The Most Important ChatGPT Feature You're Not Using

Key Concepts

• Projects: A feature in ChatGPT that allows users to organize chats into folders, upload custom instructions and knowledge bases, and create mini-automations and apps.
• Knowledge Base: A collection of files uploaded to a project that ChatGPT can reference for context and information.
• Custom Instructions: Project-specific instructions that guide ChatGPT's behavior and responses within that project.
• Project-Only Memory: A setting that confines ChatGPT's memory and context to only the chats within a specific project, preventing it from referencing unrelated conversations.
• Canvas: A feature within ChatGPT that allows for collaborative document editing and creation.
• Deep Research: A tool within ChatGPT that conducts in-depth research on a given topic, compiling information from multiple sources.
• Make: An automation platform that integrates with ChatGPT to create AI agents and automate complex tasks.
• OpenAI Playground: A testing environment for OpenAI models, including a prompt enhancer tool.

Projects: The Underutilized Powerhouse of ChatGPT

This video highlights the significant, yet often overlooked, "Projects" feature in ChatGPT, emphasizing its potential to revolutionize user experience, especially now that it's available on the free plan. Projects serve as organizational folders for multiple chats, but their true power lies in the ability to upload custom instructions and knowledge bases that apply to all chats within a project. This enables enhanced chatting, reduced hallucinations, and the creation of mini-automations and apps.

Setting Up a Project

1. Location: Projects are found in the left-hand panel, below "GPTs" and above "Chats."
2. Creation: Click "New Project" and assign a name. Default categories like "Investing," "Homework," "Writing," "Health," and "Travel" are provided, or users can create custom ones.
3. Memory Setting: A crucial configuration is the "Memory" setting, which can be set to "Access memories from outside chats and vice versa" (default) or "Project only." The "Project only" setting creates a self-contained environment, referencing only chats within that project, which the presenter prefers for most projects to maintain organization and context without overwhelming the model. This contrasts with the older method of using one long chat for a topic, which becomes difficult to navigate.

Project Features: Knowledge Base and Custom Instructions

Within a project, standard ChatGPT tools (file uploads, deep research, image creation, web search, canvas) are available. However, two key differentiators are the Knowledge Base and Custom Instructions.

Knowledge Base

• Functionality: The "Add Files" button within a project uploads documents to a knowledge base that ChatGPT can reference exclusively within that project.
• Use Cases: This can include SOPs, checklists, sales playbooks, roadmaps, design specs, definitions, brand kits, style guides, contracts, and security policies.
• Example: The presenter uploads two previous YouTube scripts to establish a writing style.
• File Limits: Free plan: 5 files; Plus plan: 25 files; Pro plan: 40 files. A workaround for the free plan is to combine multiple files into one, as the token limit (2 million) is substantial.

Custom Instructions

• Functionality: Similar to account-level custom instructions, but specific to a project. This is presented as a superior alternative to account-level instructions, which are often too broad for diverse use cases.
• Benefits: Greatly improves the user experience by providing context-specific guidance.
• Mini-Automations: Custom instructions can be used to set up automated workflows. An example is a short-form content creation system where users input a topic, and ChatGPT generates hook ideas, refines them, and guides the user through subsequent steps to create a full script.

Integrating with Automation Platforms (Make)

For more advanced automations and AI agents that can complete entire processes autonomously, platforms like Make are recommended. Make allows for the creation of "digital employees" through a no-code, drag-and-drop interface.

• Capabilities: Make integrates with thousands of apps and APIs, enabling complex automations.
• Example: The presenter demonstrates a personal assistant agent created on Make that can read emails, research topics, curate news, create content, and interact via Slack. This agent can add meetings to a calendar and generate content and images based on an article link.
• Grid Feature: Make's grid feature provides a visual overview of automation connections and data flow, aiding in troubleshooting.
• Sponsorship: Make is sponsoring this video, offering a one-month free trial of their Pro plan with 10,000 operations per month via a link in the description.

Crafting Effective Custom Instructions

The presenter details how to create effective custom instructions for a project, using a YouTube scriptwriting assistant as an example.

• Key Components: Goal, context, response format, tone, and how to use the knowledge base.
• Prompt Enhancer (OpenAI Playground):
  • Access: platform.openai.com (requires an OpenAI account and API credits, minimum $5 upload).
  • Process: Enter basic instructions in the "Chat" tab, send, and the tool generates an enhanced version in the "developer message" box.
  • Optimization: The "Optimize" button further refines the instructions based on best practices, adding clear section headers and explanatory notes.
  • Output: Copy the enhanced instructions and paste them into the project's custom instructions.
• Free Alternative (Explore Tab):
  • Method: Navigate to the "Explore" tab of custom GPTs and search for "custom instructions."
  • Tool: The "Skillip AI" custom instruction tool is recommended as a free alternative to the OpenAI Playground for enhancing prompts.

Practical Use Cases and Examples

YouTube Scriptwriting Assistant Project

• Setup: Project named "Writing," two previous scripts uploaded to the knowledge base, and custom instructions defining the role, style, tone, and banned words.
• Interaction: A new chat is initiated, and the user requests an engaging intro for a script about ChatGPT projects.
• Result: ChatGPT generates an intro that matches the presenter's conversational tone and style, referencing the uploaded scripts. This eliminates the need for repetitive prompt engineering.
• Canvas Integration: The presenter demonstrates using the "Canvas" feature within the project to write the next part of the script, showing collaborative editing and refinement.

Health-Related Project: Interactive Quiz

• Scenario: Using a project for random health topics, the presenter uses "Canvas" to create an interactive quiz based on a previous chat.
• Outcome: ChatGPT generates code for a quiz with a user-friendly interface, allowing users to check answers and receive explanations. This showcases Canvas's ability to create simple interactive applications.

Fitness Project: Workout Tracker App

• Scenario: Within a fitness project, the presenter asks ChatGPT to build a workout tracker.
• Process: Initial attempt requires refinement through natural language editing. The improved version allows users to log exercises, sets, and track progress, with the history being saved within the project. This illustrates the potential for creating mini-apps tailored to specific needs.

Short-Form Content Creation Automation

• Setup: A project with a document of hook templates in the knowledge base and custom instructions outlining an eight-step collaborative process for content creation.
• Workflow: The user provides an article or concept, and ChatGPT guides them through generating hooks, refining them, selecting visuals, outlining content, adding engagement elements, structuring the script, and final review.
• Demo: A quick demonstration shows the process of selecting hook ideas, visual hooks, engagement elements, and generating a full script. The presenter also shows how to create an image for the visual anchor using the built-in image generation feature.

Additional Tips and Tricks

• Migrating Existing Chats: Drag and drop existing chats into project folders. On mobile, use the three dots and "Move to Projects."
• Using Custom GPTs within Projects: Prefix a custom GPT's name with "@" to call it within any chat in a project, combining project structure with GPT consistency.
• Building Knowledge Bases with Deep Research: Use the "Deep Research" feature to gather extensive information from multiple sources, then add the compiled literature review to a project's knowledge base. This feature acts as a mini-agent, researching for about 10 minutes and potentially finding hundreds of sources.

Conclusion

The presenter reiterates that Projects are a fundamental, yet underutilized, feature of ChatGPT, capable of transforming it into a personal productivity hub. The ability to organize, contextualize, and automate tasks within projects offers significant benefits for life, business, and creativity. The video concludes by promoting Futuripedia, a course platform offering extensive AI education, and a related video on mastering ChatGPT.