Andrej Karpathy — Intro to LLMsY Combinator — How to get startup ideas
Your YouTube video
Paste a YouTube link. Nothing to upload.
←YouTube video
Make a executive summary
Turn a YouTube video into an executive summary
Paste a link. One page that opens with the conclusion, carries the real numbers, and says what is being asked for.
No link handy? Try one:
Free to try, no card. Your YouTube video opens in Ancher.
Caption track with timingsKey frames from the videoDescription and linked sources
Kept in parts
Every one of these stays separately addressable instead of collapsing into one block of text.
Bottom lineSituationKey numbersRisks
Executive summary
One page that leads with the conclusion and the decision being asked for.
An executive summary leads with the answer, which is the opposite of how a video is paced. Ancher pulls the conclusion out of wherever it landed, takes the figures from the frames they were displayed on rather than from a spoken approximation, and keeps whatever the speaker admitted was uncertain in a risks section instead of quietly dropping it.
The honest version
Most video summaries round the numbers away
Speakers approximate out loud — "a bit over four dollars", "call it twenty percent" — while the exact figure sits on the slide behind them. A summary built from the caption track inherits the approximation, and an executive summary that says "a bit over four dollars" is not one you can forward to anyone.
One second of one video, at 27:52, read two ways.Illustrative. The figure is invented; the gap it falls into is not.
...margin came in at, call it, give or take, which is roughly where we guided.
Gross margin, Q3 — not readThe specific never reaches the draft
Executive summary
Gross margin, Q3
71.4%
from 27:52
Lands in the draft, with its timecode
- Figures come from the frame they were displayed on, so the summary carries the number rather than the speaker's rounding of it.
- Anything the speaker hedged is preserved as a risk instead of being flattened into confidence the source never had.
- Each number keeps its timecode, so a reader who wants to challenge a figure can get to it in one click.
What comes out
Executive summary, section by section
One page that leads with the conclusion and the decision being asked for.
Bottom line
Two sentences carrying the conclusion, taken from wherever it appeared, which in a video is rarely near the start.
Situation
Scoped to what the video actually covers, so the summary does not imply a breadth the source never had.
Key numbers
Figures read off the frames they were shown on, each with its timecode, not the spoken approximation of them.
Risks
Every point where the speaker hedged, caveated or contradicted themselves — the part a confident summary deletes.
The ask
The decision being requested, or an explicit note that the video does not ask for one, which is more useful than inventing it.
Video-to-executive-summary prompt
The link on its own gets a summary. This is the instruction that produces the 5 sections above, in that order, with the rules that keep them honest. Paste it with your YouTube video — in Ancher, or in whatever assistant you already use.
Write a one-page executive summary of this video. (1) Open with the bottom line in two sentences, taken from wherever the conclusion appears. (2) Scope the situation to what the video actually covers. (3) Pull key numbers from on-screen frames, not from spoken approximations, and cite a timecode for each. (4) Carry every hedge and caveat into a Risks section. (5) State the decision being asked for, or write "No explicit ask" if there is none. Never invent a figure to fill a gap.
The detail, if you want it
How each part of a YouTube video becomes a section
| From the YouTube video | Into | Why |
|---|---|---|
| Key frames from the video | Key numbers | Slides carry the exact figure while the narration carries the rounded one; this row is the whole reason to read frames at all. |
| Caption track with timings | Bottom line | The conclusion is found by where the speaker stops hedging, which the caption track shows and the video's structure hides. |
| Caption track with timings | Risks | Hedging language is extracted verbatim, so the summary inherits the source's uncertainty rather than the model's confidence. |
| Title, channel, duration, date | Situation | Duration and publication date bound the claim — a summary of a two-year-old talk should say so on its face. |
| Description and linked sources | The ask | A call to action in a video usually lives in the description, not the audio, so the ask is often recovered from the links. |
Everything Ancher keeps from a YouTube video
Caption track with timings
Published and auto-generated captions both come through, and every line keeps the second it was spoken.
Key frames from the video
Frames are sampled and scored, so a number shown on a slide survives even though it was never said out loud.
Description and linked sources
The creator's own cited links come along, so a claim can be traced past the video.
Likes and comment counts
Reception is a weak signal about correctness and a strong one about which claims got attention.
Title, channel, duration, date
Basic provenance, kept so the finished work can attribute the video properly.
The four steps
- Add the YouTube video. Paste a YouTube URL. Ancher pulls the caption track with its timings, samples key frames from the video itself, and keeps the description links and engagement counts on the same source.
- Nothing gets flattened. Caption track with timings, key frames from the video, description and linked sources, likes and comment counts, title, channel, duration, date stay separately addressable instead of collapsing into one block of text.
- Ask for a executive summary. Ancher fills bottom line, situation, key numbers, risks, the ask using the mapping above, not a generic summary pass.
- Check it before you use it. Every specific keeps a link back to where it came from in the YouTube video, so a claim can be verified without reopening the source.
Where this goes wrong
- If a figure in the summary matches the spoken approximation rather than the slide, the frame was not read. Check any number that looks suspiciously round.
- A one-page limit is the whole discipline. If everything important does not fit, the bottom line has not been sharpened enough yet.
- Videos are edited. A summary of a published recording is a summary of what the creator chose to keep, and should not be treated as a record of the underlying event.
Questions people ask
Is this different from the summary YouTube shows?
That one exists to help a browser decide whether to watch. An executive summary exists to let someone act without watching, which needs a conclusion, exact figures and a stated ask — none of which a description-level summary has any reason to produce.
What if the video has no numbers in it?
Then the key numbers section says so rather than manufacturing metrics. A summary that invents figures to look complete is worse than a short one, and this is the single most common failure mode in AI summaries.
Can I summarize several videos into one page?
Yes, and that is usually the stronger use. A summary across four quarterly updates can show a trend that no single video states, and disagreements between them surface instead of being averaged away.
Bring your own YouTube videos
Everything you save lives in one workspace, so the executive summary is built from your sources — not from a model's memory of the internet.
Open Ancher →Not a YouTube video? The same executive summary also comes from PDF, X thread, LinkedIn post.