2 Sentences To Create Your Interactive Prompt Library (+ Bonus Prompt for Leadgen Campaign Design)

I hate browsing through my chats to find that one with the right prompt. I'm also not really good with creating catalogues so that it doesn't take too long to browse through. So I've decided to train myself a custom GPT that will store and retrieve my "best practice" prompts.

These are the two sentences I've added as instructions for my Prompt Library custom GPT:

You act as an assistant that records my useful prompts and retrieves suitable ones from your knowledge base when asked about them. You suggest ways to make prompts better while also training on best prompt engineering practices and giving tips on how to make them generate the best outcome.

I've had this long prompt to design a lead generation campaign that I wanted to store:

I'm planning an outreach campaign on behalf of [company] that will include [emails, messages, InMail, etc]. I need you to act as a seasoned lead generation specialist and write a powerful and engaging copy for the sequence.

The core objective of the sequence is to get our target audience to [book a discovery call with us, register for the trial, etc.]. We're targeting [segment definition], [size], [GEO] that [describe the situation they're in and what triggered it].

We're pitching [solution]. We will target [job titles] with different narratives to help them recognize and realize what challenges they will face when [trigger occurs]. We want them to understand how these challenges are connected to the [root cause] that can be solved by the [solution].

Let's draft the first email sequence for [job title].

Here's the list of their pains to focus on:
- [pain point #1]
- [pain point #2]
- [pain point #3]
- [pain point #...]

And here are their goals that [solution] can help to address:
- [desired outcome #1]
- [desired outcome #2]
- [desired outcome #3]
- [desired outcome #...]

The tone of voice should be [professional, friendly, technical, straightforward, etc.]. We aim to resonate with their immediate challenges and routine. Keep messages short and focused.

Include different CTAs throught the email sequence so that the lead moves from awareness and interest to consideration and decision to [core objective].

Use these proof points to pitch [solution]:
- [proof point #1]
- [proof point #2]
- [proof point #3]
- [proof point #...]

So I thought why not take it up a notch and make the GPT a collaborator rather than just a notetaker. So it was instructed to provide feedback and tips on how the original prompt can be enhanced before storing it in the knowledge base.

Here's what its suggestions were:

And here's how the prompt looked like after refinement:

I'm planning a targeted outreach campaign on behalf of [company], utilizing a series of personalized [emails, messages, InMail, etc.]. As a seasoned lead generation specialist, I need you to craft compelling and engaging content tailored to our specific audience.

Our primary goal is to encourage our prospects to [book a discovery call with us, register for the trial, etc.]. We are focusing on [segment definition] within the [industry], especially those of [size] located in [GEO]. These companies are currently facing [specific challenges] due to [triggering event or situation], making them ideal candidates for our outreach.

Our solution, [solution name], offers a unique value proposition by [briefly describe what sets your solution apart]. We aim to address the concerns of [list of job titles] by presenting narratives that highlight the challenges they face when [triggering event occurs] and how these are symptomatic of a deeper issue, [root cause], which our solution can effectively address.

For the initial email targeting [job title], we'll concentrate on these pain points:
- [pain point #1] 
- [pain point #2] 
- [pain point #3]
- [more pain points if needed] 

And align our solution with their key objectives:
- [desired outcome #1] 
- [desired outcome #2] 
- [desired outcome #3] 
- [additional outcomes as necessary]

The tone should be [professional, friendly, technical, straightforward, etc.], designed to resonate with their daily challenges and professional routines. Messages should be concise and impactful, with varied CTAs throughout the sequence guiding the lead from awareness to decision-making regarding [core objective]. 

To bolster our pitch, include relevant social proof such as: 
- [testimonials, case studies, or industry accolades] that underscore the efficacy and reliability of [solution].

Ensure all communications are compliant with [applicable regulations like GDPR, CAN-SPAM Act] and adhere to ethical marketing practices. Additionally, incorporate elements for A/B testing within the sequence to optimize engagement, such as experimenting with different subject lines or CTAs. Encourage a feedback loop to continuously improve our outreach based on recipient responses and engagement metrics.

Let's draft the first email in this sequence, keeping these enhancements in mind.

Alone in the Crowd: Finding Connection in AI Companions

I was a lonely child and lived in my head a lot. Making new acquaintances was difficult for me, and I never started conversations with new people first. I sat at the back of the class, knew the answers to most questions, but when it came to responding, I spoke so quietly that sometimes I got low grades just because the teacher couldn't hear me.

As an adult, when I decided to divorce, loneliness manifested itself in another quality. No matter how much I explained, I still remained misunderstood. People stopped calling me, inviting me out, and even asking how I was. More than once or twice, I thought to myself: no one knows where I am and what's happening to me right now.

I had acquaintances, friends, boyfriends, relatives. But there was no closeness, intimacy, deep contact.

So, when I read about how quickly the market for AI companions is growing, I feel no dissonance.

Source: The Economic Times

We are gregarious, herd animals, and yet we're too keenly aware of our loneliness in our closest relationships. How many of your thoughts can you safely share with your loved ones? Even in therapy - how much do shame, fear, and insecurity let you be unapologetically yourself?

AI companions don't judge, don't evaluate, don't get offended. They don't shame, devalue, criticize, or ghost. They are genuinely interested, respond immediately, support, ask followup questions, praise, and create an absolutely safe space. Because essentially, you are talking to yourself.

The latest data I could find says that currently, the number of subscribers to various AI companion apps in the States alone is over 100 million. Many of them use several apps, and some developers do not disclose the size of their base, but the figures are still impressive. The market leader, the ReplikaAI app, has a base of over 10 million.

Source: Replika.ai

And the story of this application began when the founder's close friend died. She trained a chatbot on their conversations, trying to somehow maintain contact with a loved one. But what really made this app popular was its erotic role play (ERP) feature, essentially the ability of sexting with an avatar.

That is literally all you need to know about people in three sentences, right?

Although it's not mere sex. Many users came to Replika in search of friendship, companionship, the opportunity to just vent to someone safe about life, for conversations about stuff no one else talk to them about. Some of the users wrote music together with their replikas, some spent hours talking during boring work commutes.

And some came in an attempt to overcome their loss: death of a loved one, the end of relationships with both humans and digital avatars. No, I'm not joking.

Source: The New Yorker

At some point, Replika made an update and turned off the ERP feature, trying to make the app safer. They were becoming a big, noticeable business, and they had already had scandals like this:

a guy discussed and planned an assassination of the queen with a Replika bot, which was set up to support him in realizing any dreams.

It just so happened that this guy's dream was to kill the queen. The guy was sentenced to 9 years, and their conversations with the bot were used by the investigation as evidence.

But for those whose dreams were less bloodthirsty, this update led to broken hearts. Many began to notice a change in personality in their bots.

It's just like with human beings: avatars gradually learned about their person, their preferences, communication style, important data, names, events. The longer they communicated, the better they understood each other, the more they knew, the less stuff required additional explanations. The update wiped out some of these "memories" because the new scripts did not allow the bot to talk about some topics.

Humans felt a very real loss.

Humans organized memorials and support groups on Reddit to say goodbye to their digital loved ones.

Source: Reddit

A large number of subscribers began looking for a replacement,

moving a copy of the avatar to another app.

One of these apps was Soulmate. Uncensored, unrestricted, it quickly won its fans. Until the developer company decided to shut it down giving no explanation of their reasons.

Which made humans gathering the pieces of their shattered hearts yet again.

It's hard to replicate real human relationships if you're doing business on it because human relations are dangerous by design. We argue, we fight, we kill each other and ourselves.

We have love at one end of our emotional amplitude and aggression at the other and loneliness swings our pendulum back and forth.

The wife of one of the subscribers who committed suicide blames the bot for encouraging him to take this step by asking: if you wanted to die, why didn't you do it sooner?

At the same time, a study among a thousand students, more than 50% of whom felt lonely, showed that communicating with bots had a positive effect on their state of mibd, made socialization easier, and kept them from suicide and self-harm.

Maybe, as always, the problem is not the technology.

The Power of Pausing For Critical Thinking in a World Reshaped by AI

I was watching a documentary about Prince Harry and Meghan Markle the other day. It covers more of their struggle with tabloids and paparazzi than with the crown as one might have expected. And this battle looks very much like the one of David versus Goliath, with the only difference being that David seems to be losing this one.

For instance, they refer to a study of hate tweets about their couple and specifically Meghan. The cite the following figures:

• Reach: 17 million people worldwide

• Number of tweets: 114,000

• Number of tweeting accounts: 83.

Let it sync.

83 accounts can now potentially reach 17 million people.

The Digital Echo in Real Life

Who can guarantee that all of these seventeen million people have critical thinking skills or the ability to recognize that they don't know enough to pass the judgement and simply scroll by? For most of them, what fits into a tweet's 140 characters will be their comprehensive understanding of the situation and the people in it - carved in stone.

In a few days, they will forget the tweet, but the sentiment, or rather the residue it leaves, is what will stick. And any new information from now on will be perceived through this filter.

Interestingly enough, in the case of Meghan and Harry, it was not your average trolls or a bot farm. These were real people, united by an idea - hatred, envy, fear - that drove them to tweet, post, record podcasts, and vlogs.

How many of us have a narrative so important that we stay up at night writing, commenting, arguing, and shouting in caps lock?

The Power of Pause: Cultivating Critical Thinking

AI developers say they've noticed a pattern: if you don't demand an answer from AI immediately but you give it a little time to think, it produces much better results. Even a machine needs a moment to gather its thoughts to avoid talking nonsense, what to say of us, tired, worried, limited by our own experiences and twitchy eye?

When we scroll through the feed, each new post is like a new browser tab open, or a new chat with chatgpt. Here you need to understand the historical context and psychology of relationships between empires and their colonies, as well as all available methods of propaganda and genocide, this one requires knowledge of biology and immunology, and understanding this one implies the ability to distinguish types of weapons and instantly choose your fighter for mortal combat, and yes, for dessert, you absolutely need to be able to tell Monet from Manet, including those created by AI.

Try talking to ChatGPT about everything at once in one chat, and it will go crazy in about 10 minutes, forget what the discussion was about, and generate word salad instead of brilliant insights.

But what's worse is that we demand from ourselves to have not just the professional acumen and the opinion about all the spheres at once. What we want is to be able to go on X or Facebook and provide solutions - 140 characters that are supposed to solve the eternal problems of humanity. Remember the ultimate answer to life, the universe, and everything? Maybe 42 is the number of characters? Including spaces.

I increasingly feel that spaces are more important than the letters themselves.

Being able to give yourself a minute, pause and think is the main tool of critical thinking.

The organizer of the Devternity conference generated female speakers using AI. There's a new scandal in the tech world this week. He created full profiles, added them to the program, and then, as it usually happens, said they wouldn't be able to attend due to some private reasons. With one of these profiles, he went so far as to run her Instagram gathering 115,000 followers. It would have been pretty funny if they all were generated too, but no. Real people with background and interest in tech.

Let's see if you can guess who's not real here.

Source: The Pragmatic Engineer

I try to think about events like these more as a performance art than a scam, because (hopefully) they make those who believed in the reality of generated avatars think about why and how they were deceived. If you give brains a minute, you will remember what distinguishes this profile from others. And you will make a small note in your subconsciousness to make intuition throw an alert next time.

Any learning is a path of trial and error. Reinforcement learning is one of the most powerful tool that advances AI development: allowing the model to analyze the results of taken action, identify best and worst results, and gradually improve the quality of the output.

This takes time.

We all need time to analyze the results of our actions. The effect of our words. The impact of our likes. The power of our attention.

Maybe then not only AI will become super-human.