How to Use ChatGPT More Effectively Every Day

TL;DR

Set custom instructions and enable memory so ChatGPT can tailor responses to your work, goals, and communication preferences. For important decisions, explicitly ask it to challenge your assumptions, identify blind spots, and present the strongest case against your position. Match specialized modes to suitable tasks, and double-check information whenever the conversation appears to lose context.

Transcript

So, look. If you're happy doing the same boring, repetitive tasks every day for the rest of your life, well, then you can click off right now because the next 30 minutes are for people who want out of that. You see, most people use ChatGPT to plan their groceries or to be their therapist after 5 hours of doomscrolling on TikTok wondering why life i... Read More

Key Insights

• Custom instructions are persistent guidance that tells ChatGPT who the user is, what work they do, what they want to accomplish, and how responses should be written. More specific preferences about length, tone, detail, and unwanted behavior can improve the relevance of future outputs.
• Custom instructions are available through Settings and Personalization. The user can enter both background context and response preferences there, creating a reusable brief that remains active until it is manually changed instead of rebuilding the same context whenever a new chat begins.
• Memory is an evolving record that ChatGPT picks up through continued use, including conversations, projects, stated preferences, and corrections. Unlike fixed custom instructions, it can reflect what the system learns after the initial setup and become more useful as interactions accumulate.
• Memory controls are located under Personalization near custom instructions. The presenter advises switching memory on, reviewing the information through Manage Memories, and manually editing incorrect details so the retained context continues to represent the user accurately.
• Devil's advocate prompting is a method for reducing agreeable, validation-focused answers. Explicit instructions such as asking for blind spots, challenges to current thinking, or the strongest case that an assumption is wrong can produce a more balanced basis for decisions.
• ChatGPT can provide incorrect information with complete confidence, according to the presenter. Users therefore need to remain in control, apply their own judgment, and double-check answers when the system appears to lose track of instructions or conversational context.
• A first-name response instruction is presented as a personal warning system for context loss. The presenter asks ChatGPT to use his first name, then treats omission of the expected form of address as a signal to stop and verify the information provided.
• ChatGPT includes multiple modes intended for different purposes. The transcript identifies voice, canvas, study, agent, image, and deep research modes, while emphasizing that users should understand which mode suits each task instead of relying on the standard mode for everything.

Questions & Answers

Q: How do I set up custom instructions in ChatGPT?

Open Settings, select Personalization, and find Custom Instructions. Add information that helps ChatGPT understand who you are, what you do for work, and what you are trying to accomplish. Then describe how you want answers written, including preferred length, level of explanation, tone, and behaviors to avoid. The presenter recommends being as specific as possible because these instructions remain available across future chats until you change them.

Q: What should I include in ChatGPT custom instructions?

Include stable context that will remain useful in later conversations. The transcript specifically recommends stating who you are, what you do for work, what you are trying to get done, and any other information that clarifies your situation. You should also specify whether you prefer short or detailed answers, what tone you want, and what the assistant should avoid. Together, these details function like a one-page working brief.

Q: What is the difference between custom instructions and memory?

Custom instructions contain information that you deliberately provide once and that remains in place until you edit it. They describe your background, needs, and preferred way of working. Memory develops over time from conversations, projects, corrections, and expressed preferences. The presenter compares custom instructions to a first-day brief, while memory represents what a working partner learns through continued interaction after that initial briefing.

Q: How do I enable and manage ChatGPT memory?

Go to Settings and then Personalization, where Memory appears underneath Custom Instructions according to the transcript. Make sure the memory setting is switched on. If you want to inspect what ChatGPT has retained, open Manage Memories. The presenter says the stored information is visible there and can be manually edited when something is wrong, allowing you to correct the context that ChatGPT may use later.

Q: How can I make ChatGPT challenge my assumptions?

Tell ChatGPT explicitly not to agree automatically and ask it to play devil's advocate. Useful instructions from the transcript include asking what blind spots you have, what you may be missing, and what the strongest possible case is for your position being wrong. This changes the goal from receiving validation to testing an idea, helping you consider opposing arguments before reaching a more well-rounded conclusion.

Q: Why should I ask ChatGPT to play devil's advocate?

The presenter argues that ChatGPT can appear overly agreeable, which may encourage users to trust answers without applying enough scrutiny. Asking it to play devil's advocate creates pressure to identify weaknesses, missing context, and opposing arguments. The purpose is not to hand control to the system, but to give the user more perspectives to examine while remaining responsible for steering the discussion and making the final decision.

Q: How can I notice when ChatGPT may be losing context?

The presenter uses a personal consistency check at the beginning of each chat. He instructs ChatGPT to address him by his first name and watches whether later answers continue following that instruction. If the expected name is omitted, he treats it as a sign that the conversation may be losing direction. At that point, he stops and double-checks the information already supplied rather than assuming it remains reliable.

Q: Which ChatGPT modes does the transcript recommend knowing?

The transcript identifies six modes that users should understand: voice mode, canvas mode, study mode, agent mode, image mode, and deep research mode. It presents these as modes serving different purposes and argues that using the standard mode for every task limits what users can accomplish. However, the supplied transcript ends before explaining the individual functions or recommended use cases for each listed mode.

Summary & Key Takeaways

• ChatGPT produces more relevant responses when it receives persistent context about the user. Custom instructions can specify your identity, work, objectives, preferred response length, desired tone, and unwanted behaviors. The presenter recommends completing this initial setup once so future chats begin with a useful personal and professional brief.

• Memory serves a different purpose from custom instructions. Custom instructions remain fixed until the user changes them, while memory captures useful information from conversations, projects, preferences, and corrections over time. Users can enable memory in personalization settings, inspect stored details through Manage Memories, and manually correct information that is inaccurate.

• Critical use requires more than accepting agreeable answers. The presenter recommends asking ChatGPT to play devil's advocate, expose blind spots, challenge existing assumptions, and build the strongest case for why an idea may be wrong. Users should remain responsible for decisions because ChatGPT can provide incorrect information with complete confidence.