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 task list
Turn a YouTube video into a task list
Pull the things someone has to do out of a walkthrough, each with the moment it was said.
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.
ActionsOwnersDatesBlockers
Task list
Actions with owners and dates, extracted rather than transcribed.
A task list from a video is only useful if every line is something a person could start on Monday. Ancher takes the instructions and commitments — the imperatives, the deadlines named out loud, the steps demonstrated on screen — and writes each as an action with the timecode it came from, then says plainly which ones have no owner, because a YouTube caption track carries no speaker labels to assign them from.
The honest version
Captions cannot tell you who said it, so we do not guess
Every tool that turns talk into tasks wants to fill an owner column, and on a video with no speaker labels the only way to fill it is to guess. A wrong owner is worse than a blank one: it gets assigned, ignored, and then blamed. The owner column stays empty unless someone named themselves or was named out loud, and the list says how many are unowned.
One second of one video, at 06:31, read two ways.Illustrative. The figure is invented; the gap it falls into is not.
...so before you do anything else, change this setting right here to or none of the rest will work.
Setting shown on screen — not readThe specific never reaches the draft
Task list
Setting shown on screen
Retention: 90 days
from 06:31
Lands in the draft, with its timecode
- An action is written as something a person could start, not as a topic that was discussed.
- Owners are filled only from a name spoken aloud; everything else is marked unassigned rather than inferred.
- Dates come from what was actually said — "by the end of the quarter" stays as written rather than being resolved into a date nobody agreed to.
What comes out
Task list, section by section
Actions with owners and dates, extracted rather than transcribed.
Actions
One line per thing to be done, written as an instruction rather than as a summary of a discussion.
Owners
Filled only where a name was spoken; everything else is marked unassigned, with a count.
Dates
Deadlines exactly as they were said, including the vague ones, rather than resolved into calendar dates.
Blockers
Anything named as a prerequisite, so an action that cannot start yet is visibly not startable.
Where each came from
A timecode on every line, so a disputed task can be checked in under a minute.
Video-to-task-list 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.
Extract a task list from this video. (1) Write one line per action, phrased as something a person could start, and skip anything that is a topic rather than a task. (2) Fill an owner only where a name was spoken aloud; otherwise mark it unassigned and count them at the end. (3) Record deadlines in the exact words used, including vague ones. (4) List anything named as a prerequisite as a blocker. (5) Put a timecode on every line. Do not invent owners or resolve vague dates into calendar dates.
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 | Actions | Imperatives and commitments are a speech pattern, and the caption track is the only place they appear in full. |
| Key frames from the video | Actions | In a walkthrough the step is performed rather than described, so the frame is what turns "change this setting" into a task with a value in it. |
| Caption track with timings | Dates | Deadlines are spoken, usually vaguely, and keeping the original phrasing is more honest than inventing precision. |
| Caption track with timings | Blockers | Prerequisites show up as conditionals — before, until, once — which are findable in the transcript and invisible in a summary. |
| Title, channel, duration, date | Where each came from | Duration and publish date bound the list, so a task from a two-year-old walkthrough is not mistaken for current guidance. |
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 task list. Ancher fills actions, owners, dates, blockers, where each came from 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 video is not a meeting. Nothing in it was agreed to by your team, so treat every line as a candidate task rather than a commitment.
- Rhetorical imperatives read exactly like instructions. "Think about what that means for your funnel" is not a task, and a list full of those is noise.
- Steps in a walkthrough go stale with the interface they were filmed on. Check the publish date before working through a settings sequence.
Questions people ask
Why is the owner column mostly empty?
Because YouTube caption tracks carry no speaker labels, so there is nothing to assign from unless a name was said out loud. We would rather hand you a list with an honest gap than one where a third of the owners were guessed.
Does this work for a recorded meeting on YouTube?
The actions extract fine, but the same speaker-label limit applies. If you need owners, the recording source handles an uploaded file with better separation than a published video does.
Can it tell a real deadline from an aside?
It keeps the phrasing rather than the judgement — "by Friday" and "at some point" both survive as written. Deciding which one counts is yours, and it is a decision you can make in seconds once the phrasing is preserved.
Bring your own YouTube videos
Everything you save lives in one workspace, so the task list is built from your sources — not from a model's memory of the internet.
Open Ancher →Not a YouTube video? The same task list also comes from PDF, recording, web page.