Frequently Asked Questions About Lewis Jackson Self-Improving Trading Agent Framework
23 answers covering everything from basics to advanced usage.
// Basics
What is a well-defined goal in the context of the self-improving trading agent?
A well-defined goal includes both a specific success definition (e.g., 15% monthly return, minimum Sharpe score of 1.2) and a specific failure definition (e.g., drawdown exceeding 12% in any 30-day window). The agent uses this polarity — toward-goal vs. away-from-goal — to orient every improvement cycle. Without both poles defined precisely, the self-improvement loop has no direction and 90% of trading agents end up flying blind.
What is a Hermes-readable ledger?
A Hermes-readable ledger is a structured file of all historical trades — both wins and losses — converted into a format the Hermes agent can parse, score, and learn from. It's generated automatically during Phase 3 scaffolding. If you have existing trade history, the framework converts it into this structured format so Hermes can analyze patterns from day one rather than starting with zero data.
What is a read-only cycle in the Hermes trading agent?
The first Hermes review cycle is read-only: Hermes observes, scores, and produces a markdown analysis of your strategy but does not write any changes to the live strategy. This exists to validate that Hermes has correctly understood your strategy and goals before it begins making modifications. You must manually approve the transition to live write mode by editing the Hermes trading strategy YAML after reviewing this first cycle's output.
Is the Lewis Jackson trading agent framework free to use?
The oneshot prompt is available for free through Lewis Jackson's 01 Systems community. Claude Code requires an Anthropic API subscription. Railway is free up to a significant usage threshold, after which paid tiers apply. There are no licensing fees for the framework itself or for Hermes. Your primary costs are the Claude Code API usage, any Railway overages beyond the free tier, and the trading capital you allocate to the agent.
What is the 01 Systems community?
01 Systems is Lewis Jackson's free community resource where all oneshot prompts from his YouTube videos are stored, versioned, and updated based on community feedback. It's the canonical source for the latest prompt versions — always retrieve prompts from here rather than transcribing from video recordings. The community also provides support, shared learnings, and feedback that drives prompt improvement over time.
// How To
How do I authenticate Railway for 24/7 agent hosting?
When the onboarding agent prompts Railway authentication, it attempts a Railway CLI login. If the interactive login fails inside the Claude Code session, open a split terminal, paste the provided login command, complete browser authentication, then return and type 'done, continuing.' If you don't have a Railway account, create one during this step — it's free up to a significant usage threshold. This ensures your agent runs 24/7 regardless of your local machine's state.
How do I define a failure threshold for my trading agent?
A failure definition must be specific and measurable — not vague like 'losing money.' Examples: 'drawdown exceeding 15% in any 30-day window,' 'two consecutive months below 5% return,' or 'Sharpe score dropping below 0.8 over a 60-day rolling window.' The agent uses this failure boundary as the negative pole in its improvement orientation. Without it, the self-improvement loop cannot determine when a strategy iteration has moved in the wrong direction.
How do I get the latest oneshot prompt for the trading agent?
Always retrieve the latest oneshot prompt from Lewis Jackson's 01 Systems community, specifically under Classroom > YouTube Video Prompts. Do not manually transcribe prompts from video recordings — those become outdated as the prompt is iterated based on community feedback. The latest version supersedes any version shown in a video. The 01 Systems community is free to join.
How do I review and approve a Hermes improvement cycle?
Use the check-in commands provided in the final configuration summary to inspect cycle outputs. Hermes produces scored hypotheses, identifies the single variable it changed, and reports whether the result moved toward or away from your goal. If the cycle moved toward your success definition, approve it as the new baseline. If it moved away, the insight still informs the next hypothesis. Track directional progress over multiple cycles rather than judging any single cycle in isolation.
What if I don't have an existing trading strategy to start with?
In Phase 2, select path B — tell the agent you have no strategy and want a basic one scaffolded. The onboarding agent creates a solid baseline strategy appropriate for your chosen asset. Success and failure definitions are set conservatively given your capital size. Hermes then begins the self-improvement loop from this baseline, changing one variable per weekly cycle to compound improvements toward your defined goal over time.
// Troubleshooting
Why is my Railway login failing during setup?
Interactive Railway CLI login can fail inside the Claude Code terminal session due to browser redirect issues. The fix is to open a split terminal (a second terminal window), paste the Railway login command provided by the agent, complete the browser-based authentication in that split terminal, then return to your Claude Code session and type 'done, continuing.' This is a known friction point documented in the workflow, not a system failure.
What if Hermes misinterprets my strategy in the first read-only cycle?
This is exactly why the first cycle is read-only. Review the markdown analysis Hermes produces and verify it accurately reflects your strategy parameters, asset, success definition, failure thresholds, and score weights. If anything is misinterpreted, correct the strategy document and goals document before approving the transition to live mode. Never flip to live trading before this review — the read-only cycle is your safety valve against misconfigured loops managing real capital.
What if the agent flags my success definition as unrealistic?
The agent sanity-checks your success definition against your starting capital. If your target is mathematically impossible — like targeting $1M/month on $10 starting capital — it flags this before proceeding. Revise your success definition to realistic bounds. A useful rule: define success as a percentage return rather than an absolute dollar amount, and research historical Sharpe ratios for your chosen asset class to set achievable thresholds.
What happens if my trading agent goes offline?
If you hosted locally instead of on Railway, your agent stops executing when your machine shuts down or loses network. This violates the Reliability principle and breaks the continuous improvement loop. The fix is to deploy on Railway — the framework is designed around 24/7 cloud hosting. If your Railway deployment goes down, check your Railway dashboard for deployment errors and redeploy from your latest strategy state.
// Comparisons
How does the Lewis Jackson trading agent compare to using a static trading bot?
A static trading bot executes predefined rules without learning. When market conditions change, performance degrades until a human manually updates parameters. The Lewis Jackson framework adds a closed-loop self-improvement cycle: the agent analyzes its own trade outcomes weekly, identifies whether results moved toward or away from a defined goal, and iterates one variable at a time. Over weeks, this compounds into strategy evolution that a static bot can never achieve on its own.
How does the self-improving trading agent compare to manual backtesting and optimization?
Manual backtesting requires you to form hypotheses, change parameters, run tests, and interpret results — all by hand. The Lewis Jackson framework automates this entire loop: Hermes forms hypotheses, changes one variable, observes live (not backtested) outcomes, and promotes winning versions. It also enforces the scientific method discipline that most manual optimizers break by changing multiple variables simultaneously. The tradeoff is that the automated loop operates on live capital, making the read-only first cycle critical.
Can I use this framework for forex or stocks, not just crypto?
The framework is asset-agnostic in design — the target asset is a required input, and you can specify any tradeable instrument: Bitcoin, Ethereum, forex pairs, stocks, or other assets. The key constraint is that you need API access to execute trades on your chosen market. The oneshot prompt and Hermes self-improvement loop work identically regardless of asset class. However, most community examples and the creator's demonstrations focus on cryptocurrency markets.
// Advanced
Can I run multiple self-improving agents on different assets simultaneously?
Yes, you can deploy multiple agents, each targeting a different asset with its own success definition, failure definition, and Hermes improvement loop. Each agent should run as a separate Railway deployment with its own strategy document, goals document, and trade ledger. The critical rule still applies to each: one variable change per cycle per agent. Do not have agents share strategy parameters across assets — each self-improvement loop must be independent to maintain clean learning signals.
What is the weekly cadence and why does it matter?
The weekly cadence is Hermes's default review and improvement cycle frequency — once per week. This interval allows enough trades to accumulate for statistically meaningful analysis. If a secondary agent like Cornelius is also involved, Hermes offsets its review by 3 days to prevent simultaneous conflicting parameter updates. Changing the cadence too frequently (e.g., daily) risks making decisions on insufficient data; too infrequently (e.g., monthly) slows the compounding improvement effect.
How do score weights work in the Hermes trading agent?
Score weights are configurable parameters in the strategy document that determine how different trade outcomes are weighted relative to your defined goal. For example, you might weight risk-adjusted return (Sharpe) more heavily than raw return. Hermes owns and adjusts these score weights as part of its self-improvement process. During the read-only first cycle, verify that the initial score weights align with your priorities before approving live mode.
What happens if a Hermes cycle makes the strategy worse?
If a cycle's single-variable change moves the strategy away from your goal, that version does not become the new baseline. Instead, the previous version remains the baseline, and the negative result becomes data that informs Hermes's next hypothesis. This is the scientific method in action — failed experiments are not wasted; they narrow the search space. Over multiple cycles, both positive and negative results contribute to the agent converging on better-performing strategies.
Can I manually override a Hermes improvement suggestion?
Yes. You can review any Hermes cycle output and reject its proposed variable change before it's applied. The framework is designed with human-in-the-loop approval, especially during the initial read-only cycle. For ongoing cycles, the check-in commands let you inspect proposals before they're promoted to the new baseline. However, frequent manual overrides can disrupt the systematic improvement process — use them for genuine misalignment, not personal preference.
Why does the framework use Claude Code specifically?
The oneshot prompt is architected for Claude Code's terminal-based interaction model. Claude Code can detect your OS, manage file creation, orchestrate multi-phase guided flows, and handle Railway CLI authentication — all from a single pasted prompt. The framework's zero-friction setup depends on these capabilities. While the underlying principles (scientific method loop, well-defined goals, single-variable testing) are transferable, the specific oneshot prompt automation is built for Claude Code.