Forwarded from (φ (μ (λ)))
I don't know what programmers of today need to realize that programming is first and foremost an activity of human thought, like any other field such as mathematics or engineering. Just because engineering has made improvements in producing certain tools, that does not replace the activity of engineering itself.
Is the average engineer of today's significantly better than the engineers during Renaissance? Programmers have a habit of fixating their own activity on certain tools, be it a programming language or any of these "silver bullets" that supposedly can replace the activity of programming as a whole.
Brooks said it decades ago, there are no silver bullets. Programming will always be hard, no matter whether your IDE starts connecting magically with your thoughts or can write bloated prototypes within a few minutes.
This remains the case for all engineering fields much older than software engineering, no amount of "innovation" in machines ever replaced good engineers entirely.
Forget even the activity part, there is a deep theoretical (read: mathematical and logical) problem at the heart of automating computation that has simply not been touched due to all attention being put into stochastic methods.
Do we even theoretically know if all computation can be automated? Do we even know what such computation theoretically looks like? We are not even asking these questions, we are throwing random statistical shit at the wall and praying it sticks most of the time.
A few years ago, somebody told me that whatever the industry accepts as mainstream is technically "the best way". If the rise of stupid LLMs has shown anything it is the exact opposite of that, industry—like any other market—cares nothing other than what brings profit to the private entities.
Is the average engineer of today's significantly better than the engineers during Renaissance? Programmers have a habit of fixating their own activity on certain tools, be it a programming language or any of these "silver bullets" that supposedly can replace the activity of programming as a whole.
Brooks said it decades ago, there are no silver bullets. Programming will always be hard, no matter whether your IDE starts connecting magically with your thoughts or can write bloated prototypes within a few minutes.
This remains the case for all engineering fields much older than software engineering, no amount of "innovation" in machines ever replaced good engineers entirely.
Forget even the activity part, there is a deep theoretical (read: mathematical and logical) problem at the heart of automating computation that has simply not been touched due to all attention being put into stochastic methods.
Do we even theoretically know if all computation can be automated? Do we even know what such computation theoretically looks like? We are not even asking these questions, we are throwing random statistical shit at the wall and praying it sticks most of the time.
A few years ago, somebody told me that whatever the industry accepts as mainstream is technically "the best way". If the rise of stupid LLMs has shown anything it is the exact opposite of that, industry—like any other market—cares nothing other than what brings profit to the private entities.
Telegram
(φ (μ (λ)))
For those who might still be under the misconception that the current bubble of "AI" will somehow absolve all problems of modern software engineering and take their jobs:
The essence of a software entity is a construct of interlocking concepts: data sets…
The essence of a software entity is a construct of interlocking concepts: data sets…
👍5
I'm not into AI news, but Anthropic claiming that Claude wrote a functioning C compiler "from scratch" is so hillarious. Like, it's wrong in so many levels that I'm actually surprised how the blog post was even approved by someone.
Forwarded from shakhzod's lab (Shaxzod Qudratov)
1. Essentials of Compilation - https://t.me/haskelluz/42505/42929
2. Parser Combinators - https://t.me/haskelluz/42505/42930
3. Scrap your Boilerplate - https://t.me/haskelluz/42505/42931
2. Parser Combinators - https://t.me/haskelluz/42505/42930
3. Scrap your Boilerplate - https://t.me/haskelluz/42505/42931
Telegram
kei in Haskell O'zbekiston
[Meetup 2026] Essentials of Compilation: An Incremental Approach
https://www.youtube.com/watch?v=Ni4B8KpRca4
Code:
- Master branch: https://github.com/haskelluz/m26-eocia.ml
- Solutions branch: https://github.com/haskelluz/m26-eocia.ml/tree/lvar
Slides:…
https://www.youtube.com/watch?v=Ni4B8KpRca4
Code:
- Master branch: https://github.com/haskelluz/m26-eocia.ml
- Solutions branch: https://github.com/haskelluz/m26-eocia.ml/tree/lvar
Slides:…
Engineering Notes
Is anyone still using ChatGPT for programming assistance?
ChatGPT when I forget to say “do not write excessive comments” explicitly:
😁28
Google Chat is cooking some really useful features recently. If you’re a small org/team and use Teams or Slack, I’d recommend giving Google Chat a try.
👍7
I don’t like Anthropic as a company, especially because they’re now starting to ruin FLOSS as well. But you know what, mad respect to them for consistently showing a middle finger to the US DoW.
👍21
I’ll be honest, I don’t like working with vibe coders at all. From my experience (both at work and online communities), they’re annoying to work with, to teach, and to learn from. They almost never do their job properly, often screw up your part, know nothing about debugging, and more often than not, argue over stuff that they don’t understand (at least from my experience).
I’m not saying vibecoding shouldn’t be a thing. It’s great to be able to vibecode a personal website or a simple telegram bot without actually learning programming. However, if you’re getting paid to do software engineering professionally, for god’s sake, please understand what you’re doing, or at least, don’t be pain in the hole for someone else in your team.
I’m not saying vibecoding shouldn’t be a thing. It’s great to be able to vibecode a personal website or a simple telegram bot without actually learning programming. However, if you’re getting paid to do software engineering professionally, for god’s sake, please understand what you’re doing, or at least, don’t be pain in the hole for someone else in your team.
👍44👀3👎2
Turns out, accidentally using
docker-compose instead of docker compose is enough to feel nostalgia.🍾9