> ## Documentation Index
> Fetch the complete documentation index at: https://docs-d0.donutbrowser.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Distilling Videos and Threads

> Drop in a two-hour keynote or a KOL's latest call. Your bot pulls out the parts that affect a trade, then checks whether the market already moved on them.

## The Scenario

Most trade-relevant information arrives in formats that are hostile to traders: a 98-minute keynote, a 40-tweet thread, a podcast. Watching all of it is expensive. Skipping it means missing the one paragraph that matters.

And a summary alone isn't enough. Knowing what someone said doesn't tell you whether it's tradeable — that depends on whether the market has already priced it in. Your bot does both halves: extract the signal, then cross-check it against flows, order books, and equities.

## Try It

* [Pull the trade-relevant parts out of this keynote and check whether the market has already priced it in:](https://t.me/donut_d0_agent_bot?text=Pull%20the%20trade-relevant%20parts%20out%20of%20this%20keynote%20and%20check%20whether%20the%20market%20has%20already%20priced%20it%20in%3A) *(paste a link after it)*
* [What is this KOL calling, and how accurate have his past calls been?](https://t.me/donut_d0_agent_bot?text=What%20is%20this%20KOL%20calling%2C%20and%20how%20accurate%20have%20his%20past%20calls%20been%3F)

## What You Get Back

### A long video, reduced to what's tradeable

> **You:** 6/1 Jensen Huang COMPUTEX 2026 keynote livestream runs over an hour — pull out the trade-relevant parts (how this AI narrative hits crypto sectors), and check whether the market has already priced it in.

> **Bot:** 📹 "Jensen Huang COMPUTEX 2026 Keynote" — NVIDIA Keynote (98min)<br />
> **Key announcements**
>
> * Unveiled next-gen AI acceleration architecture, big jump in inference compute
> * AI data-center capex + sovereign-AI demand keep expanding
> * Heavy focus on Physical AI / robotics and AI-agent deployment
>
> **Trade-relevant extraction**
>
> * AI narrative bullish → crypto AI sector (RENDER / TAO / FET etc.)
> * Also watch NVDA and AI-compute-chain equities
>
> **Market cross-check**
>
> * ✅ Crypto AI sector net inflows warming up over the last 7d, narrative not fully played out
> * ⚠️ NVDA firm after hours — part of the equities side may be priced in
> * ⚠️ On-chain AI tokens highly concentrated; chasing highs is risky
>
> **Conclusion:** AI narrative leans bullish medium-term; equities have partly reacted while the crypto AI sector lags — a catch-up candidate. Light position, avoid chasing.

The cross-check is the part that makes this useful. A summary would have stopped at "bullish AI narrative." The reply tells you where that's already in the price and where it isn't.

### A KOL call, with their track record attached

> **You:** This KOL is calling a trade again. Show me what he said this time, and whether his past calls were accurate.

> **Bot:** 🐦 @CryptoHayes (Arthur Hayes · \~731K followers)<br />
> **This signal**
>
> * Asset: \$HYPE
> * Direction: Long
> * Catalyst: "Hyperliquid fee-buyback flywheel + HIP-3 perp expansion; \$HYPE's value capture is underpriced"
> * Timeframe: short-to-mid (1–3 weeks)
>
> **Hit-rate backtest (last 90 days)**
>
> * Similar calls: 11
> * Direction correct: 7/11 (63.6%)
> * Average return: +8.2%
> * Max loss: −22% (two high-beta alt calls drew down hard)
>
> **Assessment:** Above-average hit rate overall, but high-beta names (like \$HYPE) are volatile. Small position with a 10% stop.<br />
> \[Set as monitor] \[Copy trade]

Note what the max-loss line does. A 63.6% hit rate sounds good until you see that the misses ran −22%. Both numbers get reported, because only one of them tells you how to size.

<Warning>
  Hit rates, average returns, and backtest windows in these examples are sample output from a specific moment, shown to illustrate the format — not a performance claim, and not a prediction. A KOL's past accuracy says nothing certain about their next call.
</Warning>

## When to Use

* **A keynote or earnings call you can't sit through.** Get the tradeable minutes without the other 90.
* **A KOL you're unsure about.** Before you act on a call, see how the last ten went.
* **A thread that's already moving price.** Ask whether the move has run its course.
* **Cross-asset narratives.** When something in equities or macro spills into crypto sectors, this is the fastest way to find which tokens it touches.

## Tips

* **Paste the link and the question together.** `"<link> — is this priced in?"` gets you further than the link alone.
* **Name the sector you care about.** "How does this hit AI tokens" gives a sharper answer than "summarize this."
* **Ask for the counter-argument.** Follow up with `"what would make this wrong?"` before you size a position on it.
* **Turn a call into a monitor.** If a signal looks worth watching but not worth taking, ask the bot to watch it — see [Scheduled Automations](/use-cases/scheduled-automation).

## Related

* [Market Analysis](/features/market-analysis) — the data the cross-check runs against
* [Market Research](/use-cases/market-research) — token-level research workflows
* [Thorp Strategy](/features/thorp-strategy) — how extracted sentiment feeds the signal engine
* [Acting on Thorp Signals](/use-cases/thorp-signals) — what happens when a signal clears the bar