How Do Beginners Build a Self-Improving AI Trading Agent?
For Complete beginners who want to start algorithmic trading with AI · Based on Lewis Jackson Self-Improving Trading Agent Framework
// TL;DR
The Lewis Jackson Self-Improving Trading Agent Framework lets complete beginners deploy an autonomous trading agent without an existing strategy or deep coding knowledge. The oneshot prompt scaffolds a baseline strategy for your chosen asset, sets conservative goals given your starting capital, and deploys everything to Railway for 24/7 operation. Hermes then improves the strategy one variable at a time on a weekly cycle. You need a terminal, a Railway account, Claude Code access, and realistic expectations about returns relative to your capital.
Can I build a self-improving trading agent with no coding experience?
Yes. The Lewis Jackson framework uses a oneshot prompt architecture — you paste a single prompt into Claude Code and it handles environment detection, strategy scaffolding, file creation, cloud deployment, and Hermes installation automatically. You interact through guided prompts, not raw code. You should be comfortable using a terminal (command line) and following step-by-step instructions, but you don't need to write Python, JavaScript, or any programming language.
The critical skill isn't coding — it's defining clear success and failure definitions. 'I want to make money' is not a goal the agent can optimize toward. '10% monthly return with max drawdown under 8%' is. The quality of your goal definition determines the quality of every improvement cycle that follows.
What if I don't have a trading strategy to start with?
In Phase 2 of the setup, select path B — tell the agent you have no existing strategy. The onboarding agent will scaffold a basic baseline strategy appropriate for your chosen asset (e.g., Bitcoin, Ethereum). This baseline isn't random — it's a solid starting point with defined entry/exit rules, position limits, slippage tolerance, and conservative parameters.
Hermes then begins the self-improvement loop from this baseline. Each weekly cycle tests one variable change: maybe a tighter stop-loss in week one, a different entry signal in week two. Over time, these verified single-variable improvements compound into a strategy that's genuinely adapted to your asset and market conditions.
The advantage of starting from a scaffolded baseline is that Hermes has a clean, well-defined starting point. The disadvantage is that it has no historical trade data to learn from initially — the first few cycles will be exploratory.
How much starting capital do I actually need?
There's no minimum enforced by the framework, but the agent will sanity-check your success definition against your capital. Targeting 100x returns on $50 will be flagged as mathematically impossible. A more realistic approach for small capital: define success as a modest percentage return (e.g., 5-8% monthly) with tight drawdown limits (e.g., max 10% drawdown).
Remember that trading involves real financial risk. The self-improving loop doesn't eliminate risk — it systematically improves your strategy's risk-adjusted performance over time. Start with capital you can afford to lose while the agent's first improvement cycles are still exploratory.
What does the setup process actually look like for a beginner?
Here's the concrete flow:
1. Get the oneshot prompt from the 01 Systems community (free to join) — always use the latest version, not one transcribed from a video.
2. Open Claude Code in your terminal and paste the full prompt.
3. Phase 1: The agent detects your OS and available runtimes. Confirm when prompted.
4. Phase 2: Choose path B (no existing strategy). The agent scaffolds a baseline. Set conservative, measurable success and failure definitions.
5. Phase 3: The agent creates all files and the Hermes-readable ledger.
6. Phase 4: Connect the necessary trading APIs.
7. Railway authentication: Log into Railway for 24/7 cloud hosting.
8. Hermes installation: The agent installs Hermes automatically.
9. Review: Check the final configuration summary. The first Hermes cycle is read-only.
The entire process is guided — you're never left staring at a blank terminal wondering what to do next.
What mistakes do beginners make most often?
The most common beginner mistakes are: setting unrealistic success targets relative to starting capital, skipping or vaguely defining the failure threshold ('I don't want to lose money' is not a failure definition), and flipping to live trading mode before reviewing the first read-only Hermes cycle. All three are avoidable by following the framework's principles before touching any tooling.
Your next step: join the 01 Systems community, retrieve the latest oneshot prompt, and start by writing down your success definition, failure definition, target asset, and starting capital before you open a terminal.
// FREQUENTLY ASKED QUESTIONS
Do I need to understand the Sharpe ratio to use this framework?
A basic understanding helps. The Sharpe ratio measures risk-adjusted return — a higher Sharpe means better returns relative to the risk taken. The framework uses it as a quantitative success/failure threshold. You don't need to calculate it manually; Hermes handles that. But knowing that a Sharpe of 1.0+ is generally considered acceptable and 2.0+ is strong helps you set realistic goals.
What's the cheapest way to get started with this trading agent?
The oneshot prompt and 01 Systems community are free. Railway has a free tier with generous usage limits. Your main cost is Claude Code API access (Anthropic subscription). Start with the minimum capital you're comfortable risking while the agent's early improvement cycles are exploratory. Avoid scaling capital until you've reviewed several Hermes cycles and confirmed the agent correctly understands your strategy.
Can the self-improving agent lose all my money?
Yes, trading always involves risk of capital loss. The framework mitigates this through defined failure thresholds (e.g., max drawdown limits) and the read-only first cycle that lets you verify Hermes's understanding before live modifications begin. However, no automated system eliminates market risk. Start with capital you can afford to lose and set conservative drawdown limits as a hard stop.