Video summary
BLP #38: Gini loh cara manfaatin AI buat trading yang benar!
Main summary
Key takeaways
Finance-focused summary (from the subtitles)
Core market / trading views (no explicit backtested performance metrics)
- The speakers argue that markets are (largely) efficient, meaning:
- If everyone could get the same “alpha,” it wouldn’t be alpha anymore.
- In a highly accessible environment, informational edges get competed away.
- AI cannot reliably make market predictions (directional forecasting), even with sophisticated models—outputs can change based on confidence/temperature and re-prompting.
- The practical framing is to shift from “prediction” to:
- Probabilities
- Risk management
- Analytics
Explicit recommendations / cautions
- Don’t believe “fully automated trading agents will make money while you sleep.”
- The guest calls this “bullshit”, noting that automation-only claims often ignore market complexity and changing regimes.
- Evaluate any “beat-the-market” goal versus Bitcoin buy-and-hold (BTC).
- The guest claims strategies that try to outperform often underperform BTC after:
- costs,
- invalidation,
- and regime changes.
- They emphasize that long-term BTC holders have a simpler path, with drawdowns they view as comparatively “not that big” relative to returns.
- The guest claims strategies that try to outperform often underperform BTC after:
- Backtests and “holy grail” strategies are treated skeptically.
- Mentions include unreliable backtest results (“power test garbage”).
- Even strategies that work early can fail later because:
- the market changes,
- and strategies require ongoing updates.
- Strategy applicability is pair-dependent.
- One approach won’t transfer cleanly across Bitcoin vs. Hyperliquid vs. ETFs/others.
- Parameters must be re-tuned for regimes (uptrend/downtrend/sideways).
- Risk management matters more as time in the market increases.
- “Season traders” are described as thinking in probabilities and possibilities, not certainties.
- The goal is to manage risk so outcomes remain survivable.
Crypto / instruments mentioned (tickers/assets)
- Bitcoin (BTC): repeatedly used as the benchmark for buy-and-hold and as the target outperform baseline.
- Hyperliquid: treated as a trading venue / pair context (not necessarily as a direct ticker in the subtitles).
- “ETs” / “ET”: mentioned generically (no specific ticker).
- US “XBT / Z XBT”: referenced as a “detective” account used for on-chain clustering (not clearly presented as an instrument).
- Trump tweeting: referenced as a potential news/narrative input (macro/event catalyst).
- CFTC: referenced in a regulatory/macro shock context.
- Funding rate: referenced conceptually for long/short positioning decisions.
On-chain / smart money / analytics approach (methodology framework)
The guest describes an analytics-first AI use case:
- Not prediction, but monitoring and scoring on-chain actions—especially for Hyperliquid traders.
Implied workflow / framework
- Track on-chain smart money / wallet positions, especially wallets with large portfolios.
- Use a scoring system to identify which traders are “best.”
- Send notifications when top/scored traders open positions.
- Emphasis: decision support, not fully automated execution.
Specific construction details mentioned
- A Hyperliquid tracker that:
- Monitors traders above stated portfolio thresholds ($2M / $5M).
- Alerts on positions opened above a notional threshold ($10M).
- Includes metrics such as:
- win rate
- EV (expected value)
- confidence score / conviction
- average size
- total size
- An additional example described:
- Watch “Wales” (large participants):
- 60–70% are said to be shorting everything.
- That behavior later reversed into a short squeeze after a catalyst (e.g., “Trump said something” and “CFTC flew”).
- Watch “Wales” (large participants):
Tools / automation (relevant for trading ops)
While the broader discussion is about AI agents, the finance-adjacent angle includes:
- Tracking wallet balances across on-chain/off-chain wallets:
- aggregating personal portfolio holdings and daily changes.
- Building a data aggregator / analytics layer for trading insights:
- The subtitles repeatedly stress it’s “just a data aggregator… news / smart money position… not something you can automate that easily.”
Numbers & thresholds explicitly stated
- Event dates (promo/conference context):
- November 11–12, 2026 (Jakarta)
- Trading/analytics thresholds:
- Portfolio size: $2M / $5M
- Position notional alert threshold: $10M
- Positioning example: 60–70% shorting everything → later short squeeze after a catalyst
- AI trading/indicator subscription claim:
- A subscription mentioned as 2,000 (currency unclear; later implied to be an expensive subscription type)
No explicit BTC price, yield, multiple, or portfolio return figures were provided.
Disclosures / disclaimers
- No explicit “not financial advice” disclaimer appears in the provided subtitles.
- The discussion is framed largely as personal opinion (e.g., “in my opinion,” “I don’t believe it,” “bullshit”).
Presenters / sources mentioned (at end)
- Josh Gult (main guest)
- Host / other participant repeatedly referenced as “Andreas / Andreas Ar topic”
- Website: andreasartopic.com
- Event promo source: Web Thweek Asia / Web3 Week Asia
- Other mentioned figures (not presented as co-hosts): Bis Yugo, Trump (as a news driver), CFTC (regulator)