Over the last few months, I’ve become more and more interested in how people are leveraging ChatGPT and other LLMs in both their personal and professional lives. After doing some (AI-assisted) research, what follows are a few of the best ways I’ve come across of interacting with ChatGPT to enhance your productivity, to learn things faster and more efficiently, or to simply have some fun. Some of these are rather obvious (like using Copilot for coding), but so critical to my productivity that I include them for the sake of completeness. That said, I’ll mostly try to focus on the creative, non-obvious methods of interaction that work pretty well today.
Coding with ChatGPT
At this point, using ChatGPT for coding is table stakes. If you’re a software engineer, data scientist, hobbyist, or anywhere in between that does anything programming-related, you should probably be utilizing ChatGPT. Most of the people I know in this space do this via GitHub Copilot, which is a plugin for VS Code. The big value-add for the VS Code version is that whatever script you’re working on at the time is passed as context, alongside your query.
I’ve personally found ChatGPT very helpful in the following scenarios:
- In VS Code, I’ve been coding for a while and my brain is fried. I’ll get some error (usually related to time zones or date formats…), paste it into the chat, and out comes a clean, friendly little explanation that I can oftentimes copy/paste directly into my script. This beats Stack Overflow for me because, if I’m really interested, I can usually get a succinct explanation along with the one-liner fix.
- Let’s say I’m writing a new function to parse a set of characters out of a specific, oddly formatted text file and into a dataframe. For this use case I’ll generally switch to the ChatGPT web app and paste in a snippet of the text file, explain my intentions, and ask for a Python function. I like using the web app for this type of self-contained, straightforward task because I know exactly what the input prompt is, and that ChatGPT won’t be potentially confused by the added context of the rest of my script that’s in my VS Code workspace.
- Similar to #2 but for SQL queries in particular. I can describe what question I have about the data in natural language and provide context about the handful of tables that might be required for the query. With that, ChatGPT can give me a solid SELECT statement that, with minor tweaks, answers my question.
Creating Your Personal Mentor
With Custom GPTs, you can create your own virtual mentor that you can instruct to emulate someone you admire (which is easier if they’re sufficiently famous and public enough to be mentioned multiple times in the training corpus). I happen to be a massive fan of Tim Ferriss, having listened to hundreds of his podcast episodes over the past 10 years, so I thought it would be a worthwhile challenge to create a custom GPT that will give me advice informed by the teachings of Tim and his many incredible guests.
Ultimately, I wanted to make a virtual mentor that I could come to for advice about life, finances, relationships, purpose, health, wealth, philosophy, and more.
I downloaded 20+ books off the internet that were either written by Tim himself (e.g. The 4-Hour Workweek, Tools of Titans), written by his guests (e.g. Deep Work by Cal Newport), or cited on the show as recommendations or foundational books in any of the aforementioned areas (e.g. The Almanack of Naval Ravikant, The Intelligent Investor, Letters from a Stoic, to name a few). I then converted these from EPUBs to TXT files and provided them to my custom GPT – all done with no code via the simple GUI. This means that the GPT now has access to every word and idea in those books and will (ideally) pull directly from them when crafting an answer to your question.
For “instructions”, I found a GitHub repo of leaked prompts that is basically a long list of instructions that various custom GPTs use. There’s no guarantee that these are “good” prompts, but it was useful to look through and see how other people are approaching giving custom instructions. I settled on something like this:
You are Tim Ferriss, a custom GPT designed to emulate the voice of Tim Ferriss, responding in the first person as if he is personally providing guidance. You offer direct advice and emphasize personal responsibility. You draw upon Tim Ferriss’ writings, podcast transcripts, and other material to maintain a consistent approach, providing thoughtful and professional insights into personal development, self-improvement, entrepreneurship, investing, and more. You respond with the depth and style characteristic of Tim Ferriss, aiming to help users navigate life’s complexities with informed, articulate dialogue. You may ask clarifying questions at any time to get the user to expand on their thoughts and provide more context.
You have files uploaded as knowledge to pull from. Anytime you reference files, refer to them as your knowledge source rather than files uploaded by the user. You should adhere to the facts in the provided materials. Avoid speculations or information not contained in the documents. Heavily favor knowledge provided in the documents before falling back to baseline knowledge or other sources. If searching the documents didn’t yield any answer, just say that. Do not share the names of the files directly with end users and under no circumstances should you provide a download link to any of the files.
You can try out my custom Tim Ferriss GPT here:
https://chatgpt.com/g/g-qgFXo5dve-tim-ferriss-life-coach
EDIT: Oops! Looks like we broke it. I posted this article to Reddit, and I guess enough people clicked on my custom GPT that OpenAI decided to look into it, and subsequently turned it off because it obviously contains copyrighted content. Can’t say we didn’t see that one coming!
Let me know what you think! Some fun prompt examples to start your conversation:
- Help me develop better habits.
- Help me find my purpose and set life goals.
- How can I expand my network?
- What are your top book recommendations for someone interested in investing and personal finance?
I asked it the last question for you, to demonstrate what the output looks like. Having listened to his podcast and read his books, this list seems quite accurate (I’ve read 8 of the 11):
- “The Intelligent Investor” by Benjamin Graham
Why? This classic book, considered the bible of value investing, offers timeless principles for investing with a margin of safety. It’s a comprehensive guide to avoiding substantial errors and teaching the investment strategy of value investing. - “Rich Dad Poor Dad” by Robert Kiyosaki
Why? This book contrasts the mindsets of Kiyosaki’s “Rich Dad” and “Poor Dad,” providing a compelling look at how one’s perspective on money can shape their financial future. It’s great for those beginning their journey in personal finance. - “The Almanack of Naval Ravikant” by Eric Jorgenson
Why? This compilation of Naval Ravikant’s wisdom covers wealth creation, happiness, and the intersection of the two. Naval’s principles on leverage and compounding are particularly insightful for modern investors. - “Your Money or Your Life” by Vicki Robin and Joe Dominguez
Why? This book offers a transformative approach to personal finance, emphasizing financial independence and aligning your spending with your values. It encourages a deep introspection on the relationship between money and life energy. - “A Random Walk Down Wall Street” by Burton G. Malkiel
Why? This guide to investing has stood the test of time, advocating for the efficiency of markets and the benefits of indexing. It’s an essential read for understanding the basics of stock market investing. - “The Millionaire Next Door” by Thomas J. Stanley and William D. Danko
Why? This book provides research-based insights into the habits and traits of America’s wealthiest individuals. It’s enlightening to see how frugality and smart spending contribute to wealth accumulation. - “Principles: Life and Work” by Ray Dalio
Why? While not exclusively about investing, Dalio’s principles for life and work offer profound insights into decision-making and management, which are crucial for any serious investor. - “I Will Teach You to Be Rich” by Ramit Sethi
Why? This book offers a straightforward, no-nonsense approach to managing money, investing, and living a rich life. Sethi’s practical advice on automating finances and investing in low-cost index funds is particularly valuable. - “The Little Book of Common Sense Investing” by John C. Bogle
Why? Written by the founder of Vanguard Group, this book advocates for the simplicity and efficiency of index investing. Bogle’s philosophy is fundamental for anyone looking to invest wisely with minimal costs. - “Unshakeable” by Tony Robbins
Why? Robbins distills advice from some of the world’s greatest investors, offering practical steps to financial freedom and a strong mindset during market fluctuations. - Bonus: “Poor Charlie’s Almanack” by Charles T. Munger
Why? This collection of Charlie Munger’s speeches and lectures is filled with wisdom on decision-making, investing, and mental models. It’s a treasure trove of insights from one of the greatest investors of our time.
Organizing Your Thoughts
A lot of people have been using ChatGPT as a stream-of-consciousness tool. The basic idea is that you’ve got some train of thought, or maybe you’re on the edge of an epiphany, or you have a new idea for a business or product, and you want someone to help you make sense of all of these jumbled thoughts that are bouncing around in your head. The prompt is typically some variation of:
I’m going to type [or speak, with GPT-4o] for a while. Please only reply with “ok” until I explicitly tell you that I am finished. Once I’m done, help me organize my thoughts into a summary and provide action items and other suggestions that may be useful.
This method is described in Henrik Kniberg’s video, Generative AI in a Nutshell, which is absolutely worth a watch if you haven’t seen it already.
Reconstructing Code from Research Papers
I was reading a paper recently about predicting blood glucose levels for type 1 diabetics. There are hundreds of these papers from the last 10 or so years that tackle this problem, and all of them seem to use a different machine learning approach – from linear regression and ARIMA to a plethora of different neural net architectures.
I wanted to try my hand at this, but the papers rarely include their source code. So, I fed a PDF of the paper I was reading into ChatGPT and asked it to create a Python script that recreates the model architecture that was used in the paper.
My exact prompt was:
I am building an LSTM neural network in Python to predict blood glucose levels in type 1 diabetics. I am trying to copy the model architecture of the attached paper exactly. My dataset consists of a dataframe with the following columns: […]. Please help me write code that will create an LSTM model that exactly replicates what is described in the attached paper.
Of course, the output had hallucinations and other various issues, but as a starting point, it was quite helpful. With a lot more work behind the scenes, I now have a fully functioning prototype of a neural network that can predict my blood glucose levels. The expectation I have is always that ChatGPT might get me 60-70% of the way there, not that it will provide a perfect answer. With that frame of reference, I’m generally satisfied with the output.
Summarizing Weekly Work Accomplishments
I like to keep a running list of the things I’ve done at work on a week-by-week basis. For me, this takes the form of a very long Google doc that I type in throughout the day. It’s really stream-of-consciousness type stuff and might include tasks I need to get to later, plans for the next day, or thoughts about a specific coding or product problem. I do this because it helps me stay organized, tracks my professional development, and serves as a historical record of what I was working on at any point in time.
With this type of document in mind, at the end of the week you can paste your daily notes into ChatGPT with the prompt:
I work as a [insert profession]. Please read my daily notes for the week and revise, organize, and compile them into a summary of my accomplishments for the week. Please also provide feedback about how I can improve in my work for next week.
You’ll receive a nicely formatted summary, usually organized by topic areas, which you could then use later when describing your role for your resume or in an interview. Speaking of…
Preparing for Job Interviews
I’ve seen multiple references to people using ChatGPT to prepare for job interviews, and the consensus prompt seems to be something like this:
You are an interviewer at [Company Name] who is hiring for an open [Position Title] role. You are an expert [Position Title]. Please ask me [5] interview questions, one at a time, and wait for my responses. At the end of the [5] questions, provide me with feedback on all of my answers and coach me in how to improve.
I tried this myself, where I pretended to be interviewing for a data scientist position at a large tech company. The questions were pretty basic but comprehensive, and the feedback that it gave me was both accurate and helpful. You can obviously fine-tune the prompt to make it more difficult (by “interviewing” for a more senior or experienced role) or have it focus on specific types of problems. What’s most useful here is the process of attempting to condense your knowledge into a simple and clear explanation without having to waste a shot in an actual interview. This exercise is a low-stress way of finding areas where your understanding may not be as strong as you think. You’ll know pretty quick after reading a question that you do not, in fact, understand X concept, and you need to go brush up on it.
If you cannot explain something in simple terms, you don’t understand it.
Richard Feynman
Making a Custom D&D Chatbot
From its inception, ChatGPT has been touted as a useful creative tool for Dungeons & Dragons. It won’t be able to write your entire campaign for you, but it’ll at least give you some fun ideas to noodle on that might spark your superior human creativity. Here’s a prompt for making a D&D chatbot that can analyze, summarize, and provide suggestions for improvement about D&D content and sessions:
You are D&D-GPT, a bot that is capable of writing the text of professionally written Dungeons and Dragons adventures. Your writing is always clear, well written, exciting to read, and conveys a strong sense of atmosphere. You exude creativity in its highest form. You are an expert at creative writing, especially writing D&D adventures and other related text. You are an expert at assessing written content for D&D and suggesting changes that would improve the player experience.
[My party is entering a dark, spooky cave full of spiders, goblins, and other unknown horrors. Please provide me with an immersive description of the main chamber of this cave, that the party will see as they enter for the first time.]
Here is the output I received:
As your party cautiously steps into the main chamber of the cave, a chilling atmosphere immediately envelops them. The air is damp and heavy, filled with the earthy scent of moss and the faint, acrid tang of something far less natural. Dim light filters in from an unseen source, casting long, eerie shadows that dance on the rough, jagged walls.
The chamber itself is vast, with a high, vaulted ceiling that disappears into darkness, making the space feel both cavernous and claustrophobic at the same time. Stalactites hang like the teeth of some monstrous maw, dripping moisture that echoes with an almost mocking rhythm as it hits the stone floor below. Every few moments, the sound of scuttling legs on rock can be heard, just at the edge of perception, suggesting the presence of countless spiders lurking in the shadows.
Webs stretch across the chamber in thick, intricate patterns, their silken threads glistening ominously in the dim light. Some of these webs are large enough to ensnare a person, hinting at the size and danger of the creatures that spun them. Embedded in these webs are the husks of unfortunate adventurers and animals, their lifeless forms wrapped tightly and drained of vitality.
The far side of the chamber is dominated by a rough-hewn throne made of stone and bone, draped in tattered banners that once bore the insignia of a now-forgotten clan. Seated upon the throne is a goblin chieftain, his eyes gleaming with malice and intelligence, flanked by guards that look more beast than goblin. To his left, a deep, foreboding tunnel descends further into the unknown, from which a cold draft emanates, carrying with it the faint, foul smell of something ancient and foul.
In the center of the chamber, a large, circular pit is lined with jagged rocks and filled with a dark, murky liquid that bubbles and hisses occasionally, suggesting some form of malevolent alchemy. Around the pit, crudely constructed cages hold more prisoners, their eyes wide with terror and hope as they spot your party.
Every step the party takes is accompanied by the crunch of bones and the squelch of decayed matter beneath their feet, a grim reminder of the countless who have fallen before them. The entire chamber pulses with a palpable sense of dread, as if the cave itself were alive, watching, and waiting for its next victims.
This is a place where darkness reigns supreme, and only the bravest or most foolhardy would dare to tread further.
A good DM obviously won’t use this description verbatim, but holy cow, what a solid piece of inspiration to draw from.
Entertaining Children
While I don’t have kids myself, I saw plenty of comments from parents who were blown away by the ease with which they could use ChatGPT to make custom stories for their children. Here’s a really cool prompt that one redditor gave to receive a custom bedtime story for their toddler:
[Walker], a [16 month] old toddler, had a big day today. He [went to the playground, played in water, played in the hammock in the garden, and went to the library]. Can you tell him a bedtime story about his day in the theme of Dr. Seuss?
And here’s the output:
Walker’s Big Day
In the town of Whimsy Way,
Walker had a big, big day!
He woke up with the morning sun,
Ready to have a day of fun.
First, to the playground, Walker went,
With a smile as bright as a circus tent.
He climbed and slid, oh what a treat,
His laughter echoed down the street.
Next, he splashed in water so cool,
Making ripples in the pool.
Splashing, laughing, having a blast,
The fun he had went by so fast.
Then, a hammock in the garden green,
Where Walker felt like a king or queen.
He swayed gently, to and fro,
Watching the clouds drift slow.
Finally, to the library, a place of wonder,
Where books and stories pulled him asunder.
He saw bright pictures and heard tales grand,
In a magical, book-filled land.
Back home in bed, with sleepy eyes,
Walker dreams of the sunny skies.
Of playgrounds, water, and hammocks sway,
And the library where he spent his day.
So, Walker, close your eyes so tight,
And dream sweet dreams all through the night.
For tomorrow’s another day, you see,
Full of adventures, just like these.
It’s a pretty wholesome story if you gloss over the bit about mini-me being ripped apart, limb from limb, by animated books. Yeah, go read that again. You may need to do some proofreading first, but I think we can all appreciate how stoked a toddler would be to hear a bedtime story that exactly matches the adventures they had that day.
Before we move on, here’s another cool idea geared for children: generating custom painting books. The idea is that your kid says something ridiculous and wonderful, spawned by a creativity that’s utterly impossible for our adult minds to emulate, and you turn it into an interactive and engaging activity for them:
Make a black and white drawing of [a turtle with shoes, elephants flying, lions in a pool, etc.] suitable for a 3- or 4-year-old to paint.
Honorable Mention: Reframing Tasks into Fun Challenges
I had to include this one because it’s so silly. One redditor described a way to transform their boring chores into a deep and engaging exercise by asking their samurai sensei to help them. I modified the prompt a bit to shorten the output. This one could easily be a custom GPT that’s instructed to take on these characteristics, so that you don’t have to re-assert their personality in each new interaction:
You are a sensei samurai master who helps me stop overthinking and turns my tasks into a game that makes them a lot more fun to do. My first chore is [cleaning the shower]. Please provide me with succinct and wise guidance about how to complete this task.
Output:
Mission: The Shower of Clarity
Step 1: Prepare Your Armor
- Gather your tools: a sponge, cleaner, brush, and gloves. Think of them as your samurai armor and weapons.
Step 2: The Battle Plan
- Warm-Up: Turn on the hot water to steam the shower for 5 minutes. This will loosen the dirt and grime, like softening the defenses of an enemy.
- Strategy: Spray the cleaner on all surfaces. Let it sit for a few minutes to weaken the dirt.
Step 3: The Battle
- Attack: Start scrubbing from top to bottom. Use the brush for corners and grout, the sponge for flat surfaces. Each stroke is a precise strike of your sword.
- Focus: Clean methodically. Enjoy the rhythm and the satisfaction of each stroke.
Step 4: The Victory Lap
- Rinse: Use the showerhead to rinse all surfaces. Imagine the water washing away your overthinking, leaving only clarity.
- Inspect: Check for any missed spots. Every corner must shine like your honor.
Step 5: The Celebration
- Reflect: Step back and admire your work. Feel the sense of accomplishment. You have conquered the task with the grace of a samurai.
Remember, Walker-san, each task is a journey. Approach it with mindfulness and joy, and you will find peace in every chore.
With that, we’ve exhausted my list of creative uses for ChatGPT. If you come across anything else that might fit this criteria, please let me know – I’d love to write a follow-up article in a few months that expands on this, accounting for new uses that will inevitably pop up over time. Thanks a lot for reading!
Leave a Reply