Techstructive Weekly #72

Week #72

A good productive week (I am not sure, I am saying that in the last month of 2025). I would be writing a yearly review in a couple of weeks and this week might lift my spirits up. I was fresh, given time off from work, not really, but given the time to experiment.

I read a lot, fiction, technical, even philosophical, and random Hacker News articles. I read 2 books of Harry Potter. I haven’t read Harry Potter, I have watched the first 3 movies, but not books. I enjoyed it, it was subtly different from movies, I missed the quote “What an Idiot” from the movies, it was not in the book, i was disappointed but the atmospheric adventures was amazing read.

I even played with golang after a while, I streamed after ages. I felt back. I might be making much more comebacks. Have lot of ideas for projects, I know how to deal with such situations, now, do one thing. Don’t get into a trap of overthinking and porcastinating and delaying for prefection traps. Just start one project or one improvement and lets see where it leads me.

Quote of the week

“Never trust anything that can think for itself if you can’t see where it keeps its brain”

-- Mr. Arthur Weasley, Harry Potter and the Chambers of secret (Chapter 18)

I don’t know how I picked it up. But it hit me right on. LLMs are like Tom Riddle’s diary, aren’t they? I feel so relatable here. Software developers like me are Ginny and they hallucinate us to do certain things in a certain way and we follow it, like in a trance, and it makes the software collapse. What a relatable quote. It makes me think that reading fiction is healing. For nerds like me, I would just read and escape the world.

Read

Craft Software that make people feel something
1. So good. Just do it because there is a curiosity. Inspiration is also cool and need for software to make.
When programming becomes repetitive, the odds of you creating something that makes people go “wow” are reduced quite a bit. It isn’t a rule, of course. You need to be inspired to make inspiring software.
1. This is aspiration, the level of it is high here.
This is what I’m talking about: taking time to build something so that once people try it, they remember it for as long as they live.
If you’re going to vibe code, why not do it in C or even Assembly
1. Yeah! This is a valid point, If you are not caring about the code, why bother with the language.
2. Choose a language which computers can understand the best, x86 or even machine code, 1s and 0s.
3. Stephen says to create a VOPL vibe-oriented-programming-language which suits LLMs. Maybe this is what it will look it, who knows.
4. But the point is strong, if vibe coding is not caring about programming, why bother choosing tech stack and languages, just let it choose whichever it is familiar and good at just like a good’ol developer.
Why Transformers must hallucinate
1. This is a spicy take. The points are valid
  1. Averaging is a critical mistake
  2. They miss checking whether an answer exist or not
  3. Always answers will guarantee hallucination
  4. It should asses the question, then generate and not generate and then asses.
Please don’t automate science
1. Boy, it takes courage to speak this! Well spoken.
They are here because they love research and want to contribute to advancing human knowledge. If you take the human out of the loop, meaning that humans no longer have any role in scientific research, you’re depriving them of the activity they love and a key source of meaning in their lives. And we all want to do something meaningful. Why, I asked, do you want to take the opportunity to contribute to science away from us?
1. This hits it harder
Science automation is coming whether we want it or not, and we’d better get used to it. The train is coming, and we can get on it or stand in its way.
I think that is a remarkably cowardly argument.
It is up to us as a society to decide how we use the technology we develop. It’s not a train, it’s a truck, and we’d better grab the steering wheel.
1. There are bangers after bangers
Making human intellectual or creative work redundant is something we should avoid when we can, and we should absolutely avoid it if there are no equally meaningful new roles for humans to transition into.
1. I want to quote each and every paragraph it seems, this is so good, almost like it comes out of my mouth
You could further argue that working on cutting humans out of meaningful creative work such as scientific research is incredibly egoistic. You get the intellectual satisfaction of inventing new AI methods, but the next generation don’t get a chance to contribute. Why do you want to rob your children (academic and biological) of the chance to engage in the most meaningful activity in the world?
1. So true, I have been here when there was model after model releases in a week, i think in August-September. It was a wild month. I was overwhelmed, I didn’t get a chance to slow down. AI can do a lot of things, can produce a lot of things, and I cannot handle it at that pace. I need time to absorb, it makes productive, true, but it quickly overshoots the danger productive bar. The moment where you are too much productive that you lose track of every context in your head.
Software never Fails
1. It never fails. It does what it was intended to do. If the intended thing and the actual thing was different than that’s a developer problem and not the software’s. It did what was written as its not a magic wand that will do what you thought to do.
I read more than I write, do you?
1. Yes, this is true for me too. Reading needs to be more or rather at certain point, balanced from writing.
2. You can’t just keep on reading and produce nothing. You will have to reflect on what you have consumed. This newsletter is exactly that.
3. After I got a full time job, I suddenly had a lot of time, I realized i needed to dump my learning somewhere and I was following Register Spill by Thorsten Ball and quickly felt the urge to log my reading and interesting things I find throughout the week. There were a lot of things, I took for granted, they got lost and most of the things didn’t stick. I started this and it helped me realize the ample amount of time I have to learn, explore and tinker on stuff. It was liberating. Hence writing the 72nd edition of this. Its fun.
Vibe Coding is made and depressing
1. Sigh, The frustration is quite evident
I just feel sad with how AI has bastardized my profession, which I worked hard for the last 15 years. There is no best practices anymore, no proper process, no meaningful back and forth
1. I can even feel it with 5 years of coding, man has been doing it for 15 years! Humans gets too excited when they can produce code i think.
The Gemini API Key Frustration
1. Yeah! have you set up Google products without opening and closing a bunch of tabs
2. Here you are in 2025. Wait a minute, was PaLM a thing? wasn’t google notes to be shut down?
3. What is happening, what is AI Studio, Vertex AI, Jules, Antigravity, Gemini CLI, Gemini models of course, dug sneaked into various products, geese. Google!
A TSP game I wanted for 10 years: built in 4 hours
1. This is quite a good thing, I am not good at frontend, i let ai do it, i do what i am good at, writing backend. Really? need to see it carefully again.
Growing Software
1. What a great analogy. Software is not architect a building but its growing a garden. You don’t have a layout already built, some parts are clear not all. Software is ever changing. You need to build something, observe and change constantly just like a gardener. Gardener doesn’t plant a bunch of plants and forgets, but rather it nurtures them, observes and then takes care of them.
2. Software is not something you build, its something you grow. Its a slow process.
How HTML changes in EPUB
1. This is quite intuitive. I didn’t knew epub is a collection of XHTML documents. Its quite obvious now.
2. Because e-book has to be scaled from each character, so it is xhtml or some variant of it. Zooming, Changing fonts, all happens at all levels or doesn’t look good. So that is the perfect use case for it.

Created

SQLog: Create Strict Tables
Live streaming Advent of Code Day 1 and 2 in Golang
YouTube Shorts on FreeCodeCamp Daily Challenges

Watched

Software is getting worseDouble click to interact with video
- Yes, This is true, from I have not observed actively, but looks valid and intuitive - We need these three ingredients for a good software - Passion - Trying to solve a problem with care - Developers been funded or are self sustainable

As the software gets old it get big

As the software gets big, it gets worse

These are not the things responsible for bad software (might be but not only)
- AI Slop
- Not using Rust (or any other framework or language)
- It’s an Electron app (or using a particular framework or language)
STRING is actually an integer type
- Everything is flexible unless you type strict
- SQLite is flexible, as I have said its a double-edged sword until you don’t want it to be.Double click to interact with video
Just use Postgres
- Wow! I love this. I want to write a technical book too. It is such a great adventure to be in. But burnout seems to be stronger there.
- I love Postgres can be used as a message queue, gen ai application, full text search I knew and JSON was obvious.
- Getting deep into the tech is important, I need to focus on thing at a time.Double click to interact with video

Learnt

Using Strict mode in table creation in SQLite.
- If you enable strict option while creating a table, you need to specify either of the 4 types (int, text, blob, any), Don’t use any, it defeats the purpose of using strict option.
- You will lose the flexibility entirely when you type strict, can’t even ignore the type before a column while creating the table.
Code generation is getting cheaper
- Look around, there are 10s of agentic CLIs and what not, everyone is commoditizing the AI tokens, its getting cheaper too. The speed at which it generates stuff (not quality, quantity) matters in code, if you know what you are doing, then you can ster it quickly to get results quick. I have learnt it the hard way, by vibing hard and after 4 years of hard work. It finaly feels I can understand the productive side of AI coding.

Tech News

OpenAI Launches GPT-5.2
- Everyone says it will be a best model for a while, and then silence. It just amalgamates into the slop. Nothing major I can see, 0.01% maybe.
Google Rolls Out Reimagined Gemini Deep Research
- This is something cool, but deep research is neither deep nor its a research, its just summarising the links into a hallucinated piece of document.
Agentic AI Foundation Launched Under Linux Foundation
- Wow! Anthropic in opensource, who would have thought.
- Anthropic donated MCP, OpenAI with Agents.md (I am actually confused, what is a agent and file name belonging to a company, why they own it?), Goose from Block.

That’s it from the week. I am excited to code more in parallel, by spawning the agents in the background and reviewing the code. Yes, I no more write code by hand. I like to when I do, but as a software developer and employee, I don’t write code as I used to a year back. I spawn code agents and heavily review its output. Embrace the way it is, adapt the skills, there’s no other way.

For more news, follow the Hackernewsletter (#774th edition), and for software development/coding articles, join daily.dev.

See you next week!

Happy Coding