How Do I Turn My Crypto Strategy Into a Self-Improving Agent?

Why can't I just keep manually tweaking my crypto strategy?

Manual strategy tweaking is slow, inconsistent, and prone to changing multiple variables at once. When you adjust your stop-loss and your position size and your entry signal in the same session, you have no idea which change drove the result. The Lewis Jackson framework solves this by enforcing the scientific method: Hermes changes exactly one variable per weekly cycle, observes the outcome, and only promotes improvements that move toward your defined goal.

For crypto traders with existing strategies, the biggest leverage is in Phase 2, path A. You point the onboarding agent at your existing strategy file — your entry/exit rules, position limits, slippage tolerance — and it extracts everything into a structured format. Your trade history (wins and losses) becomes a Hermes-readable ledger that gives the self-improvement loop a running start instead of cold-starting from zero data.

How does the framework handle my existing trade history?

During Phase 3, the agent converts your historical trades into a structured ledger that Hermes can parse, score, and learn from. If you have 50 trades over 6 weeks, Hermes can immediately identify patterns — like whether your losing trades had systematically larger position sizes than your winners. This is the kind of insight that takes humans weeks to discover but Hermes surfaces in its first read-only cycle.

The key requirement is that your trade history must be honest and complete. Cherry-picking only winning trades corrupts the data and violates the Accuracy principle. Feed in everything — the Hermes agent needs the full picture to form valid hypotheses.

What success and failure definitions work for experienced crypto traders?

Your success definition should be specific and realistic given your capital and asset volatility. Examples that work well: '15% monthly return with Sharpe ≥ 1.2 and max drawdown ≤ 12%' or 'outperform buy-and-hold by 5% per month with drawdown under 10%.' Your failure definition is equally important: 'two consecutive months below 5% return' or 'drawdown exceeding 15% in any 30-day window.'

The framework will flag success definitions that are mathematically impossible relative to your starting capital. If you're trading with $5,000, targeting $100,000/month isn't a stretch goal — it's an error that wastes improvement cycles.

How do I transition from the read-only first cycle to live trading?

After Hermes completes its first read-only cycle, review the markdown analysis it produces. Verify it correctly identifies your strategy parameters, understands your success and failure definitions, and has accurately parsed your trade ledger. If everything checks out, edit the Hermes trading strategy YAML to enable live write mode.

Do not skip this step. The read-only cycle exists specifically because experienced traders often have nuanced strategy logic that could be misinterpreted. A misconfigured Hermes loop making live modifications to your strategy with real capital is the most expensive mistake in the framework.

What should I do after the agent is running?

Monitor improvement cycles weekly using the check-in commands from the final configuration summary. Each cycle will show: the variable Hermes changed, the hypothesis behind the change, and whether the result moved toward or away from your goal. Approve strategy promotions when cycles improve performance. Over multiple weeks, you'll see compounding improvements — each small, verified gain building on the last.

The next step is to get the latest oneshot prompt from the 01 Systems community and paste it into Claude Code to begin your setup.