How to Build a Second Brain With ChatGPT

To build a second brain with ChatGPT, create a project for one continuing goal, add a small set of reliable sources, and write instructions for how answers should use them. Review useful results and save a short current-state note so future work starts from something you have checked.

Keep the original documents and your approved conclusions distinguishable. fenbs can hold the project decisions and progress that connected assistants need across AI apps, while ChatGPT remains the place where you explore the material. Start with one real project rather than uploading your whole digital life.

What should you set up first?

1. Choose one outcome, such as preparing a workshop or researching a product change. Write what a useful result would look like.
2. Create a ChatGPT project and give it a name that identifies that outcome. Put continuing work about the same outcome there.
3. Add the few sources needed for your first question. Include an index explaining what each source is and which version is current.
4. Add project instructions that separate source evidence, your conclusions, and unresolved questions.
5. Ask a specific question, check the answer against its sources, and save only the useful result and next step.

OpenAI explains that ChatGPT projects↗ (opens in a new tab) bring related chats, uploaded files, sources, and instructions together. Project instructions apply across its chats. Check the current documentation for the controls available in your account instead of building your system around an assumed file allowance.

A small first project makes mistakes visible. If the workshop covers accessible slide design, begin with the course outline and an approved checklist. You should be able to tell whether an answer used the checklist before introducing several years of presentation notes.

Which material belongs in the project?

Choose material because it can change an answer. A current requirement, a primary reference, and an example of an acceptable output usually help more than a large collection of vaguely related articles. Leave unrelated interests in their own notes system.

Give source files recognizable names and include version information where it matters. Mark old material as superseded, or remove it from the working set after preserving your own copy. A file named “final” is not enough when two different documents both claim to be final.

The broader AI second brain guide explains how reusable knowledge fits together. For this setup, focus on the smaller question: which evidence does the next ChatGPT task actually need?

What instructions make the answers easier to check?

Write a short operating agreement for the project. Name the goal, the source rules, the expected output, and the situations that require a question. OpenAI’s prompting guidance↗ (opens in a new tab) recommends supplying the context and boundaries that matter to the result.

Goal: Help prepare a workshop for first-time presenters.
Use the current outline and checked reference files.
Separate supported facts, suggestions, and open questions.
For factual claims, name the source and section you used.
If the supplied material does not answer a question, say so.
Do not turn a suggestion into an approved decision.
End substantial work with what changed and what needs review.

This is an instruction to follow, not proof that every answer will comply. Test it with a question the sources cannot answer. If ChatGPT fills the gap confidently, correct the response and revise the instructions before relying on the system for more consequential work.

Keep stable instructions separate from changing project state. “Use plain language” may remain useful for months. “The organizer has not approved the outline” needs updating as soon as approval arrives. In fenbs, Decisions and rules give connected assistants a place to read those human choices before continuing.

How do you turn a conversation into reusable knowledge?

Start a focused chat for a distinct result, such as comparing two teaching examples. Ask for a short answer tied to the supplied sources, then check the important statements yourself. Save the conclusion in your own words, with the source location and any limit on its use.

For example, a discussion might produce this note: “Use the simpler chart in the introductory exercise because the audience has not covered statistical uncertainty. The advanced example stays in the optional appendix. Organizer review is still needed.” That note distinguishes a proposed teaching choice from a completed approval.

Avoid saving every response as a permanent reference. Keep intermediate drafts when you need their history, but label them clearly. The reusable record should tell you what is reliable now, not require you to reconstruct the entire conversation to find the latest answer.

What should you check before trusting the setup?

Run a short practical test. Ask where a particular requirement came from, which outline is current, what remains unapproved, and what the next task should be. Open the cited sources and compare the answer with your record.

Look for missing exceptions as well as false statements. A summary can sound accurate while leaving out the one condition that changes your choice. If an answer cannot show where a claim came from, treat that claim as unresolved until you verify it.

Upload only information you are allowed to share with the service and account you use. A ChatGPT project is not the original source archive, and another AI app will need its own access to any documents you want it to read. The AI brain overview explains that distinction between an assistant and the record it uses.

How to do this with fenbs

1. Create a free account, free for one person with no credit card, and add the project whose knowledge you want to preserve.
2. Add the approved decisions to Decisions and rules. Add source references to Links and listings, with enough explanation to identify the material.
3. Create tasks for the next concrete work, such as checking a workshop example or getting the outline reviewed.
4. Use the integration guide to connect a supported assistant over MCP, a standard way for an AI app to use an outside tool. Follow the ChatGPT guide for the connection available to your account.
5. Ask the assistant to read the context at the start and save checked progress and the next step before stopping. Share document content separately when the assistant needs it.

fenbs does not import your ChatGPT chats or read their history. If you only need one personal reference notebook and do not hand work between assistants, your existing notes may be enough. Keep fenbs for the active decisions and progress that benefit from a separate record.

How do you keep the system useful as work changes?

Review the working set when a source changes or a milestone is completed. Replace obsolete instructions, label replaced conclusions, and remove sources that no longer help the current goal. Preserve necessary originals in the archive you control.

At the end of a session, write three things: what you checked, what remains uncertain, and where to begin next. When trying another assistant, give it that record and the required sources. The separate Copilot setup guide covers that app’s document workflow; the second-brain topic collects other approaches.