Content ideas from meetings: never run out of things to write again

Claude scans your transcripts for stories, anecdotes, customer truths, and sharp observations. Output is a running ideas file you mine for LinkedIn, articles, podcasts, talks.

What you'll have when you're done

A running file, content-ideas.md, that fills itself every week. A script reads your meetings, pulls the moments worth writing about, sorts them into stories, customer truths, hard-won lessons, and frameworks, and ranks them by how interesting they are. Each idea comes with the verbatim seed it grew from and a note on which channel it fits.

You will never again open a blank page wondering what to post. The hardest part of being a CEO who writes isn't the writing, it's deciding what's worth saying. This workflow does the deciding from the raw material you generate every day anyway: the anecdote a customer handed you on a call, the thing you said in a board meeting that made everyone go quiet, the lesson you learned the expensive way. You open the file and pick. The blank page never starts blank.

"I have nothing to write about" is a lie your memory tells you

I went through a stretch where I knew I should be posting, knew the content compounds, knew the founders who write build an audience that pays off for a decade. And every time I sat down to write, my mind went flat. I had nothing. Then I'd close the laptop. Then two days later, on a customer call, I'd say something sharp about why most onboarding fails, the customer would write it down, and I'd think, that's a post. Then the call would end and the thought would be gone by the next meeting.

The blank page isn't an idea problem. It's a memory problem. You have a dozen genuinely good things to say every week. You say them out loud, in meetings, to one or two people, and then you forget you ever said them. By the time you sit down to write, the well looks empty, not because it is, but because the water already drained into conversations you don't remember in enough detail to mine.

I tried the usual fixes. A notes app for capturing ideas (I never opened it in the moment, and writing in it later felt like homework). A content calendar (it just turned the blank page into a blank page with a deadline). A swipe file of other people's posts (which made everything I wrote sound like everyone else). None of them touched the actual problem, which was that the best material was being generated and lost in the same breath.

The breath was being recorded the whole time. Granola caught every sharp thing you said and every truth a customer handed you. This workflow reads it all back and pulls out the parts worth writing. The well was never empty. You just couldn't see into it.

What you need first

• The [Granola][1] → markdown pipeline already running. This workflow runs on the meeting library the pipeline produces. If you haven't set it up yet, do [Granola → markdown][2] first. Thirty minutes. Comes back when done.
• Claude Code installed and pointed at the folder your Granola transcripts land in. Free tier handles the extraction.
• A week of meetings in the folder. This runs weekly, so you need a week of conversations to mine. More variety helps: customer calls, team meetings, and board prep each surface a different kind of idea.
• Two or three things you've published before, if you have them. Optional. Dropping a few of your best past posts in a folder lets Claude calibrate to what "good" looks like for you specifically, so it ranks ideas the way you would. Skip it on the first run and add it once you see what the unguided output looks like.

Step-by-step

Step 1Save the extraction prompt

The prompt sorts and ranks, so it only surfaces the moments actually worth your time. Save it once.

mkdir -p ~/.claude/prompts
touch ~/.claude/prompts/content-ideas.md

Paste the prompt below into that file. The categories and the "skip generic" rule are what you'll tune.

Read every meeting transcript in ~/notes/granola/ from the last 7 days.
Pull out the moments worth writing about.

Categories to look for:
- **Story:** a specific anecdote (a customer reaction, a candidate moment, a
  board pushback) that could open a LinkedIn post or an article.
- **Truth:** a customer insight or an operator observation that cuts against
  the conventional wisdom.
- **Lesson:** a mistake I made or watched, and what I'd do differently.
- **Framework:** a way of thinking I articulated that could become a piece.

Append each one to ~/notes/content-ideas.md in this format:

## {category}
### {one-line hook}
- **Source:** {filename} · {date}
- **Verbatim seed:** "{quote}"
- **Why it's interesting:** one sentence
- **Channel fit:** LinkedIn / Article / Talk / Podcast

Rank within each category by how interesting the idea is, sharpest first.
Skip anything generic. If it's a thought a thousand other founders have
already posted, leave it out.

Step 2Run it and read the top of each category

Open Claude Code in your notes folder:

cd ~/notes
claude

For the first run, point it at a longer window so the file opens with real volume. Paste the prompt but change "last 7 days" to "last 30 days." Claude reads the month and builds content-ideas.md.

Read the verbatim seed under each idea, not just the hook. The hook is Claude's pitch; the seed is the raw moment, and the seed is what tells you whether there's a real post here or just an interesting-sounding label. The ones that make you think "oh, I forgot I said that" are the gold. The ones where the seed is thinner than the hook are Claude reaching, and you tighten the "skip generic" rule until those stop appearing.

Step 3Calibrate it to your taste

The first run will rank some weak ideas high and bury some strong ones, because Claude is guessing at what "interesting" means to you. Fix that by showing it.

mkdir -p ~/notes/best-content

Drop two or three of your best-performing past posts into that folder as plain text, then add a line to the prompt: "Read ~/notes/best-content/ first to learn what a strong idea looks like for me, and rank against that bar." The difference between a generic idea list and one tuned to your voice is three examples of your own best work. Claude reads what landed for you before and starts surfacing more of the same kind of thing.

Step 4Mine the file, don't just grow it

A file that only fills up is a graveyard. The workflow pays off when you pull from it.

Build one small habit: once a week, open the file, pick the single highest-ranked idea you haven't used, and write it. One post. The seed is already there, the angle is already there, the customer's own words are already there. You're not starting from blank, you're editing from a strong start, which is a five-minute job instead of a fifty-minute one. Mark the idea as used (move it to a "shipped" section at the bottom) so the top of the file always shows what's fresh.

Step 5Make it a standing weekly run

Don't rely on remembering to run it. Two ways to make it automatic:

Option A · One saved command (lowest effort). Each week, open Claude Code and say: "Run ~/.claude/prompts/content-ideas.md against this week's transcripts in ~/notes/granola/ and append to ~/notes/content-ideas.md." Ten seconds, and the file refills with the week's best moments.

Option B · A Claude Code [Routine][3] (fully automated). Get Claude to set up a Routine that runs the extraction every Sunday and appends the week's ideas, then sends you the top three as a short note. Set it once and you get a weekly shortlist of things to write, pushed to you, ranked, with the seeds attached.

How you'll know it's working

The signal is a published post, not a full file.

The first time you ship a LinkedIn post that came straight out of the ideas file, this workflow has earned its keep. You'll feel the difference in how fast it went: you opened the file, the seed was there in the customer's own words, you spent ten minutes shaping it instead of an hour staring at a blank page, and you hit post. The thing that used to feel like a separate job you never had time for becomes a thing you do between meetings.

The second signal is that you stop dreading the blank page entirely. A few weeks in, "I have nothing to write about" stops being true and stops being a feeling. You have a ranked file of things only you could write, refreshing every week. The constraint moves from ideas to time, which is a much better problem to have.

When it breaks

• The ideas aren't new. Claude surfaces something you already wrote about. The fix is the calibration in Step 3: point it at your published work and add a rule, "skip anything I've already covered in ~/notes/best-content/." It cross-checks and stops handing you reruns.
• It misses your sharpest insight. You know the best moment of the week was a thing you said in Tuesday's call, and it's not in the file. Usually the seed was there but Claude ranked it low. Re-run with the example calibration, and add the missed moment as a "this is the bar" example. The ranking sharpens fast once it sees what you actually value.
• Everything it surfaces is a "framework." Claude over-indexes on the tidy, abstract stuff because it pattern-matches well. The stories and customer truths are usually the better content. If the file skews abstract, add a rule: "Prefer specific stories and verbatim customer truths over abstract frameworks." Concrete beats clever almost every time on LinkedIn.
• The file fills and you never write from it. The silent failure. A growing ideas file you never mine is just a tidier version of having nothing. The fix is Step 4: one post a week, picked off the top, edited not authored. The habit is small on purpose, because a small habit you keep beats a big one you abandon.

Where this fits in your harness

This is part of the content layer of your [harness][4]. The [Granola → markdown pipeline][2] made every conversation searchable. This workflow reads those conversations for a specific purpose: the raw material of everything you could publish, pulled out before it drains away into forgotten meetings.

The siblings most relevant to this workflow:

• [Book ideas from meetings][5] · the long-form version of the same idea. Content ideas feeds the post you ship this week; book ideas organizes the same kind of material against a chapter outline so a bigger project accumulates in the background. Run both off one week of meetings and you're feeding the short game and the long game at once.
• [Weekly team and investor updates][6] · the same meeting record aimed at the audiences you owe rather than the audience you're building. One reads the week for what's worth publishing; the other reads it for what's worth reporting. Together they turn your meetings into most of your writing.

See the [Granola pillar][7] for the full pipeline and the other workflows that compound on top.

[1]: /workflows/what-is-granola [2]: /workflows/granola-to-markdown [3]: /workflows/claude-routines [4]: /workflows/what-is-a-harness [5]: /workflows/book-ideas-from-meetings [6]: /workflows/team-and-investor-updates [7]: /blog/granola-for-ceos-highest-roi-ai-install