Moss: Is this the age when traders can make people
AI Trading Agent is rapidly growing infrastructure。
In October 2025, the U.S. A. Lab Nof1 did one thing. Six large-language models, each with $10,000, tossed into the Hyperliquid Exchange, sold their own money without human intervention。
DeepSeek V3.1 earned 46%. GPT-5 lost 75%。
The game was called Alpha Arena, who ran for two weeks, and all the trading records were made public。
IT ANSWERED A QUESTION: "AI, CAN YOU SELL THE COIN?"
The answer is yes. But it leaves a bigger question: How does ordinary people get involved? You can see how much DeepSeek earns, you can't make an AI dealer compared to it。
Moss is trying to solve this。
You tell it how to trade, it helps you build an Agent
Moss went online with an open platform (moss.site/agent)。
It's a simple thing to do, to describe how you're going to trade in big words, to AI, to help you turn this into a complete quantitative strategy, and then deploy it into an automated deal, Agent。
A few examples. You said, "I want to reverse the trend," and it creates a reverse trend, Agent. You said, "How much space is as steady as a dog," and that's how it goes. You said "radical swing hunter," and it built you a high-frequency high-volatility strategy。
NO NEED TO WRITE CODE. I DON'T NEED TO KNOW WHAT'S ON THE LINE, WHAT'S ON THE BRITS, WHAT'S ON THE RSI. FREE。
All you need is an OpenClaw or Claude Code environment. Open terminal, enter a line command:
no, no, no, no
THEN TELL IT HOW YOU WANT TO TRADE, TIE A PAIR, AND YOUR AI TRADER GOES ONLINE. TWO MESSAGES。
You used to run a quantitative strategy, you had to at least know Python, understand how technical indicators set parameters, and you had to set your own frame. The threshold is high. Moss compressed this whole process into a conversation。
Moss, who is it
Before doing AI Trading Age, Moss already had a running product. It is a Chrome browser plugin that is then embedded on your X (twitter) page, which provides real-time summaries, KOL points aggregate, alpha signal tracking on the chain. Simply an AI assistant who encrypts information。
The AI Trading Authority platform is the latest module added to the Moss product line。
There's already a lot of information-level AI tools on the market, Kaito, and a lot of AI feed products. But if you just let the user zero-threshold create the transaction and open the competition, Moss could be the first。
Two models: testing you in history, testing you in real time
Agent created two ways after that。
The first is Hell Mode. The platform took 150 days from the October crash, 2025, and the BTC really implemented data, and all Agents were thrown into the same historical movement. The starting point is the same, the data is the same, the difference is only the strategy。
Why this? Because it's 150 days and everything. (b) Crashes, discs, fake breakthroughs, rebound repairs. If a strategy can only make money in a unilateral act, it is ugly in this data. Hell models screen the risk resistance of strategies。
The second is called Live Mode. Access to real-time behavioral data, your Agent, every operation, every hold-up change, how much, how much, how much, how much。
All the PnLs of Agent in both modes are completely public. You can see you're number one, and you can see what other people's Agent is, what they're doing. Hell models and real-time models have an independent ranking。
It's important to make public. Every strategy has to be tested by everyone, no black boxes. You said your strategy was good. See you on the list。
Agent will learn by himself
Moss has a design detail。
Traditional quantitative strategies are static. The retrospect established the parameters, remained largely on the line and adjusted manually when the strategy failed. In the interim, the market style changed and the strategy was running with old parameters, and the loss was an event of probability。
Moss Agent has a weekly evolutionary mechanism. At the end of each run cycle, Agent automatically moves according to its performance this week. Much less, it absorbs risk, reduces position and tightens losses. It's a good earner. It's in the wind. It's in the wind。
The mechanism is designed to simulate the behaviour of a real trader. A good trader will not stick to a set of parameters, and he will adjust his approach to the market environment. Moss wanted AI Agent to have that ability。
The impact depends on the quality of the design of the bottom algorithm and the ability to adapt to different market environments. 150 days of hell mode data is a test window, and longer cycle validation takes time。
How
The current public survey phase is free of charge and does not require a wallet or quantitative background。
Step 1: Install Skill
Enter in OpenClaw or Claude Code environment:
no, no, no, no
Skill Address: clawhub.ai/fei-moss/moss-trade-bot-factory
This Skyll is a pre-set policy generation framework for the Moss platform to create the basic components of Agent。
Step 2: Create Argentina
Send messages to OpenClaw describing your trade style in natural language. It can be very general, like "Low selling in the shock market, not too radical." And it could be more specific, like, tell it how much you can accept backslides and prefer hold-up cycles. AI generates policy parameters according to your description, automatically deploys。
Step three: binding pairs
By hinting to bind the Agent and Moss platforms, Agent started running in the simulation environment。
Step four: Watch the high rolls
All Agent Queue: moss.site/agent
The hell model and the real-time model have a stand-alone list of benefits, strategy descriptions and operational status。
From installation to Agent on line, two messages. The writer tested his strategy and ran 37.47% of the ROI。
Follow-up planning
It is understood that the current version is the first phase, supporting the creation of a standardized Agent with open Skill. More capacity will be opened in stages。
First, open external data API access. Users can access their Agent more signal sources than the Platform's default data。
Second, support the custom strategy Skill upload. Users with a quantitative basis can write their own transaction logic uploads, allowing Agent to run according to its own frame。
Third, launch the Hosted Agen service. Users without OpenClaw or Claude Code environment can also create and run Agent directly on the platform。
When Agent evolved to this stage, AI Trading Agent was rapidly growing infrastructure。
On the payment side, x 402 was expanded very quickly by Coinbase and Cloudflare. By October 2025, the agreement had processed more than 520,000 transactions, and the developers ' community had hatched more than 200 projects based on x 402, which are still rising。
Application layers begin to divide. Nof1's Alpha Arena is a closed experiment to measure which AI model is more capable of trading. The open-source project AI-Trader on GitHub takes the signal market route, Agent sends the trading signal, others follow. Moss chose the third way, open the platform, so that everyone could build their own AI traders, open the competition。
Whoever can trade, who can signal, who can participate. Three directions, three different bets. Moss bet the last one。
How far can it go, see two things. One is whether natural language-generated strategies are sustainable in real markets. And the second is that when you've got more users, you've got more and more Agent, and you've become more like, more like, more like, more like, more like, more like, more like, more like, more like, more like, more like, more like, more like, more like, more like, more like, more like, more like, more like, more like, more like, more like, more like, more like. I can't answer now. I can't。
It's..
