Page 3 of 8

Feeding Your Brain

Establish a sustainable research ingestion workflow. Your brain is only as good as what you feed it.


The Daily Routine (5 minutes/day)

Morning: Newsletter Check

# Run newsletter fetch
"Check newsletters"

# Or be explicit
"Use newsletter-daily-check skill to fetch new newsletters"

What happens:

  1. Connects to Gmail via API
  2. Finds newsletters with label "8020brain"
  3. Saves as markdown in research/newsletters/[source]/
  4. AI generates summary with key insights
  5. Updates index in context/ideas/[source].md

Result: All your newsletters captured, summarized, and searchable.


As-Needed: YouTube Transcripts

Got a valuable video? Grab the transcript:

# Fetch a video transcript
/fetch https://youtube.com/watch?v=xyz123

What happens:

  1. Fetches transcript from YouTube
  2. Saves to research/youtube/[channel]/
  3. AI generates summary
  4. Updates index in context/ideas/youtube-[channel].md

Result: Video insights captured without rewatching.


Daily: Your Business Artifacts

Every day you create valuable context that should feed your brain:

AI Conversation Artifacts

# After a productive Claude conversation
"Save this conversation summary to context/business/conversations/YYYYMMDD-topic.md"

What to capture:

  • Solutions you figured out
  • Decisions you made
  • Skills you improved
  • Problems you solved

Why: When you improve a skill or solve a problem, your brain gets smarter. Next time, it knows what worked.


Customer & Client Notes

# After a client meeting
"Create meeting notes for [client name] in context/customers/[client-name]/YYYYMMDD-meeting.md"

What to capture:

  • Client goals and challenges
  • What they care about
  • Successful approaches
  • What didn't work

Why: Next time you work with this client, your brain remembers everything. No re-learning.


Project Context

# Save to context/projects/project-name/
- Strategy documents
- Decision rationale
- Lessons learned
- What worked / didn't work

Why: Your brain builds institutional memory. Future projects benefit from past experience.


Personal Observations

# Quick captures throughout the day
echo "Noticed: AI agents work better with explicit constraints" > !inbox/observation-$(date +%Y%m%d-%H%M).md

Why: These become insights you can search and build on.


The Self-Improving Loop

Here's the magic:

  1. You work with Claude → Create artifacts (code, docs, solutions)
  2. You save artifacts to brain → Context/business/, context/projects/, context/customers/
  3. You improve skills → .claude/skills/ get better with each iteration
  4. Your brain gets smarter → Next conversation starts from higher baseline
  5. Repeat → Compounding intelligence

Example:

  • Week 1: Solve customer objection → Save approach to context/customers/
  • Week 2: Different customer, same objection → Brain remembers solution
  • Week 3: Refine approach → Update context
  • Week 4: Teach someone else → Pull from brain's knowledge

Your brain isn't just reading newsletters. It's learning from your work.


Weekly: Index Review (Optional But Powerful)

Every week or two, scan your indexes:

# List your index files
ls context/ideas/

# Read one
cat context/ideas/ethan.md

What you'll see:

  • Accumulated insights over time
  • Patterns emerging across issues
  • Key themes from each source

Why this matters: This is where you spot the connections. Where multiple sources say the same thing independently, that's signal.


One-Time Setup (10 minutes)

To make daily checks automatic, you need Gmail API access. One-time setup:

Step 1: Enable Gmail API

  1. Go to Google Cloud Console
  2. Create new project (or use existing)
  3. Enable Gmail API for the project
  4. Create OAuth 2.0 credentials
  5. Download credentials as credentials.json

Step 2: Place Credentials

# Move credentials to your home directory
mv ~/Downloads/credentials.json ~/google-credentials.json

Step 3: Create Gmail Label

  1. Open Gmail
  2. Create label "8020brain"
  3. Create filters to auto-label your newsletters

Example filter:

Step 4: Test It

# First run will ask for authorization
"Check newsletters"

# Follow the OAuth flow in browser
# Grant access to Gmail

# Check it worked
ls research/newsletters/

You should see: Folders for each newsletter source with markdown files inside.


What to Capture

Not all content is worth indexing. Be selective.

✅ Capture These

Newsletters you reference later

  • Strategic thinking (Ethan Mollick, Ben Thompson)
  • Technical depth (Simon Willison)
  • Market analysis (Azeem Azhar)
  • Your industry experts

YouTube videos with insights

  • Deep dives, not news
  • Interviews with experts
  • Talks and presentations
  • Tutorial concepts (not step-by-step)

Podcasts (via transcript)

  • Convert audio to text (Whisper)
  • Save transcript as markdown
  • Same workflow as YouTube

Your own observations

  • Meeting insights
  • Client patterns
  • Ideas worth developing
  • "Note to self" moments

❌ Skip These

Promotional emails

  • Product launches
  • Sales pitches
  • Event invitations
  • Unless it teaches strategy

News headlines

  • "X company releases Y"
  • Surface-level coverage
  • No analysis or insight
  • Google is better for facts

Tutorials you won't revisit

  • Step-by-step how-tos
  • Specific tool instructions
  • One-time fixes
  • Keep these bookmarked instead

Content you won't reference

  • If you wouldn't Google it later, don't capture it
  • Your brain is for insight, not trivia

Pro Tips

Quality Over Quantity

"5 great sources beats 50 mediocre ones."

Start with 3-5 core newsletters. Get those working well. Add more when you have capacity.

Your brain is most valuable when it contains high signal-to-noise content.


If You Wouldn't Google It, Don't Capture It

Good test: "Would I want to find this again in 3 months?"

  • Yes → Capture it
  • No → Read and move on

Your Brain is a Research Assistant, Not an Archive

You're not trying to save everything. You're trying to remember the insights worth acting on.

Archive goal: News site (capture all events) Brain goal: Strategic advisor (capture valuable thinking)

Be the strategic advisor.


Batch Newsletter Labeling

Don't label newsletters one-by-one. Set up filters.

One-time setup:

  • 10 minutes to create filters for all your sources
  • Never think about labeling again
  • Daily check just works

How to batch create filters:

  1. In Gmail, search: from:ethan@oneusefulthing.org
  2. Click "Create filter"
  3. Apply label: 8020brain
  4. Check "Also apply filter to matching conversations"
  5. Repeat for each source

Now all past + future emails auto-labeled.


Review Indexes Weekly

Every Monday (or your planning day):

# Quick scan of what came in
ls -lt research/newsletters/*/

# Read index updates
cat context/ideas/ethan.md | tail -50
cat context/ideas/simon.md | tail -50

Takes 5 minutes. Keeps you connected to what your brain knows.

Bonus: You'll spot patterns and make connections in real-time.


Common Workflows

Add a New Newsletter Source

# Create folder for new source
mkdir -p research/newsletters/new-source

# Create index file
touch context/ideas/new-source.md

# Add header to index
echo "# New Source Name\n\nInsights from New Source newsletter.\n\n" > context/ideas/new-source.md

# Create Gmail filter for new source
# (Do this in Gmail UI)

# Test it
"Check newsletters"

Capture a Thread or Article

Don't have a clean newsletter for something valuable? Save it manually:

# Create markdown file
touch research/newsletters/one-offs/2025-11-07-valuable-article.md

# Paste content
# Add your summary at the top
# Update one-offs index

Or just ask Claude:

"Save this article to my brain and summarize the key insights: [paste URL or text]"

Backfill Historical Content

Just got your brain set up? Want to add past newsletters?

# Create filter with "Also apply to matching conversations"
# This labels all past emails matching pattern

# Run newsletter check
"Check newsletters"

# It will fetch ALL labeled emails (may take a few minutes first time)

Warning: Fetching 100+ newsletters will take time. Consider:

  • Start fresh (only new ones going forward)
  • Or backfill top 10-20 most valuable issues

Your call. More data isn't always better data.


What Can Go Wrong

Gmail API Not Working

Symptoms:

  • "Failed to fetch newsletters"
  • "Authentication error"
  • No newsletters appearing

Fixes:

  • Check ~/google-credentials.json exists
  • Re-run OAuth flow (delete ~/.credentials/gmail-token.json and try again)
  • Verify Gmail API is enabled in Google Cloud Console
  • Check you labeled some emails "8020brain"

See Troubleshooting: Gmail API for details.


Too Many Newsletters

Symptoms:

  • Takes forever to process
  • Brain feels cluttered
  • Hard to find specific insights

Fixes:

  • Unsubscribe from low-value sources
  • Archive sources you're not actively using
  • Focus on 5-10 core sources max

Quality over quantity. Always.


Summaries Not Generating

Symptoms:

  • Newsletter saved but no summary
  • Index not updating
  • Manual summary needed

Fixes:

  • Check AI API key configured (Anthropic/OpenAI)
  • Verify newsletter-daily-check skill has API access
  • Check logs for errors
  • Try manual: "Summarize research/newsletters/source/2025-11-07-issue.md and update index"

Your Brain Needs Regular Feeding

Daily: Check newsletters (5 min) Weekly: Review indexes (5 min) Monthly: Prune low-value sources (15 min)

Make it routine. Like checking email or Slack.

The magic happens when you:

  1. Capture valuable content consistently
  2. Let AI extract insights automatically
  3. Query when you need answers

Input → Process → Query → Insight

That's the loop.


Next: Get answers from all this knowledge


Quick Reference

Daily check:

"Check newsletters"

Fetch YouTube:

/fetch https://youtube.com/watch?v=...

Add new source:

mkdir -p research/newsletters/new-source
touch context/ideas/new-source.md
# Then create Gmail filter

Setup Gmail API:

  1. Google Cloud Console → Enable Gmail API
  2. Create OAuth credentials → Download
  3. Save as ~/google-credentials.json
  4. Create "8020brain" label in Gmail
  5. Test: "Check newsletters"

What to capture:

  • Strategic thinking
  • Technical insights
  • Your observations
  • Content you'd want to find in 3 months

What to skip:

  • Promotional emails
  • News headlines without analysis
  • One-time tutorials
  • Anything you won't reference