Access Granted

    5-Day Second Brain
    Challenge Hub

    A Second Brain is an AI context layer that knows your work, so your AI stops starting from zero. In five days you build the smallest useful version: context, memory, workflow, one work signal, and a daily overview.

    Quick Start

    1. 1. If you are unsure, choose ChatGPT Projects. Claude.ai web is the second easiest option.
    2. 2. Do not connect tools on Day 1. Copy/paste your context and one small signal manually first.
    3. 3. Do one day at a time. Each task should take 15-30 minutes.
    4. 4. Use the copy/paste prompts as starting points, then adapt to your real project.
    5. 5. Do not skip the acceptance checks. They tell you whether the setup actually works.
    Why Bother

    What You'll Be Able To Do by Day 5

    Most people use AI like a search box and start from scratch every time. By Friday you'll have an AI setup that knows your work and does real things for you. No coding required.

    Your morning brief

    Type one command and get today's priorities, pulled from your own projects and notes, not generic advice.

    Real work, finished

    Turn a messy task into a done draft with a built-in quality check, in minutes instead of hours.

    It remembers you

    Stop re-explaining yourself. Your AI keeps your context, goals, and decisions between sessions.

    Your weekly compass

    Ask what to focus on this week and get a clear, prioritized answer built from your own work.

    Content in your voice

    Teach it your brand voice and guidelines once. Then posts, emails, and slide outlines come back sounding like you, not generic AI.

    The full-system example

    My own /overview command pulls active projects, yesterday's activity, meetings, replies, pipeline, content queue, outreach, tasks, blockers, and business metrics. /overview-slides turns the same briefing into a dashboard-style visual deck.

    Before You Start

    Choose Your Work Surface

    Start with one surface and follow the full steps. The challenge is tool-flexible, but your setup should not be vague: each route below tells you exactly where the context goes, what to ask first, and how to know it worked.

    Beginner default: ChatGPT Projects. Claude user default: Claude.ai web. File-based default: Codex app. Power default: Codex CLI or Claude Code CLI. Gemini and Antigravity remain explicit test routes until their launch issues are complete.

    Beginner translation

    Words That Look Scarier Than They Are

    If stuck: use ChatGPT Projects
    Work surface: the app you choose for the challenge. If you are unsure, choose ChatGPT Projects.
    Context: the short explanation of who you are, what you are working on, and what good output looks like.
    Memory: reusable notes the AI should remember next time, such as goals, decisions, projects, and examples.
    Signal: one current input from your real work, such as today's calendar, task list, inbox notes, or CRM notes.
    Connector or MCP: a way for the AI tool to read another app. Skip this at first; paste text manually until the basic setup works.
    AGENTS.md / CLAUDE.md / GEMINI.md: plain text instruction files for local-file tools. App-first users do not need these files.

    Setup router

    Pick one route. Unsure? Use ChatGPT Projects.

    10 routes

    No terminalOpenAI web / desktop10-15 min

    ChatGPT Projects

    Fastest beginner route when you want persistent project instructions, uploaded reference files, and a clean chat workspace.

    Beginner Translation

    • Start here if you are unsure. You do not need a terminal, code editor, or MCP setup.
    • Create one Project, paste your Day 0 context into Project instructions, then use the first prompt below in a new Project chat.
    • Skip connectors on the first pass. Paste calendar/tasks/notes manually until the daily overview is useful.

    Step By Step

    1. 1Open ChatGPT and create a new Project named Second Brain - [work area].
    2. 2Add project instructions: tell ChatGPT to read your context first, ask before assuming, keep outputs concrete, and end serious work with an acceptance check.
    3. 3Upload or paste your context note plus any small reference files that should shape every answer.
    4. 4Start a fresh chat inside the Project and ask it to summarize your role, current focus, constraints, and quality bar.
    5. 5Correct the summary until it sounds like your real work. Save the corrected version back into the Project instructions or context file.
    6. 6Save the Day 5 daily overview prompt in the Project so you can rerun it without rebuilding the setup.

    Copy This After Setup

    Finish the steps above first, then paste this into your chosen tool. If a word does not apply, delete it.

    First prompt for ChatGPT Projects
    You are my Second Brain operator for the 5-Day Challenge.

Use the context, project knowledge, and connected sources I explicitly provide. If a connector is unavailable, ask me to paste or export that source instead of guessing.

First, summarize:
- who I am and what work area this Project covers
- what matters this week
- what good output looks like
- what you still need before you can build a reliable daily overview

Then build today's mini /overview:
1. date and current work context
2. today's schedule or focus blocks
3. active projects and stale/waiting items
4. one work signal I provide or connect
5. memory updates worth saving
6. blockers and risks
7. top 3 actions with why each matters

End with: "Save to memory" bullets that I can paste back into Project knowledge.

    Optional: Connect Later

    • Beginner move: skip connectors on Day 1 and paste your calendar/tasks/notes manually.
    • Google Drive/Docs: connect through ChatGPT Apps/Connectors if your plan/workspace exposes it; use it for goals, project docs, examples, and operating notes.
    • Gmail + Calendar: connect only after the text version works; ask for today's meetings, unanswered important emails, and follow-ups, not a full inbox dump.
    • GitHub/Linear/Jira: connect if your daily overview needs issues, PRs, or roadmap work; otherwise paste a tasks export.
    • Rule: use connectors read-first. Do not ask ChatGPT to send, delete, invite, or update external systems during the challenge.

    What To Save As Memory

    • Project instructions: role, work context, quality bar, privacy rules, and how to format the daily overview.
    • Project knowledge: profile, goals, active projects, decisions, examples, and voice/style notes.
    • Signal notes: one pasted/exported source per day, such as calendar, CRM, tasks, emails, or meeting notes.
    • Memory update rule: after each session, ask what should be saved and paste only durable items back into project knowledge.

    Final Brief Checklist

    • Header: today's date, current work context, and one-line status.
    • Schedule: meetings or focus blocks for today; if no calendar connector, use pasted calendar bullets.
    • Active work: projects, tasks, deadlines, stale items, and anything waiting on someone else.
    • One signal source: inbox, CRM, project board, docs, content queue, GitHub/Linear, or calendar.
    • Memory growth: new decisions, reusable facts, examples, and missing context to save.
    • Risks and blockers: overdue work, unanswered messages, unclear ownership, or bad assumptions.
    • Top 3 actions: ranked, concrete, and tied to why each matters today.
    • Optional slide mode: return an 8-card briefing with title, critical alerts, schedule, pipeline/tasks, projects, inbox/brain health, memory, and top 3.

    Research note

    These steps are grounded in the official setup docs linked inside each guide. The separate launch tests still need to confirm which routes become the recommended defaults for real participants.

    Daily Build Notes

    What Each Day Actually Does

    Most AI setups fail because they begin with a toy prompt. Your first move is choosing one real context: a project, offer, client, team, content system, or operating area.

    Beginner version: pick one thing you actually need help with this week. Do not try to build a Second Brain for your whole life.

    The output of Day 0 should be a usable context note. In ChatGPT or Claude, paste it into Project instructions. In file-based tools, save it as context.md or the tool's instruction file.

    Day 0 context builder prompt
    Help me turn this work area into a Second Brain context note.

Return:
1. role and work area
2. current focus
3. active projects
4. quality bar
5. privacy rules
6. likely signal sources
7. missing context questions

Use this work area: [sales / client delivery / product / content / ops / current project].
    Beginner example
    I am a founder running a 5-day launch.
This setup is only for the launch, not my entire business.
This week I care about: emails, landing page fixes, participant questions, and daily priorities.
Good output is short, practical, and ready to paste or act on.
The first signal source will be: today's task list.
    context.md starter
    # Context
I am [role]. I am building/operating [work area].

# Current Focus
- [Project or active responsibility]
- [Deadline or priority]
- [What matters this week]

# Quality Bar
Good work is specific, concise, accurate, directly usable, and tied to my real constraints.

# Privacy Rules
- Do not expose private customer, employee, financial, or health data unless I explicitly paste it for this task.
- Ask before using connected sources for anything beyond reading and summarizing.
- Never send, delete, invite, charge, or update external systems without explicit approval.

# Daily Overview Target
Every overview should include schedule, active work, one signal source, risks, memory updates, and top 3 actions.

# Candidate Signal Sources
- Calendar:
- Tasks/project board:
- Docs/Drive/Notion:
- Inbox/Slack:
- CRM/database:
- GitHub/Linear/Jira:

# Rules
- Ask before assuming missing business context.
- Prefer small shipped steps over big vague plans.
- End serious work with an acceptance check.

    Go deeper: Context Engineering guide →

    Full Challenge

    Copy the Whole Challenge

    Use this as your working checklist

    Paste it into your repo, notes app, or project board. Complete the acceptance check before moving to the next day.

    5-day-second-brain-challenge.md
    # 5-Day Second Brain Challenge

Build a working AI daily briefing in 5 days.

Beginner default:
- If you are unsure, choose ChatGPT Projects.
- If you already prefer Claude, choose Claude.ai web Project.
- Do not connect tools on Day 1. Paste text manually first.

Pick one work surface before Day 1:
- ChatGPT Projects: no-terminal route for project instructions, uploaded files, and optional Apps/Connectors.
- Claude.ai web Project: no-terminal route for Claude project knowledge, custom instructions, and web connectors.
- Claude Desktop / Cowork: desktop route for Claude chat, local MCP, and Cowork local-folder tasks.
- Claude Code Desktop: Claude Code app route with local folders, visual diffs, sessions, and optional MCP.
- Codex app: OpenAI desktop route for local files, AGENTS.md, reviewable diffs, and optional MCP.
- VS Code / Cursor: editor route for file context, selections, diagnostics, git diff, rules, and MCP.
- Codex CLI: terminal route for OpenAI-powered local file edits, AGENTS.md, and codex mcp.
- Claude Code CLI: terminal route for Claude-powered repo work, CLAUDE.md, /mcp, and claude mcp.
- Gemini CLI: terminal route for Google/Gemini testing with GEMINI.md and gemini mcp.
- Antigravity: Google agent route; test desktop, IDE, or CLI separately.

Do not switch surfaces mid-challenge unless the first one is blocked.

Plain-language glossary:
- Context: who you are, what you are working on, and what good output looks like.
- Memory: reusable notes the AI should remember next time.
- Signal: one current input from real work, such as calendar, tasks, inbox notes, or CRM notes.
- Connector/MCP: a way for the AI to read another app. Skip this until the pasted version works.

Optional starter memory structure for file-based routes:
mkdir -p memory signals
touch AGENTS.md context.md daily-overview.md
touch memory/profile.md memory/goals.md memory/projects.md memory/decisions.md memory/examples.md
touch signals/calendar.md signals/tasks.md signals/inbox.md

For Claude Code use CLAUDE.md instead of AGENTS.md.
For Gemini CLI use GEMINI.md instead of AGENTS.md.
For ChatGPT or Claude web, put the same content into project instructions and project knowledge.

Optional connector / MCP priority after the basic setup works:
1. Calendar: meetings, deadlines, focus blocks.
2. Tasks or project board: assigned/open/stale/waiting work.
3. Docs or Drive or Notion: strategy, examples, operating notes.
4. Inbox or Slack: unanswered important messages and follow-ups.
5. CRM or database: customers, pipeline, metrics, read-only first.

Safety rule:
Connect one source first. Read-only first. Never ask the AI to send, delete, invite, update records, or operate on a full inbox/database during the challenge unless you explicitly approve that exact action.

Day 0: Pick one real work context
- Choose one area: sales, client delivery, product, content, ops, fundraising, or a current client/project.
- Write context, current focus, quality bar, constraints, privacy rules, and one signal source you may connect later.

Day 0 prompt:
"Help me turn this work area into a Second Brain context note. Return:
1. role and work area
2. current focus
3. active projects
4. quality bar
5. privacy rules
6. likely signal sources
7. missing context questions"

Acceptance check:
- Your AI can answer: "What am I working on, what matters now, and what does good look like?"

Day 1: Teach AI the context
- Put your context where your chosen tool reads it.
- App-first route: paste it into Project instructions or Project knowledge.
- File-based route: put it in AGENTS.md, CLAUDE.md, GEMINI.md, rules file, or context.md.
- Ask the AI to summarize the context back to you and name missing context before doing work.
- Save the corrected summary back into memory.

Day 1 prompt:
"Read my context and memory before helping. Summarize:
1. who I am
2. what work area this setup covers
3. what matters this week
4. what good output looks like
5. what you are not allowed to assume
6. what context is missing
Then build a mini /overview for today using only available context."

Acceptance check:
- The next AI answer is clearly more useful than a blank chat.

Day 2: Run the work loop
- Pick one messy open loop.
- Ask for plan, acceptance criteria, smallest useful action, review, and memory update.

Day 2 prompt:
"Help me finish [task].
First create:
1. objective
2. plan
3. acceptance criteria
4. missing context
Then complete the smallest useful step.
Before calling it done, review the output for wrong assumptions, missing source material, privacy risk, and quality issues.
End with memory updates worth saving."

Acceptance check:
- You have a finished artifact plus a short review note.

Day 3: Add memory
- Add only reusable context that should change future answers.
- Store profile, goals, projects, decisions, examples, signals, and missing context separately.

Day 3 prompt:
"Extract durable memory from this session.
Return only:
- profile
- goals
- projects
- decisions
- examples
- signals
- missing context
Exclude temporary chatter. Mark uncertainty."

Acceptance check:
- Your AI can continue without you repeating the basics.

Day 4: Add one work signal
- Pick one source of truth: calendar, tasks, docs, Drive, Notion, inbox, Slack, CRM, database, content queue, GitHub, Linear, or Jira.
- Use a connector/MCP only when you would otherwise paste the same data repeatedly.
- If no connector is available, use a pasted/exported signal note.
- Beginner version: paste 5-10 bullet points from that source into the chat.

Day 4 prompt:
"Build a safe signal note for [source].
Include:
1. connector/MCP/export path
2. exact fields to read
3. what to ignore
4. privacy rules
5. update rhythm
6. failure fallback
7. how it appears in the daily overview"

Acceptance check:
- Your Second Brain knows where one real input comes from and how to use it safely.

Day 5: Build the daily overview
- Read context, memory, current tasks, and one signal source.
- Return a daily command you can rerun tomorrow.

Day 5 prompt:
"Build today's /overview.
Include:
1. date, current work context, and one-line status
2. schedule and focus blocks
3. active projects, tasks, deadlines, stale items, and waiting items
4. one signal source and what changed
5. memory updates worth saving
6. risks and blockers
7. top 3 actions ranked by importance
For every important claim, say which context, memory file, or connected source it came from."

Optional slide mode:
"Turn the overview into an 8-card briefing:
1. title + date + 3 status stats
2. critical alerts
3. schedule + focus blocks
4. projects/tasks/pipeline
5. inbox or signal source
6. memory growth
7. blockers
8. top 3 actions"

Acceptance check:
- You have a daily briefing you can run again tomorrow.

Day 6: Decide the path
- Keep DIY if the starter system already helps.
- Use the full Second Brain if you want memory, skills, agents, workflows, and quality checks wired around your real work.
- If you receive a personalized preview, treat it as a sketch of what your full system could become.

    What To Do After Day 5

    If the challenge helped, you have two paths: keep building DIY, or get the full Second Brain system wired around your real work.

    Second Brain 2.0

    The complete Second Brain system: memory, skills, agents, workflows, quality checks, and setup support.

    DIY $197 · Kickstart $597 · DWY $2,497

    See Packages
    Newsletter

    Get practical notes on AI workflows, launches, and Second Brain improvements.

    For operators building with AI in public.

    Read More