What’s next?
Heyy there folks!
It's been long since I have published anything, but better late than never! I was finding it hard to build the habit of writing every week, and after the second post, it felt more of a work than something I enjoy, hence I stopped writing entirely. But thanks to the 17 people who subscribed after the second post, who indirectly pushed me to write again.
Researching, filtering, and curating topics for the newsletter takes a lot of effort (kudos to everyone who does), but I realized it's not something I enjoy. However, researching really helped me push boundaries and learn new stuff. So, I have decided to reduce the amount of research so that I don't get bored but still push myself a little. How does that sound? idk, let's experiment!
Other than that, I'll try to produce something each week (excluding this newsletter) and share it here. What do I mean by 'produce'? It can be anything, such as a mini project, a technical blog/article, or a YouTube video (I have been wanting to start YouTube for so long, but the effort it takes is surreal.) If you like this idea too, I think it'll be fun to do it together! Let’s connect on Discord, share our progress, and keep each other accountable each week. Who’s in?
Here's the article which inspired me to do so — https://training.kalzumeus.com/newsletters/archive/do-not-end-the-week-with-nothing
Here's my reading list in case you're interested — https://readinglist.adarshdubey.com/
That’s all about the newsletter! I’ll try to keep the rest of the newsletter short for this week.
Interesting AI trends and developments
ChatGPT Search is trash?
So OpenAI recently announced it’s all new Search feature, and a lot of people are saying it’s kinda trash. Well, lucky me, I applied to be it’s beta tester and got in!
ChatGPT Search comes with a Google Chrome Extension, basically replacing Google (or whatever search engine you use) with ChatGPT. Here’s a video of me using it —
For me it’s cool to try the search feature, but I don’t really want to replace the search engine I use. I feel more comfortable using traditional search engines. Moreover, I find Perplexity a lot better than ChatGPT Search. But hey, who knows? This might be the future.
Here’s the announcement article — https://openai.com/index/introducing-chatgpt-search/
Claude’s new analysis tool
So Claude can now analyze data. It’s great if you generally work with excels and other stuff. Even though I have studied data visualization, I generally find it very boring to analyze the data myself by writing code in a Jupyter notebook. Here’s what Anthropic says about this new tool —
When you need precise, verifiable answers from data, Claude now works more like a real data analyst. Instead of relying on abstract analysis alone, it can systematically process your data—cleaning, exploring, and analyzing it step-by-step until it reaches the correct result.
It basically writes and runs JavaScript code to do all this within the chat, and also shows visualized graphs and some insights.
The announcement includes a demo video with some explanation. It’s a 2 minutes read so I highly recommend you going through that. Here’s the announcement article — https://www.anthropic.com/news/analysis-tool
Well, there are a lot other news, but all of them include announcement of some new model. I feel it gets repetitive so I have decided to not include those.
What have I been up to?
Micrograd implementation in C++
After finishing Andrej Karpathy’s ‘Introduction to back-propagation’ lecture, I thought it’ll be fun to implement Micrograd in C++.
Why not C? My goal with re-implementing Micrograd in some other language was to understand the working behind Micrograd and re-think how back-propagation works. I didn’t want to spend a lot of time fighting (or learning new stuff) with a language but rather focus on the implementation part. Hence, I chose C++.
Implementing Micrograd in C++ was really fun and I feel pretty confident with Back-propagation now. If you’re a nerd too, check it out here — https://github.com/inclinedadarsh/incligrad (obviously I have named it ‘Incligrad’)
Understanding and building RAG applications
So recently I have been exploring applied AI, and RAG has to be the first step. If you don’t know what RAG is, here’s my attempt —
RAG are AI(LLM) chatbots/agents that takes in a data source and then answers your question based on that data source. In easy words, “Chat with your PDF”, but with any data source.
To get start, I completed this short course from Deeplearning.AI — LangChain: Chat with your data.
Currently I’m reading articles and blogs to get started building one of my own. I’m thinking of building “Chat with any YouTube video” to start with, let’s see how it goes!
Conclusion
Well, that’s it for this week! I hope you enjoyed reading this edition. I’m constantly looking for feedback to improve, so if you have one please feel free to reach out to me.
To end with, if you doubt yourself, here’s a tweet for you (must read) — https://x.com/arpit_bhayani/status/1835651124155613442
Hey, did you like this post? Please let me know by messaging me on Discord or tweeting about it, make sure to tag me (@inclinedadarsh).
You can connect with me on Twitter, Discord, WhatsApp or you can find all other links here.
Thank you for your time.
This post took a while, but I enjoy reading your Substack posts—keep it up!