Founders don't have an information problem. They have a synthesis problem. The information already exists — customer call recordings, investor transcripts you haven't reopened since the call ended, twelve versions of the pitch deck stacked in a Drive folder, midnight memos never read again. Everyone feels guilty about not using any of it, opens Notion, sees the graveyard, closes the tab.
The reason every note-taking system fails is structural. Retrieval is hard, capture is easy, and every tool on the market optimizes for the easy side — infinite hierarchies you'll never reopen, bidirectional links you'll never query, a graph view of a graveyard. None of them are bad tools. They're solving the wrong half of the problem. The fix isn't a better filing cabinet. It's an entity that can read every document at once and find the patterns hiding in the cross-section.
1. Structure a folder Claude can actually reason across
The whole system fits in one folder. Pick somewhere you'll actually remember — Desktop or iCloud Drive — name it second-brain, and don't move it. Inside, eight subfolders mapping to the functions a founder accumulates documents around, so Claude knows what kind of document it's looking at without reading it first:
second-brain/
├── customers/ calls, support patterns, feedback, churn, NPS
├── sales/ prospect calls, deal notes, pipeline reviews
├── fundraising/ investor calls, deck versions, data room docs
├── team/ one-on-ones, hiring notes, interviews, reviews
├── product/ roadmap, feature decisions, feedback synthesis
├── marketing/ positioning docs, campaign briefs, drafts
├── finance/ model snapshots, runway notes, P&L commentary
├── legal/ contracts, NDAs, terms reviews, IP, employment
└── personal/ podcast quotes, essay excerpts, reflectionsFormats mix freely inside each — markdown for your own notes, PDFs for deck exports, Excel for model snapshots, plain text for transcripts pasted from Fathom or Granola. Claude reads all of them. The one convention every file follows: YYYY-MM-DD-subject-type.ext — e.g. 2026-04-12-acme-customer-call.md, 2026-03-18-pitch-deck-v9.pdf. Date first so files sort in the order events happened, which is the order you'll want to reason about them.
2. The 10-second capture habit
The system fails if capture takes longer than ten seconds — you'll skip it exactly on the days you most need it. The workflow, by document type:
Calls — record with whatever you already use (Fathom, Granola, Otter, Fireflies, Loom), export the transcript, drag it into the right subfolder, rename with date and subject.
Pitch decks — every version you export to send to an investor, drop a dated copy in fundraising/. Don't judge whether the version matters. Save them all — comparing v3 to v17 six months later is enormous.
Financial models — a monthly snapshot in finance/. The working model lives wherever you build it; the snapshot is for Claude, freezing a moment so it can compare June's assumptions to October's reality.
Memos, decisions, hiring notes — write them directly inside Claude. Open Cowork or Claude Desktop, paste the content, say "save this as a note in team/ with today's date and the subject." No friction.
Small things (podcast quotes, essay paragraphs) — identical: paste, "save this in personal/ with today's date and a one-word topic."
The note format matters less than people think. The minimum: a one-line title, the date in the filename, the source (who/when/where), the raw content, and one or two sentences from you on why you saved it. That last one is what nobody does and what makes synthesis actually work — a folder of verbatim transcripts is inert; a folder where every note carries your own read on it gives Claude an anchor to reason from:
Acme Corp customer call
Date: 2026-04-12
Attendees: Sarah Chen (VP Ops, Acme), Ale, Marco
Source: Fathom transcript
What happened
Sarah walked us through the renewal conversation. She's
pushing back on the price increase but the underlying issue
is her team only uses two of the five modules. The other
three feel like wasted budget.
Verbatim quotes
"We're paying for a Ferrari and we drive it to the supermarket."
"If you could split out the analytics module separately I
could probably get a yes from procurement tomorrow."
Why I saved this
Third time this quarter a customer has asked to split
modules. This is becoming a pricing problem, not a feature
problem. Should pull every call where modules came up.Two minutes of writing after the call. That's enough.
3. Connect Claude to your second brain
Four steps, twenty minutes. Install Claude Desktop from claude.com/download and sign in — Cowork lives only in the desktop app, not the web version.
Create the second-brain folder, open a new session, click select folder, and point it there. Claude can now read, list, and create files in any subfolder. Have it create the structure for you:
second-brain/
├── customers/ customer calls, support patterns, feedback,
│ churn analysis, NPS
├── sales/ prospect calls, deal notes, pipeline reviews, demos
├── fundraising/ investor calls, pitch deck versions, data room docs
├── team/ one-on-ones, hiring notes, candidate interviews,
│ performance
├── product/ roadmap, feature decisions, customer feedback
│ synthesis
├── marketing/ positioning docs, campaign briefs, content drafts
├── finance/ financial model snapshots, runway notes,
│ P&L commentary
├── legal/ contracts, NDAs, terms reviews, IP, employment
└── personal/ podcast quotes, essay excerpts, books, founder
reflectionsIf most of your work already lives in Notion, Google Drive, Linear, or Slack, don't migrate anything — connect via MCP instead. The mental model is identical either way: give Claude read access to a corpus, and it reads across it. The folder is just the simplest version of the pattern; the synthesis prompts below work the same regardless of source.
Optional: install Obsidian (free) and open the same folder as a vault for a graph view of how your notes link. It reads plain markdown natively, changes nothing about how Claude works, and is purely for founders who find that visualization useful.
4. 15 prompts that find the patterns hiding in your own work
Each prompt names the source files, names the synthesis task, asks for specifics rather than summaries, and tells Claude to surface what you haven't asked about — that last part is what produces the unexpected insight.
Customers and sales
1. The pre-call briefing — run 30 minutes before any meaningful call.
Read every interaction we've had with [company or person]
across all subfolders in my second brain. Brief me as if you
were my chief of staff before this meeting. Cover:
relationship history, last three interactions in detail,
anything they've said about pain points or priorities,
commitments we've made, and the three things I should not
forget to bring up. Open the actual files, do not infer from
filenames.2. The recurring objection finder — run at quarterly planning or before a pricing decision.
Read the last [N] customer and sales call transcripts in
customers/ and sales/. Tell me the three objections, concerns,
or requests that come up most often, ranked by frequency. For
each, include three verbatim quotes from different customers
and tell me which segment each seems to cluster in. If you
find a pattern I haven't asked about, flag it.3. The churn root cause — run quarterly or after a churn spike.
Read every customer call and support note in customers/ from
customers who churned in the last six months. Find the signal
that appeared before they churned, the thing nobody acted on
at the time. Quote the specific lines. Tell me the leading
indicator I should watch for in current customers.4. The deal post-mortem — run after every lost deal you cared about, and every closed-won one.
Read everything in my second brain related to [deal name or
company]. Walk me through the deal chronologically: where it
started, where momentum was strongest, the moment things
shifted, what objection or signal we underweighted. Tell me
what I'd do differently knowing what I know now.Fundraising
5. The pitch evolution
Compare the pitch deck versions in fundraising/ from [date A]
to [date B]. For each major section (problem, solution,
market, traction, ask), tell me what got sharper, what got
softer, what was added, what was removed. If something
powerful got cut, flag it so I can consider putting it back.6. The investor objection map — run after every batch of investor meetings.
Read every investor call note and email exchange in
fundraising/ from the last [N] months. Cluster the concerns
into themes, ranked by frequency. For each theme, quote the
specific investors and tell me whether my latest deck answers
it directly or sidesteps it.7. The story consistency check — run before a raise.
Compare the narrative in my latest pitch deck with what I've
actually been saying in customer calls, internal memos, and
team all-hands over the last quarter. Where does the public
story diverge from the operating story? Quote both sides.Finance
8. The model vs reality check — run before sending an updated model to the board or committing to a hire.
Read my most recent financial model snapshot in finance/.
Then read the last eight weeks of customer calls, sales
notes, and team memos across all subfolders. Find where the
model assumes something the rest of my data contradicts or
doesn't yet support. Cite the file and quote for each.9. The runway stress test — run monthly.
Read my latest financial model and runway calculation in
finance/. Then read the last quarter of hiring notes, churn
signals from customers/, and any team or product memos. Tell
me which revenue, hiring, or retention assumptions are most
fragile right now, with supporting evidence.Team
10. The team blind spot finder — run quarterly, per direct report.
Read the last six one-on-ones I've had with [person] in
team/. Identify any pattern or theme they've raised more than
once that I haven't addressed. Quote the specific lines. Tell
me what they're actually asking for, beneath the surface ask.11. The hiring decision review — run before a hard performance conversation.
Read all interview notes for [candidate or hire] in team/.
Surface every concern raised, by whom, and how it was
resolved or dismissed. Tell me whether any concern is showing
up now in their performance, with evidence from 1:1 notes.12. The 1:1 theme tracker — run quarterly.
Read every 1:1 note in team/ from the last quarter, across
all reports. Cluster the recurring themes by report and
across reports. What is everyone collectively raising that
I'm hearing one report at a time?Strategic synthesis
13. The contradiction hunter — run when you sense you've changed your mind on something but haven't named it.
Read everything I have written or saved about [topic] over
the last [N] months. Find places where my recent thinking
contradicts my older thinking. Quote both sides. Tell me
which version reflects what I now believe.14. The postponed decision finder — run monthly.
Read every memo, 1:1 note, and strategy doc from the last
sixty days. Find the decisions I've been postponing without
naming them as decisions. Quote the lines where I'm circling
the same question without resolving it.15. The cross-brain pattern finder
Read across all subfolders in my second brain from the last
[N] months. Find an unexpected connection: two pieces of
information from completely different functions (a customer
call and a hiring note, a financial model and a podcast quote
in personal/) that point at the same underlying pattern. Be
specific about what the connection is and why it matters.These fifteen are a starting set. After a month you'll write your own, sharper for being specific to your company.
5. The weekly review that makes it compound
A folder plus on-demand synthesis already clears a real bottleneck. What turns it into something that compounds is a fixed weekly review. Every Friday afternoon, twenty minutes:
Read every file added or modified in my second brain over the
last seven days, across all subfolders. Read also the most
recent weekly review in personal/. Produce a one-page
reflection covering: what I was working on this week, the
three things that surprised me, the question that keeps
reappearing in my notes that I haven't named yet, the decision
I've been postponing that my own documents suggest I need to
make, and what I should re-read next week to keep building on
this. Be specific. Quote my own notes where useful.Save the output as weekly-review-YYYY-MM-DD.md in personal/. This is the move that compounds: next week's review reads this week's as context, the week after reads two, and by the end of a quarter the reviews trace the actual evolution of what you were thinking about — what you flagged but didn't act on, what kept coming back. After six months, ask Claude to read all of them and tell you what you've been circling without naming, and you'll get an answer no founder produces alone.
