Published on by Nayar

Getting ready for PyCon US 2026. Here’s how 2025 edition went for me.

Last May, I flew from Mauritius to Pittsburgh to attend PyCon US 2025. I had 2 layovers, namely in France and Detroit. That’s was a total duration of 24hrs.

But I managed to reach Pittsburgh safely in full 😀

Going to PyCon was easy using public transportation. I was so delighted to see the banners on the streets.

I volunteered in the registration desk. I made some good friends there. I hope to volunteer even more this year.

The talks were really interesting. They cater from total beginners to pro level.

I added a pin marking the only Mauritian in the conference. In PyCon 2026, we will be at least 4 this time 😀

I was shocked to see keynote speakers openly bashing Elon Musk (whom I have great respect for) and even the sponsors of the event for being monopolies and anti-consumers sometimes. Is this why we call USA the land of the free? Free like in freespeech?

One of the gems of PyCon are hallway talks i think they call it. The topics range from tech to “how to write a book” to “what is consciousness?”. It is a great way to get to know and listen to other participants as well.

Alright.  See you in PyCon 2026 🙂

 

Published on by Nayar

Import/Export Flaw in ERPNext. Why?

I exported my Item Categories from my dev Environment and imported to prod environment using Frappe own import and export tool. Guess what?

 

It added quotes around the ID and failed to remove it during Import time. So I was unable to update my items due to ID (name) not matching. I really don’t understand how this is possible. ChatGPT says:

 

 

ERPNext is an awesome piece of software. Unfortunately built on the wrong foundation I think. A primary key cannot and should not be a string containing real data.

 

 

Published on by Nayar

Attending Biopunk Meetup in Frontier Tower

I was intrigued by this event so I decided to attend it. It was situated in Frontier Tower in San Francisco. The tower is dedicated to be a frontier in research. Each floor has people working on different aspects such as Robotech, Neurotech and AI among others.

Everyone was greeted and was welcomed to present themselves and what they are working on. It was very brief without any visual support.

After that we got to chit chat with the other guests. It was really awesome to meet scientists who are pushing the knowledge boundary. I hope one day i’ll be able to help the scientific community with my computer and programming knowledge.

 

Published on by Nayar

Vibe Coding with Convex Chef at Convex

My Friend Abdur from PyMug recommended me attend events on lu.ma. I found this awesome Vibe Coding Meetup from convex. I didn’t know convex company nor products so I decided to go there and try it out.

I was greeted by Wayne and he guided the participants on how to get started.

Since I was working on working my Frontend for my Kreole Chatbot, I decided why not ask convex chef to do it for me.

With 1 prompt, boom! It made me a whatsapp clone interface. But I wanted to interact with it. Send messages and reply back. I asked it to do that and it worked.

I was even more shocked to know that this is actually saving in the database. Crazy!!!!

Next step was to make it display the products from my chatbot. I thus asked it to create a carousel of products.

Magic once again. It looks so pretty. All the best practices from design i think are there.

But but but. When I asked it to make the add to cart functionality, it didn’t work due to some bug. My AI tokens were also depleted at that point and I started to debug manually.

I see it is a tiny error but since i was totally new to the framework, i couldn’t solve it.

And then came Food time and Demo time where everyone could showcase their vibe coded project.

 

It is such a cool office. They have guitars and drum sets!

Was a really cool experience. I invite everyone to try it out and let me know till where you are able to stretch it: https://chef.convex.dev/

Published on by Nayar

Using AI to Catch a Thief in Flic-en-Flac

A thief stole a bicycle on my porch and we have CCTV cameras installed. However, witnesses couldn’t recall the exact time, leaving us with three full days of footage to review. Watching it manually was impractical, especially since the mobile app couldn’t fast-forward remotely as it was not on the same LAN as the DVR.

We downloaded all 65 video files from the DVR, but even that was daunting. That’s when I turned to AI.

Using AI to Solve the Problem

I asked ChatGPT to write a Python script to detect bicycles in the footage.

The script worked immediately, but after some testing, I adjusted it to detect people instead—making it faster and more efficient.

It saved only frames with detected persons, reducing hours of video to just the important clips.

Tools and Setup

I ran the script on an Ubuntu machine with an NVIDIA RTX 4070 ti super GPU, which processed the footage quickly. What would have taken days was done in almost 3 to 4 hours.

The Moment of Truth

Reviewing the condensed clips, I finally found the thief on camera, riding away with the bicycle.

What do you think is the proper way to bring justice?