Skip to content

My First Two Months As A Prompt Engineering Intern

Posted on:May 28, 2023 at 07:37 AM

I left Kantar at the end of March, opening up time to dive into fresh thoughts and daring ventures. What a thrilling period this has been! With new LLMs popping up, and new tools launching almost every week, it’s a bit of a race to stay on top. But I’ve set my mind on learning these tools, integrating them thoroughly into my daily life. After journeying down this path for a good two months, today I would like to share my insights, reflections, and a bunch of stories from this adventure. In fact, I am so engrossed in this new life that it feels like my new job. I invite you to join me as I unveil my experiences in my new position: Prompt Engineering Intern 😎.

AI has now become an integral part of my life. I’ve been leveraging LLMs, specifically ChatGPT, in nearly all aspects of my daily activities – from journaling or brainstorming to programming and learning. GPT has been my teacher, gym coach, my doctor, financial intern, junior and senior developer, therapist, personal assistant, and more. I believe, quite honestly, that there is no corner of my life that I haven’t sought to enhance with some input or help from AI. The results are astounding. I’ve previously written about how this technology feels like magic, but this time, I want to share specific instances showcasing how it has transformed my life.

A small girl touching a hand of a humanoid robot

Life and career coaching

Departing from my last role presented an unknown frontier. It’s both a blessing and a curse, and it has been a source of a minor anxiety for me. Unsure of my future course, I turned to GPT for career and life coaching sessions. Together, we delved into my past experiences, achievements, and my skills. GPT’s probing inquiries helped me to better understand my needs and strengths, carving out a clearer vision of my future career. Although I don’t have all the answers yet, my apprehension has significantly lessened, and I feel better armed to confront what lies ahead.

My New Journaling Companion

I have kept a journal since I was fifteen and it is my important habit. It’s a reflective tool where I distill my experiences, engage in self-dialogue, and contemplate future scenarios. Despite its benefits, journaling can feel somewhat solitary sometimes it is great just to talk to somebody, get a new perspective, bounce some ideas to create new ones. The beauty of journaling with GPT lies in its ability to maintain privacy while providing a sounding board for your thoughts. It offers “superhuman empathy” (a quote I attribute to my friend Maciek Cielecki) and uncovers your perspective through a third-party lens. All you need to do is feed GPT a journal entry and ask about the writer’s intention. The depth of insight it brings forth might surprise or even scare you - GPT can figure out a lot about you from reading your journal. It’s also insightful to analyze your journaling style with GPT, revealing your thought process, writing style, and areas to improve. My journaling is better and feels easier now.

I convinced Revolut to pay me 500 EUR

I have had a long standing beef with Revolut. Despite being a loyal and sometimes premium customer for years, they abruptly closed my account last year without any justification or means to appeal. In fact, not only my own account was closed - two of my close relatives also had their accounts suspended due to interactions with mine. I presumed these transactions initiated some AML procedures, but Revolut refrained from revealing anything about their decision. Feeling defeated after many futile interactions with customer service, I let it slide.

Nearly a year later, armed with my newfound AI expertise, I decided to revisit the case. Assisted by ChatGPT as my legal advisor, I drafted an official complaint to Revolut. As expected, they maintained that my case was closed but suggested that I could approach the Bank of Lithuania for mediation or support. This was exactly the opportunity I needed. With the help of GPT, I penned a detailed, professional, and somewhat legalistic complaint detailing my situation, my rights, and my expectations. If it weren’t for GPT, I would have likely spent over an hour drafting the document and the result would be average. With GPT, it took only 10 minutes and the end product was great! Miraculously, four weeks later, Revolut admitted their error. They apologized, reinstated my account immediately, and even offered a 50 EUR compensation if I retracted my complaint. Encouraged by their legally nuanced response, I used GPT again to craft my reply. I thanked them for reopening my account and appreciated their compensation offer. I then expressed my dissatisfaction and frustration over the unjust account closure and proposed that a more appropriate compensation would be 500 EUR. To my astonishment, they agreed! All three accounts are now fully operational and I have some additional funds to spare, ideally for more AI subscriptions 🙂.

Rainbow smudge going through a dark rural place

No More Cover Letters for Me

Every week, I set aside a few hours to explore new career opportunities and even apply for some positions. Currently, I’m actively engaged in several recruitment processes. While securing a new job isn’t my immediate priority, I keep my job hunt to a manageable two to three hours per week. Typically, this wouldn’t be enough to sift through job postings, assess fit, prepare for recruitment discussions, and — the most time-consuming task — draft cover letters. But with ChatGPT in my toolkit, this window of time is more than sufficient to build a robust pipeline of applications.

From the get-go, I enlist ChatGPT’s help. Sometimes, I don’t even read the entire job description; I just copy and paste it into GPT and then ask whether my CV would be a good fit. If GPT gives me a green light, I delve deeper into the offer, often applying on the spot.

Honestly, I’ve never understood why so many employers still insist on cover letters. Having been on the hiring side for the past 15 years, I confess I’ve never read a single cover letter — they always struck me as monotonous, repetitive, and superficial. Now, thanks to GPT, I no longer waste time crafting the perfect paragraphs for each application. Instead, I simply input the job specifications, my profile, and my expectations for the letter. The first draft is usually quite decent (my prompts are good!) and I can finish a polished cover letter for a new job in less than 10 minutes.

Additionally, I use GPT for interview prep. It assists me by generating potential questions relevant to the position, helping me craft interesting, pertinent anecdotes from my career, enhancing my CV, and spotlighting my key strengths and unique selling points. Job hunting becomes quite enjoyable when approached this way.

… nor PPT presentations

As an executive and consultant I have an uncomfortable confession to make: I detest creating PowerPoint slides. The exact reason eludes me, but staring at an empty PPT deck fills me with unease. Perhaps some long-forgotten experience has left its mark, but the core issue remains that this task brings me no joy. I often find myself spending more time than necessary on my decks, largely due to procrastination born out of my disdain for the process.

At the same time, there is one thing I am genuinely passionate about: storytelling. I find myself naturally weaving narratives, with a firm belief that everything can be intriguing, enjoyable, and relatable if framed as a story. With this realization, I recently decided to use my storytelling skills to circumvent the task of creating a new presentation that was assigned to me.

Rather than starting with the slides and trying to grind thought the unpleasant anticipation, I decided to sit down and write down the story behind the presentation. I imagined speaking to someone, explaining the context, justifying my stance, and helping them comprehend the matter as if I were having a casual chat with a friend or family member. This storytelling process took approximately twenty minutes, during which I didn’t fret about grammar or style. I freely expressed my thoughts, meandering occasionally, much like we do in friendly conversations.

I then used this story as a prompt, and asked GPT to create a plan for a presentation. And the result was amazing - it was precisely what I needed. And it didn’t end there. Equipped with my detailed narrative and this structured plan, I used GPT again to generate the actual slide content – titles, text, and even images. The entire process was fun and effortless. I can’t wait until I use this approach again.

Bedtime Story Machine

My daughters adore bedtime stories, and I love coming up with new ones for them. I must have shared over a hundred unique stories over the years, thanks to a rule my daughter Anika established early on: repeating stories is not allowed. But there are sometimes evenings when fatigue sets in. Or perhaps you have a fantastic beginning but no idea how to conclude. Or vice versa — you have a great ending in mind but struggle with the opening. Then there are those grand story ideas that would require hours of research and planning to execute right. In such scenarios, GPT is a godsend.

A couple of weeks ago, we delved into a discussion on the four fundamental forces of nature. An idea emerged: what if I could weave a tale about these forces? Maybe a story featuring two young space explorers visiting four planets, each devoid of one of these forces. This thought formed the crux of my prompt, to which ChatGPT responded with rich and imaginative details. The resulting story was engaging, educational, and inspirational. Although GPT didn’t entirely capture my narrative approach, I was able to adapt and modify the plotline as we progressed. Any deviations from my usual style were met with laughter from my girls, amused by the uncharacteristic phrases.

We’ve also had a few engaging bedtime story sessions where we co-create the narrative with GPT. We’d start the tale, allow GPT to add a paragraph, then take over again. After a few rounds, we’d arrive at a surprising and satisfying conclusion or plot twist. This interactive storytelling is a delightful game that we all enjoy playing together.

Deep Learning with AI

My love for learning and exploring new ideas runs deep. And I’m not talking about surface-level understanding but deep knowledge similar to what you’d gain from studying at a university or pursuing a PhD. I’ve poured unimaginable hours into platforms like Coursera or OCW, and have uncovered YouTube gems like EigenChris and Pavel Grinfeld during this journey. The opportunity to access high-quality education for free and at your own pace is nothing short of incredible.

And ChatGPT can supercharge this experience tremendously. Imagine having a physics professor as your uncle who patiently guides you through the intricate details of complex concepts. Or a mentor who’s well-versed in classical philosophy but is also open to discussing more everyday topics that pique your curiosity. Visualize a conversation with a Buddhist monk on your own terms, delving deep into concepts of emptiness or self-illusion, with unlimited examples and metaphors. You could even ask this monk to help craft a bedtime story for your kids based on an intriguing Buddhist principle. It sounds and is amazing.

Tools like Mr. Ranedeer take learning with ChatGPT a step further. Though my experience with these is limited, they certainly seem promising, and I’m definitely looking forward to exploring them in the future.

Dark room with modern lights and a few standing postures

Coding on Steroids

I’ve recently returned to programming professionally and now dedicate several hours each week to working as a mid-level Python developer for Reef Technologies. Reef operates with a “seniors only” policy, and their recruitment process is quite rigorous. I consider myself fortunate to have secured a position with them and am eager to demonstrate the value I can bring to the table. ChatGPT and GitHub Copilot have been my superpowers, allowing me to bridge the gap between my strong programming fundamentals (thanks to my computer science degree!) and the pragmatic realities of working in a modern tech team with large, existing codebases.

GitHub Copilot has quickly become an indispensable tool, akin to syntax highlighting or code completion in any capable IDE. The UX of Copilot, especially in Visual Studio Code which I use, is perfecly seamless. The quality of its output has generally been commendable, although I ensure not to rely on it blindly. Often, either the first attempt is 100% or you can fix it with some minor tweaks. Of course, understanding when and where to use Copilot was a bit of a learning curve initially, but for me it comes quite naturally now.

While Copilot, with its simple UX and seamless IDE integration, is impressive, it can’t do everything. For example, the Copilot Chat interface within the IDE felt somewhat clunky to me, which is why I often turn to ChatGPT. It assists me in creating basic algorithms, sandboxing solutions, writing unit tests, learning about programming patterns, and explaining aspects of existing code that I don’t understand. GPT can review my code, providing suggestions about structure, conventions, and even potential bugs. It’s like having two additional developers on my team — an eager junior to whom I can delegate tasks and a senior who coaches me through doubts or problems. And both are incredibly fast!

Sometimes, I even ask ChatGPT to clarify something that a senior developer mentioned or wrote during a code review, particularly when they use jargon or mental shortcuts unfamiliar to me. Instead of reapproaching a human colleague for clarification and potentially causing minor frustration, I can now ask GPT to elaborate on a specific comment or feedback. It’s been a game-changer!

In my estimation, the combination of Copilot and ChatGPT has increased my efficiency as a software engineer by an order of magnitude. Of course, your mileage may vary — my situation is somewhat specific, possibly unique. I’ve learned how to leverage these tools to their full potential in my circumstances, and naturally, others will adapt them to their own patterns. I believe a productivity multiplier between 2-5x is within reach for anyone in this industry.

Closing Thoughts and Future Plans

This list could go on and on, but I feel this post is already too long. The point is, ChatGPT has been a game-changer for me; it feels like a major upgrade to the operating system of my mind. I use AI across the board and continually explore novel ways to intertwine it with my everyday routines. I look forward to other AI tools — Siri and Alexa, in particular — catching up soon. Using them presently is a stark reminder of the giant leap GPT represents; as Microsoft CEO Satya Nadella put it, they’re “dumb as a rock”.

One area of my life that remains relatively untouched by GPT is productivity and time management. I’m something of a productivity fan — GTD is a weekly habit, Inbox Zero helps me manage my email, and I use Roam Research as my “second brain” and journaling app. As of now, none of these apps or workflows have reaped the benefits of AI tools. I wish ChatGPT could integrate with my Google Calendar to assist in scheduling tasks or planning my week. I made an attempt to use the Zapier ChatGPT plugin for this but to no avail. It might be time to consider creating some plugins myself to bridge this gap. Given the support of AI, it shouldn’t be too difficult or time-consuming 🙂.

Shoutouts

All images from Unsplash: