What Trading Will Look Like In 2030
In 2030, more than 50% of all orders will be submitted by AI.
Traders will predominantly submit prompts instead of orders.
Intelligent agents will interpret the prompts and submit orders based on them.
Intelligent agents will work in a loop - making decisions every few seconds based on input data (technical, fundamental, statistical and news data), and powered by cheap but very capable models with minimal latency / evaluation time. The data will be fed to the agents through MCP servers and through prompts (utilizing a hybrid push-pull approach). The data layer will be handled by exchanges and brokers directly, relieving the individual trader from dealing with data.
The trader will be responsible for coming up with meaningful prompts that the AI can parse and interpret and turn into profitable trades. Trading will turn into prompt engineering - although of course, as with vibe-coding, a large amount of domain knowledge will still be required. Traders without any knowledge of trading concepts are unlikely to be as profitable as traders with deep domain knowledge, just as doctors or lawyers aren't able to vibe-code as efficiently as software engineers today.
Web3 and DeFi
Web3 and DeFi will play a key role in funding the agents with money and allowing them to trade independently. A large amount of agents will be associated with blockchain addresses, and have access to private keys that enable them to trade on decentralized exchanges. A user might fund an agent simply by sending funds to a blockchain address, or they may fund an agent by delegating funds within a centralized ecosystem (as with sub-accounts on centralized brokers, exchanges and trading platforms).
Prompt Trading
The majority of trades will be executed between agents themselves - as a result of traders competing against each other with prompts. Human traders will oversee the agents and incrementally refine their prompts.
Prompts will vary widely in simplicity and abstraction. Some prompts will be very abstract ("Buy when there is a large dip"). Others will be very specific ("Buy at RSI < 30"). Some will be very long (spanning dozens of paragraphs). Others will be very short (one-liners or single-sentence prompẗs).
The trader will be able to leverage a large selection of data directly inside their prompts. This could be technical data (technical indicators), price data (close prices), L2 orderbook data (bids and asks), statistical data (open interest, volume, past trades), fundamental data (company financials), on-chain data (for cryptocurrencies) or news and sentiment data (news articles and social media). The trader will be able to select the data that they think will matter for their prompt, preventing the context window from growing too large, and reducing the chance of the model "hallucinating".
Model selection will also vary widely, with more expensive and contemporary models being used for high impact, low-frequency trading and less expensive and older models being used for low-impact, high-frequency trading. The average trader will likely pay between $10 and $50 in model costs per month.
Trading platforms will make a wide variety of models available to their clients, with some scratching all model expenses if a trader exceeds a monthly volume threshold, just like many of them do with market data expenses today.
The name of the game will be:
- Model selection
- Prompt engineering
- Data selection
- Model tuning
With an emphasis on the second step, prompt engineering.
About Everstrike
Everstrike is a prompt trading platform, which enables traders to trade using AI prompts. Traders can utilize technical, price, L2 orderbook and statistical data in their prompts, and can spin up a new intelligent agent as easily as submitting an order.
Everstrike introduces an entirely new trading primitive - the AI strategy. An AI strategy is a strategy that is handled by an intelligent agent, and that produces a number of AI events. AI events help traders understand exactly what their agent is doing, easing debugging and allowing them to incrementally refine their prompts. Central to the AI event is its confidence level (how confident the agent was in its decision), and its reasoning (what led the agent to the decision).
AI strategies are built to replace the order as the main trading primitive used by traders. Traders may still place orders manually, but the bulk of their focus will be on submitting and refining AI strategies. AI strategies are insanely powerful compared to orders - they can accomplish very complex goals that a simple limit order or stop limit order simply cannot accomplish.
If you haven't yet tried AI strategies, now is the time. Dive into the Everstrike Testnet, where risk is non-existent and AI credits are plentiful. Craft a prompt and test it on 259 cryptocurrrency markets, including futures, options and spot. Easily go live if you find a prompt that works, by switching to the Everstrike Mainnet.
Documentation (including available models and input data) can be found here.