🤔 ‘Working’ Fast & Slow

If you are not living under a rock, you might have probably heard about the book “Thinking Fast and Slow”

It talks about 2 systems the first is more abrupt and unconscious and second is more slow, logical.

If you think about it all the AIs work in the former mode, the way an LLM writes is just like how our phones keyboard suggests the next words. The AI cannot think through like we do. (that’s why we need to prompt it to use scratchpads)

I've been under the weather this week and still am.

However, I felt more productive than in previous weeks because I mixed both System 1 and System 2 thinking.

I had some fast-paced days where I just powered through tasks using System 1, and other days where I used System 2 to think logically about how to simplify things.

This balanced approach helped me work on the RAG AI app for the website, scrape 22,000 AI tools, find ways to sort the data, upload a long-form video, and post regular content on Twitter, LinkedIn, Instagram, and YouTube.

🤩 22K AI apps Scraped

I went to all the website UGC or non UGC (repositories), and setup scrapers with the help of Claude.

I am not left with the task to sort it, and assign tags so it can be treated as a vector database and the AI could answer questions from it.

I will be using gpt- for sheets to assign these tags, should take me <$50 to do as this. Tough part here is fetch the pricing models, and manually review it. I feel this part I can do post launch.

👼🏻 Prompt Engineering Is Not Just Meh

I never thought purchasing prompts from websites made much sense.

I believed these Large Language Models (LLMs) were good enough on their own

I didn’t even find these ‘hyper-niche’ apps based on chatgpt wrappers useful. In fact, I still believe this is true for most cases. Apart from the rarecases where you are an advanced user and you really care about quality for instance programming, copywriting, and designing.

For these specific cases, it's definitely worth learning from the guides Anthropic has published.

Just by using simple <xml tags> in my prompts, my outputs improved by at least 50% - and I'm not exaggerating.

Plans Ahead

I've started working on the website now, finding AI tools and categorizing them into marketing and non-marketing groups.

I'm thinking about solutions to set up a search engine plus RAG tool to make this database useful.

One challenge is figuring out how to build the database of these AI tools in a way that's simple and effective, unlike WordPress. I've also been struggling to find good brand names for the AI project. I have a few shortlisted like "Bots Gone Wild," "AI Brew," and "The AI Afterlife," but I don't feel as excited about them as I did with "Ultrapreneurs."

I know stressing over the brand name is a classic startup mistake, so I'll choose the one I like best once I finish setting up the tech stack.

📊 This Week in Numbers

Social Media Growth:

7+ reels and shorts posted

1 long form video on Youtube

Thanks for joining me on this Ultrapreneur journey! Let's build something extraordinary, exposing vulnerabilities and the obstacles we face.

To your unconventional success,

Vaibhav

P.S. Got ideas to make this community even more badass? Hit reply. I would really love to talk.