Bookmarks

Certainty in uncertain times goes a long way in approaching touch situations with calm and clarity. 30 Jan 2026

What not to prematurely optimize is one of the most important engineering decisions in system design. Like how Dropbox avoided generating preview thumbnails for all uploaded video/images based on their data-backed insight that most such files will never be viewed in a search result. Saved them both storage and compute costs. 27 Jan 2026

Development has ground to a halt. The thing is plagued by security issues. As fondly as I remember MySQL from my formative years, the sad reality is it stopped being a viable option soon after its Oracle acquisition. Why Sun, why did you sell yourself? 21 Jan 2026

A good case study in picking just part of an off-the-shelf platform solution and building out the rest based on your org’s unique architecture and constraints. ‘Most important is to stay grounded in real problems rather than chasing architectural trends.’ 18 Jan 2026

Add intentional ‘thinking time’ to your workdays. Use it for reflecting on hurdles encountered, measuring the impact of a completed task, and planning for future value creation. Otherwise, all you are doing is keeping the lights on rather than investing in your growth: good for your company, bad for your career. 16 Jan 2026

TIL about Evals, the automated testing analogue to traditional unit/integration tests. Since running LLMs (for evaluation) in CI pipelines isn’t cheap, it’s good to prioritize test scenarios based on top buckets of real-world user issues. 15 Jan 2026

Dr. Gandhi’s piece confirms my prophecy about the AI dust settling by late 2025 or start of 2026. 100% automation shouldn’t be the goal. Human knowledge is priceless (goal mine as per Dr. Gandhi). Rather than replace humans with AI, augment agents with humans for the best outcomes. Good for business, good for humanity. 15 Jan 2026

In my 2017 book Artificial Intelligence for.NET (co-authored with Nishith), I described how language models of that era worked. A combination of gradient descent and supervised fine-tuning gave them predictive and classification superpowers for the given domain/context. Those models were next token predictors in the truest sense.

The attention/transformers architecture together with data-science-optimized GPUs revolutionized language models and made today’s LLMs possible. The fundamentals of pre-training and teaching models have stayed the same.

It’s incredible how a single breakthrough in software architecture can lead to generational advancements.

Beautifully written article by Alex Xu, so easy to digest. Highly recommended even if you aren’t familiar with the technical details. 11 Jan 2026

TIL that there’s a name for the inter-service communication pattern that I implemented in our microservices solution five years ago.

It’s called Choreography-based Saga. This is where each service listens for interesting events from other services to trigger actions.

The alternative is Orchestration-based Saga where a central orchestrator listens to events from and issues commands to services. 11 Jan 2026