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 mind map
Turn a YouTube video into a mind map
See the shape of a long explanation in one screen, including the parts it never joined up.
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.
CentreBranchesCross-linksGaps
Mind map
The structure of an argument, laid out so the shape is visible at a glance.
An hour of talking has a structure that is impossible to see while you are inside it. Ancher places the claim the speaker keeps returning to at the centre, branches the sub-arguments off it in the order they were introduced, draws cross-links where a later point depends on an earlier one, and marks the branches the speaker opened and never closed rather than tidying them away.
The honest version
The interesting part of a map is the branch that goes nowhere
Most visual summaries produce a tidy tree, because a tidy tree is what the format rewards. But a real explanation is not a tree — it doubles back, it depends on itself, and it leaves things hanging. A map that hides those has removed the only information you could not have got from a summary, and it looks more finished than the thinking actually was.
One second of one video, at 31:20, read two ways.Illustrative. The figure is invented; the gap it falls into is not.
...which connects back to the thing I mentioned earlier about and I will come back to that.
Diagram on screen — not readThe specific never reaches the draft
Mind map
Diagram on screen
Feedback loop, step 3
from 31:20
Lands in the draft, with its timecode
- The centre is the claim with the most time spent on it, not the video's title.
- Cross-links are drawn where a later branch depends on an earlier one, which is what makes it a map rather than an outline.
- Branches the speaker opened and abandoned stay on the map, marked as gaps rather than pruned.
What comes out
Mind map, section by section
The structure of an argument, laid out so the shape is visible at a glance.
Centre
The claim the speaker spent the most time on, which is frequently not the one in the title.
Branches
Sub-arguments in the order they were introduced, so the map preserves how the thinking was built.
Cross-links
Cross-links wherever a later point depends on an earlier one, drawn rather than described.
Gaps
Branches opened and never closed, marked as open instead of quietly removed.
Legend
What a line means here, so the map is readable by someone who did not watch the video.
Video-to-mind-map 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 mind map of this video. (1) Put the claim the speaker spent the most time on at the centre, not the title. (2) Branch the sub-arguments in the order they were introduced. (3) Draw a cross-link wherever a later point depends on an earlier one, and say what the dependency is. (4) Mark any branch the speaker opened and never closed as a gap rather than removing it. (5) Include a legend explaining what a line means. Keep it under thirty nodes and say what you merged to get there.
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 | Centre | Time on a topic is measurable from caption timings, and it is a better centre than any title written for a thumbnail. |
| Caption track with timings | Cross-links | Back-references — as I said earlier, this connects to — are literal sentences, and each one is an edge on the map. |
| Key frames from the video | Branches | Speakers draw their own structure on screen, and that diagram is a better skeleton than anything inferred from words. |
| Caption track with timings | Gaps | Promises to return to something are easy to find and almost never kept, which is what makes them worth marking. |
| Title, channel, duration, date | Legend | Title, channel and duration sit in the legend so the map carries its own provenance when it is shared on its own. |
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 mind map. Ancher fills centre, branches, cross-links, gaps, legend 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
- A map with no cross-links is an outline drawn sideways. If nothing connects across branches, the structure was not extracted, only the headings.
- Long videos produce wide maps. Past roughly thirty nodes it stops being one screen and starts being a diagram you have to navigate, which defeats the point.
- A speaker's digression can look like a branch. Time spent is the test — a topic held for four minutes is a branch, one held for twenty seconds is an aside.
Questions people ask
How is this different from an outline?
An outline is a tree and can only show hierarchy. A map can show that branch four depends on branch one, which is usually the actual insight in a long explanation and the thing a nested list has no way to represent.
What if the speaker draws their own diagram?
Then that becomes the skeleton. A diagram someone drew to explain their own argument is a better structure than anything inferred from the words around it, and the frames it appeared in are kept with it.
Can I edit the map afterwards?
Yes, and you should. The map is a reading of the video, and the moment you disagree with a placement is the moment you have understood the argument well enough to be worth trusting over the extraction.
Bring your own YouTube videos
Everything you save lives in one workspace, so the mind map is built from your sources — not from a model's memory of the internet.
Open Ancher →Not a YouTube video? The same mind map also comes from X thread, LinkedIn post, Instagram post.