Capture Conversations for Better Content Ideas

Most of your best episodes and videos didn't start in a planning session. They started while you were arguing with someone, or explaining something for the third time, or narrating a bad experience while it still annoyed you. A real position comes out of pressure. Somebody's in front of you, so you skip the hedging and get to what you actually think.

The supply isn't your problem. A discussion at dinner, a client call where the same question comes up again, a walk where you talk yourself from a vague irritation into a three-part argument. In a normal week that's a dozen usable seeds. Almost none reach your notes app, and the few that do turn up stripped — "pricing episode", written on a train six weeks ago, with the reasoning it stood on left behind.

Count the steps. Notice the thought, get the phone out, wake the screen, find the app, tap record, talk, then later add enough context that it still makes sense in a month. Six steps, while you're mid-sentence, driving, or holding a camera. So it doesn't happen. Not for lack of discipline — capture just costs more than one fragment seems worth at the time.

Two things change that. First, you stop trying to react in the moment at all. You start capture on your wrist when the dinner or the drive starts and it keeps running until you stop it, so the good line needs nothing from you while it's being said. It never starts on its own, so you do have to remember to start it. Second, what gets captured is the conversation, not a summary you dictate afterwards. Pastok records from the Apple Watch, in person or on a call, with no bot joining and nothing to install on anybody else's device.

That second one matters more than it sounds. A dictated memo gives you your recollection of a conversation, already flattened. The conversation itself gives you the objection in the other person's words, the moment you shifted position, and the sentence you landed on when you finally got it right. You can't reconstruct those.

• Anywhere somebody disagreed with you. The best of the lot. You get their objection in their words, which is the tension an episode needs, and you find out where your audience will be standing.
• Questions you've now answered more than twice. Listeners, clients, DMs, somebody after a talk. Repetition is the cleanest demand signal you'll ever get, and you'll miss it, because the three instances were nine days apart.
• An experience while it's still happening. The shoot that fell apart, the flight that told you exactly what the company cared about. Narrate it in the car while you're still annoyed. Annoyance is specific. Recollection is general.
• Ten minutes of thinking out loud with nobody there. Not a note — a monologue, where you talk yourself from "something's off about this" into a three-part argument.

You don't have to sort any of it at the time, and you shouldn't try. Say one extra sentence — why it struck you — then stop. "Note this: the cheap gear thing. Right for beginners, wrong for anyone shooting weekly." Ten seconds. Enough to find later and enough to use.

Archives fail on retrieval, not storage. A folder of recordings goes dead for the same reason a notes app does: nothing brings the good parts back up unless you go digging, and you won't go digging on planning day. So the search has to run without you. In Pastok you write a Skill once in ordinary language and pick when it runs — on demand, on a schedule, or when it spots a certain kind of conversation. That last one's best effort, not a promise. Set it for Sunday evening and the shortlist is waiting, as Markdown or a PDF.

The one to write first:

• "Every Sunday evening, go through this week's conversations. Find anything I got animated about, anything I explained more than once, and anything somebody disagreed with me on. For each, give me the topic in one line, my own words at the time, and who I was talking to. Markdown."

A few worth running on demand:

• "List the stories I've told more than twice in the last 90 days, with the most recent telling verbatim."
• "Find every point this month where somebody pushed back on something I said. Their objection in their words, mine in mine."
• "What questions have people asked me this month that I answered off the cuff?"

Two things about that search. It matches meaning, not keywords, so asking about pricing surfaces the conversation where you only ever said "what we charge". And because it's interpreting, ask it one more thing before you build anything on the answer: "Which conversations did you use?" Then go read them.

A shortlist entry is still just a topic. What separates a strong forty minutes from a thin twelve is material: a claim you hold, a story with specifics in it, an objection worth answering, and a reason it matters now. Assembling that in the hour before you record is what produces the stretch of general advice you can hear yourself sliding into. If the idea is software rather than an episode, use the same captured context to turn the conversation into a working spec before you hand it to a coding agent.

So ask for it outright: "Pull everything I've said about cheap gear this year, in order, and mark where I changed my mind." Order matters — the sequence you worked the argument out in is usually the running order for the episode. And a change of mind is the best middle an episode can have, which is exactly the thing you'd never remember unprompted.

After that the outline is mostly decided. Open on the objection, because that's the version your audience turns up holding, then answer it with what you said in the moment. One caveat if you're quoting anybody: attribution slips in a three-way conversation, so check the line against the full transcript on the web app first.

• Dictation apps. Good at one thing: you have a thought, you narrate it, you get clean text. But capture is a deliberate act, so what you keep is your summary of a conversation, not the conversation. And the archive is a flat list nothing reads for you.
• Second brain systems. They solve retrieval properly and charge you in the currency you have least of. They work if you tag, link and review on a schedule. For most people who publish for a living that sticks for about six weeks.
• DIY chains. Recorder into transcription into a notes database into a chat prompt. They work, and you maintain them forever. Every hop breaks quietly, and you find out on planning day.
• Meeting tools. Built around the calendar, so they do good minutes for scheduled calls and can't see the dinner, the drive or the hallway. Your material is mostly in those.
• Idea generators. Trend and keyword tools tell you what's being searched, which helps with a title and not with material. A subject with no position attached is what you were stuck with already.

What's left is the actual job: capture you start once and then forget about, plus a search you write once that runs itself. That's what Pastok is arranged around, and it's why the wrist isn't a gimmick. It's the difference between a system you use and one you meant to use.

• Solo commentary. Your positions are the product, so an archive of them is the raw material. Strongest fit.
• Interview shows. Topic choice is guest choice, so use it upstream: pull what you've already said about the guest's field before you write questions, and you stop asking what they answer on every show.
• YouTubers working to a script. Same problem, tighter deadline. A shortlist with specifics attached is what survives the edit.
• Coaches, consultants and freelancers who publish. You spend the week being asked the same six questions. That's a content calendar you're throwing away.
• Newsletter writers. A section you've already argued out loud drafts several times faster than one you're inventing at the keyboard.

Two caveats. If your show is researched rather than argued — history, explainers, subjects you never discuss at dinner — reading and keyword tools will do more for you than this will. And on recording other people: asking "mind if I keep this running?" feels like a lot for about a week, then it stops being a thing. Where something's told to you in confidence, don't. If that whole area isn't for you, the solo version still works — three walks a week of thinking out loud gives the Sunday search plenty to find.

Every week you make more usable material than you could publish, and most of it expires because catching it costs more than it seems worth at the time. Start capture when the conversation starts instead of trying to catch the line after it lands, and write the weekly search once instead of never. What comes back isn't a list of topics. It's a list of things you've already said well, in the words you used.

• Turn voice notes into a searchable knowledge base
• AI meeting notes for consultants
• AI meeting recorder for in-person meetings