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 content brief
Turn a YouTube video into a content brief
Paste a link. A writer gets the angle, the outline and the specifics — without having to watch anything.
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.
AngleTarget readerOutlineMust-include points
Content brief
An instruction set for a writer: angle, reader, outline, must-include points, sources.
A brief is an instruction set, not a summary, and its job is to let a writer produce the piece without opening the source. Ancher extracts the claim the speaker defends hardest as the angle, lists every number and named entity as must-include points with their timecodes, and hands over the creator's own cited links so the finished article can stand on primary sources rather than on the video.
The honest version
A brief that says "cover the main points" is not a brief
Most video-to-brief output is a summary with an outline stapled to it. A writer receiving it still has to watch the video to find anything specific enough to write with, which defeats the entire purpose. What makes a brief usable is a list of concrete things the piece must contain — figures, names, dates — and a source for each.
One second of one video, at 18:05, read two ways.Illustrative. The figure is invented; the gap it falls into is not.
...and the retention curve flattened around which is the part everyone gets wrong.
On the chart — not readThe specific never reaches the draft
Content brief
On the chart
Day 34
from 18:05
Lands in the draft, with its timecode
- Must-include points are extracted as specifics with timecodes, so a writer can verify without scrubbing the video.
- The angle comes from where the speaker spent time defending a position, not from the title, which was written for a thumbnail.
- Description links become the writer's source list, so the piece cites primaries instead of citing a video.
What comes out
Content brief, section by section
An instruction set for a writer: angle, reader, outline, must-include points, sources.
Angle
One sentence, from the claim the speaker returns to most often rather than the one the title advertises.
Target reader
Inferred from who the speaker addresses and what they assume is already known, which sets the register of the piece.
Outline
Ordered for a reader rather than a viewer, so the conclusion moves up and the setup gets compressed.
Must-include points
Every number, date and named entity, each with a timecode and a note on whether it was spoken or shown.
Sources to cite
The creator's cited links, so the article can attribute the claim to where the creator got it.
Video-to-content-brief 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.
Build a content brief from this video. (1) State the angle in one sentence, drawn from the claim the speaker defends at greatest length, not from the title. (2) Identify the target reader from who the speaker addresses. (3) Give an outline ordered for a reader, conclusion first. (4) List every number, date and named entity as a must-include point, each with a timecode and marked SPOKEN or ON-SCREEN. (5) Build the source list from the description links, not from the video. Mark anything you could not attribute.
The detail, if you want it
How each part of a YouTube video becomes a section
| From the YouTube video | Into | Why |
|---|---|---|
| Caption track with timings | Angle | Time spent defending a position is measurable in the caption timings, and it is a far better angle signal than the title. |
| Key frames from the video | Must-include points | The specifics that make a piece worth publishing are usually on a chart, and a caption scrape returns none of them. |
| Caption track with timings | Outline | Subject changes in the caption timings give section breaks, which are then reordered rather than used as-is. |
| Description and linked sources | Sources to cite | A writer citing the creator's sources produces a stronger piece than one citing the creator, and the links make that possible. |
| Likes and comment counts | Target reader | Which claims drew comment volume is a rough read on which audience actually showed up, independent of who the channel targets. |
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 content brief. Ancher fills angle, target reader, outline, must-include points, sources to cite 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 the must-include list contains no numbers, dates or names, the brief is not specific enough and the writer will produce something generic.
- Briefing a writer to follow the video's structure produces a transcript in prose. The outline has to be reordered for reading before it goes out.
- Sponsored segments read exactly like editorial in a transcript. Check any product mention in the must-include list before a writer builds a section around it.
Questions people ask
Can the writer skip watching the video entirely?
That is the test of whether the brief worked. Every must-include point carries a timecode, so the writer can spot-check a specific claim in thirty seconds rather than watching the whole thing. If they have to watch it, the brief was too thin.
How do I avoid duplicating the creator's own write-up?
Build the brief from several videos making the same claim. Where they disagree is an angle no single creator has published, and it is also the part that is not already indexed.
Can Ancher write the piece too?
It can draft from the brief in your voice, but the brief is the artifact worth having — it is what lets a human decide the angle before any words get spent on it.
Bring your own YouTube videos
Everything you save lives in one workspace, so the content brief is built from your sources — not from a model's memory of the internet.
Open Ancher →Not a YouTube video? The same content brief also comes from PDF, X thread, LinkedIn post.