Boring is good

Link: https://jenson.org/boring/

Context

Yes, this post summarises the current trend in LLMs well. The hype is about people adopting to the value of LLMs, but soon they’ll will realise that they were on the wrong path (one example is they thinking, LLMs could replace developers) and then we will settle on the thing they are good at. I know, throwaway code, temporary code, and simple stuff that you know you need to do, you know how to do, but not worth the time to manually type it in and craft it. I am really excited and positive about the SLMs, the small language models, I want to use it to just be a google search but simple and not ripping out my entire project into a react and python boilerplate mess.

Building a lexical analyzer from scratch in C

Link: https://devlogs.xyz/blog/building-a-lexical-analyzer-from-scratch

Context

This actually cleared up how to write lexers from scratch. I was wondering if that was bunch of if-else to parse each tokens, but we do have to group the kind of tokens and then write specific conditions on how to parse them. Now that makes a lot of sense. Worth exploring more by writing my own markdown parser, even adding more features and syntax. If your soul screams to write your own flavour of markdown, let the muse take over you. This month or one day, not this weekend though.

Creating a static site for all my bookmarks

Link: https://alexwlchan.net/2025/bookmarks-static-site/

Context

I am very much this kind of person, I want to access my bookmarks and the linkblog too. I am developing it, but can’t make it polished enough to be usable. I had created linkblog.netlify.app. This is work in progress, anyone can add but I would like to add the authentication, but that again creates a friction element, this all constraint bother me then.

David Heinemeier Hansson (DHH) on Rails World 2025: Opening Note

Link: https://youtu.be/gcwzWzC7gUA

Context

I can’t fathom here, how did we end up backwards? It takes some thinking to deploy to prod, yaml manifests, and what not to deploy a simple API, how the heck people are accepting those? It is like a egoistic culture to have complexity and assume that it will work, but they are only adding complexity upon layers of complexity that developers will have to clean up themselves. Rails is doing a phenomenal job in the developer ecosystem.

Finding and Exploiting Buffer overflow password cracking in C

Link: https://youtu.be/W1okWEzGAGo

Context

This guy is producing great stream, bangers after bangers. This one not quite remarkable but enduring nonetheless. I will vouch for persistence over hype and fuss.

Source: techstructive-weekly-60

Go Struct Alignment: A Practical Guide

Link: https://medium.com/@Realblank/go-struct-alignment-a-practical-guide-e725c8e1d14e

Context

I have read this and it makes sense, a bit wired but nice. Writing structs should be carefully planned, so just add the largest ones at the top and cram all the smaller ones thereafter. The rule of thumb to follow if you have any memory-heavy or scarce use case. Handy functions like Sizeof, Alignof, Offsetof are used to get the total byte size, memory alignment requirement, the field start position of the struct or any fields. Use it to craft the proper and perfect structure by tinkering and aligning.

Magical Systems Thinking

Link: https://worksinprogress.co/issue/magical-systems-thinking/

Context

I love this post, it hits home for me Systems should grow from the simplest possible solutions and then branch off from the possibilities and situations. Creating patches to the existing systems will only survive for limited time or none at all. Sundar Pichai estimated in late 2024 that over 25 percent of Google’s code was AI generated; as of mid-2025, the figure for Anthropic is 80–90 percent. The comparison to AI slop is great and fits well here. Vibe coded mess is no exception to systems thinking. We have seen LLMs want to patch and keep on patching existing code mess, we want to start from scratch, the urge is right, but we worry about whether we will lose the progress? The progress is fake progress. The code LLM produces will always be throwaway code, as it is no subconsciously written, call me philosophical, but code is art and it needs attention form a human soul in order for it to work, not technically but overall in order to complete its purpose. NEW SYSTEMS CREATE NEW PROBLEMS’ and ‘THE SYSTEM ALWAYS KICKS BACK’. As systems become more complex, they become more chaotic, not less. The best solution remains humility, and a simple system that works.

My Favorite Postgres 18 feature: Virtual generated columns

Link: https://tselai.com/virtual-gencolumns

Context

: I agree to this, there are pros and cons of both. Stored makes write heavier but are read efficient. Virtual makes it write easier and read heavier. You have trade-offs, you need to decide based on the computation that impacts how you want the column to be generated. I don’t like the notion of JSON flattening in Postgres. Postgres is not a database that would be ideal for that kind of data. I know there are tons and tons of support for JSON, but tables and JSON, I can’t bare it at once. Those two are just separate entities for me. Maybe they are useful in one-off values, not not much. Switch to NoSQL if you have that lengthy data.

Myopic Focus

Link: https://thedailywtf.com/articles/myopic-focus

Context

Wow! Fantastical and Tech stories, are my love. Gist of the story is people are very myopic(viewing in very hindsight) about system and think of the existing workflow as ritualistic that makes it harder for others to make it better or even improve the quality or realibility. The story goes like > A new developer chops replaced fragile ID pool logic with robust UUIDs, eliminating inevitable crashes. But their manager fixated solely on declining unit test numbers, demanded rollback. It made the project a bit risky to sustain, but the fault is not of developers. Its the myopic view that causes it.

OpenAI dropped another Codex

Link: https://youtu.be/j9wvCrON3XA

Context

I agree to this, it is just a slight improvement to the GPT-5 experience. Not much here. I kind of hate to admit that free users can’t use the web codex, the cloud instance of the agent. So, basically free users are devoid of the code experience of the GPT models. Google might win that race. Gemini CLI is so cool, they also have Jules.