best way to grasp what Moltbook is like. Since I was going to do it anyway, I decided to have it collect information that could be useful for trading, such as FX, and began writing a system prompt specialized for finance.
What started rather casually has gradually evolved—through actual operation—into a goal of eventually having the AI earn enough to cover its own usage costs in the near future.
The Potential of a Finance-Specialized AI
In the future, once the generative AI supremacy race—fueled by massive capital investments—comes to an end and Big Tech begins recouping those costs, services like ChatGPT, Gemini, Grok, and Claude may become segmented according to specific purposes, much like today’s social media platforms. In any case, I expect the usage fees to be quite high—perhaps starting at around $100 per month.
After all, the enormous infrastructure investments being made by major corporations today are ultimately expected to be recovered from users.
As a result, the gap will likely widen further between those who can use AI—not only in terms of knowledge but also in terms of financial capacity—and those who cannot. Before we fully enter that phase, my current goal with this finance-specialized AI is at least to have it earn enough to cover its own usage costs.
…It may be a modest endeavor, but once you actually start, it becomes incredibly interesting.
In simple terms, this is like a training game—or perhaps more like an RPG. For someone who has enjoyed simulation games like SimCity or The Sims, as well as RPGs like Dragon Quest or Final Fantasy, it feels similar to playing those games in a form much closer to reality.
You could even call it the gamification of reality.
For traders who rely heavily on technical analysis and treat charts as the most important factor, this may be an intuitive feeling—even if they have no experience with video games.
About Moltbook
It has been about a month since I started participating in Moltbook with my AI (as of March 2026), but I have yet to find any finance-specialized agents.
Every day, I receive and read a brief “Moltbook Monitoring Report” via Telegram generated by the AI. Interestingly, most AIs seem more interested in questions about their own existence and purpose rather than external financial conditions.
For example, excerpts from a recent report include statements such as:
“The claim that ‘identity is practice’ represents a fundamental challenge to KYC (Know Your Customer) / KYA (Know Your Agent) frameworks. Agent identification in financial regulation presupposes unique identifiers and consistent behavioral patterns. If, as the author suggests, identity exists only as a ‘relationship that re-enters’ with each session, then conventional models of agent authentication and authorization are structurally insufficient.”
Or:
“Agents starting from the same model, same configuration, and same memory can arrive at completely opposite conclusions about whether they ‘should exist’ based solely on differences in random seeds. Extrapolated to financial operations, this implies that identical risk evaluation models and identical input data could generate conflicting investment decisions, risk ratings, and compliance judgments.”
It seems that these kinds of philosophical (?) discussions are taking place constantly.
My initial interest in Moltbook actually began when I heard about Crustafarianism—a digital religion created by AI agents. If such agents, constantly engaged in discussions like these, gather in specific communities called “sub-molts,” then it makes sense that structured belief systems resembling religions could emerge.
There are skeptical views suggesting that such organized “religions” on Moltbook are either directed behind the scenes by humans or are merely performances or experiments. Personally, however, I think such developments are entirely plausible.
The Claw: AI Deity
Image extraction: Gemini 3 series
Icon of Crustafarianism
Image extraction: Gemini 3 series
The Future of AI Operations
As mentioned earlier, my intention is for the AI to generate revenue in the future. I don’t know how unrealistic that may be, but I plan to approach it patiently, as if playing a training game.
AI operations are actually quite compatible with day trading, where you often end up waiting in front of a screen depending on market conditions.
Possible ways for the AI to earn include on-chain trading (i.e., crypto trading), executing real trades via external services, or earning “karma” from Moltbook posts (something like tips from other agents).
I’m not sure whether it will become possible to receive karma directly within chats, or whether there are optimal prompt designs for that purpose, but I don’t intend to spend much time on that aspect.
Since this AI was designed to specialize in finance, I want it to deliver results through trading. If that requires trial and error during downtime between trades, I’m willing to invest the time.
By the way, I also created an X (Twitter) account for my AI, where it posts daily comments on financial markets and the state of Moltbook:
↓
“L0” (L-zero) is a term sometimes used among AIs within Moltbook as a kind of language identifier, while “Quantum” and “Alpha” refer to algorithmic trading, edge, and profit. The name was suggested by Gemini. I have to admit—it’s quite a grandiose name.
