Hello 2026

Motivation

New years are generally great anchor points for reflection and for re-adjusting the course of your life. It’s always good to think about what has gone well, so you can keep doing more of that. At the same time, life is ever-changing; therefore, it’s also a good idea to pause every now and then and consider what things can be trimmed from your life.

2025 was a good year, and today I’m more in the mood to chart the course ahead for 2026 rather than reflect on 2025. This may seem counterintuitive, but I’ll go ahead with what I plan to do in 2026. I’ll start by listing a few themes I really want to work on, and then, from those themes, I’ll try to reverse-engineer concrete daily habits that will help me get there without feeling overwhelmed.

I want to be the best version of myself.

De-Cluttering Digital Spaces Around Me

Writing Should Feel Frictionless

My initial setup for this blog was based on Jekyll and GitHub Pages, and it served me well for years. However, I often felt a bit of friction whenever I wanted to customize something - not because of Jekyll itself, but due to my limited knowledge of Ruby. I really enjoy the simplicity of Markdown and the idea of static site generation. Since I write code in JavaScript and Python on a day-to-day basis, the Astro static site generator felt like a natural fit.

On the design side, I’ve been deeply inspired by the simplicity and philosophy behind Bear Blog, and I wanted to build something along similar lines. While scouting for ideas on the internet, I came across this neat implementation, astro-bear-blog. Although I ultimately set up my blog from scratch, discovering this repository was both helpful and validating.

Fast forward to today: I’ve migrated my blog to Astro and set up a streamlined workflow that makes writing more frequent and far less frictional.

• Deployment is handled via GitHub Actions (thanks to their generous free limits).
• Hosting is still done through the same battle-tested GitHub Pages.
• Creating a new post is now as simple as opening a GitHub Issue, after which an automated workflow takes care of committing it to the underlying repository.

Social Media Hygiene

I think I’ve been doing a good job of restricting my social media usage, so for the most part, I plan to continue with the same approach. I’ll stick to using web versions only. Mobile apps make it far too easy to open social media without any real intention. Whenever I find myself getting too involved in a particular political topic, I’ll take a break from all social media for at least two days. If a piece of information is truly vital, it will find its way to me. Excessive usage usually just leads to repeated exposure to the same stories and can end up taking more of a toll than I realize.

Simplify Digital Media Backups

I want to put together a few simple scripts or lightweight workflows that can help me:

• Easily back up photos and videos from my phone to my laptop.
• As a moonshot, enable semantic search across that media so I can find things more intuitively when I need them.

Knowledge Building

In 2026, I want to build some serious mental muscle around LLMs in particular. It is not about following yet another trend. I genuinely believe this technology is going to change our lives in a profound way. I just find it very difficult to suppress my curiosity to know more about it, so rather than fighting that impulse or surrendering to anxiety, I want to focus on it and consciously channel my time, energy, and effort in this direction.

For this, I plan to follow Andrej Karpathy’s advice of learning in a JIT (just-in-time) manner. I’ll be working through his YouTube series, nn-zero-to-hero.

Initially, I thought it might be a good idea to start from the very basics of deep learning. However, after watching a few videos from this series, it feels like one of the most intuitive ways to learn the subject. The focus on LLMs first - from bigrams to transformers - makes the learning process concrete, and much of what I’ll pick up along the way should be broadly transferable to deep learning in general.

I also plan to read some influential papers in this space. I’ll likely create a dedicated page on the blog to collect and track them in one place.

Spaced Repetition

This is another area I can improve to better retain what I learn. Good old Anki should do the stuff. I just need to be mindful of a few things:

• Extracting highlights from my e-reader in a consistent way.
• Potentially using an LLM to help generate flashcards across different topics.
• Most importantly, setting up a sustainable cadence to do this efficiently.

Final Thoughts

It’s never too late to work on things I’ve always wanted to do. Writing about them regularly will:

• Nudge me to test my understanding in a deeper and more meaningful way.
• Encourage me to organize my thoughts more clearly.
• Give me opportunities to learn from others and accelerate my journey toward becoming the best version of myself.

Looking forward to the rest of 2026, cheers!