Deep Dive
Setting Up GPT-6 Astra with TradingView
Miles walks through the technical setup required to connect GPT-6 Astra to TradingView. You need three components: the TradingView desktop app from tradingview.com/desktop, an MCP bridge installed from GitHub (with a full setup guide available free in the Miles Finance School community), and either the MCP method for speed or a local connector using GPT's computer use feature. Miles emphasizes that for trading, the MCP method is faster and more efficient because you want chart updates instantaneously when voice prompting. The alternative local connector works but introduces lag since GPT has to click around your computer. He notes that the magic of this integration is automation—you're not just manually adding indicators yourself, you're having AI execute commands at scale.
Building Support/Resistance and Custom Indicators
Miles demonstrates how GPT Astra can analyze Bitcoin charts and identify the three most defensible support and resistance zones based on repeated price reactions at swing points. When tested on a live chart, it accurately identified the key zone at $76–$76.7k with evidence from April and May daily reactions, plus specific invalidation rules. He then layers in visible range volume profile (VRVP) to challenge the initial analysis, showing how point of control at $64,000 and value area high align with structural pivots. The tool can create custom Pine Script indicators on demand—he builds a market context dashboard that displays volatility, yearly high, reference high, daily close, 200-day direction, and Bollinger Bands in one view. This gives traders a daily snapshot in seconds instead of manually assembling data themselves, and you can ask GPT to modify or expand indicators based on your exact trading biases.
Backtesting Strategies and Finding Profitable Setups
Miles shows how to ask GPT Astra to research distinct setups, define measurable entry triggers and invalidation rules, then backtest strategies directly in TradingView. He backtests a Donkey and Breakout strategy on Bitcoin daily from October 2017 to 2026, which returned $2.49M on $100k starting capital—or 2499% on $10k. Over the past two years alone (mid-2023 to now) it returned 92%, outperforming buy-and-hold on that timeframe. However, he notes the strategy only has a 43% win rate and performs best in sustained bull markets; in sideways conditions it suffers false breakouts and serious drawdowns. Miles emphasizes this is why you need to test thousands of strategies and pick ones that match your market regime bias. He's built a backtesting engine that runs 1,441 strategy tests in 3-4 hours, which he'll release separately on his channel with truly profitable setups.
Automated Scanning and Morning Brief Alerts
Miles demonstrates a scanner prompt that automatically reviews his entire August altcoin watchlist on the 4-hour timeframe, flagging coins near key support/resistance levels he's marked. The scan identified VVV as having the clearest pullback structure while noting that most altcoins on the list have broken recent support. For continuous monitoring when you're offline, Miles outlines two options: run a local device like an old Mac Mini or laptop 24/7, or use a VPS (virtual private server) paired with Hermes agent orchestration. Finally, he shows how to build an automated morning brief that sends alerts at 9 a.m. Dubai time, scanning all marked zones for actionable trades and breaching alerts throughout the day. The critical caveat: this requires you to manually draw your setup zones on the charts yourself—GPT reads binary proximity (near zone or not), so human accuracy on level placement is essential. Without that discipline, the alerts lose their edge.
Limitations and the Best Use Case
Miles identifies speed as his main complaint—the charting and analysis are solid but slower than he'd like for real-time trading. He also notes that GPT Astra is better than Claude Fable 5.1 for this specific use case because Astra excels at agentic work, computer automation, and speed, while Fable is smarter at planning and orchestration but not ideally suited for real-time chart interaction. The strongest use case Miles found isn't replacing your discretionary trading, but becoming a learning companion: you no longer need to Google what VRVP or RSI means—you ask questions about what's actually happening on your chart. You can build pattern recognition by asking GPT why price is reacting to macro events, bonds, CPI announcements, or Trump statements, developing intuition over time. Miles concludes that becoming AI-native by backtesting strategies, creating custom indicators, and learning inside TradingView will make you a better trader as this technology matures.