HNHacker News
TopNewBestAskShowJobs

xiphias2

8,476 karma · joined September 7, 2018

submissionscomments
xiphias2··on A ChatGPT clone, in 3000 bytes of C, backed by GPT-2 (2023)
Everybody who has worked on a large/huge code base knows that most of it is there for historical reasons: it's always safer to add a new feature than to replace something old, as most of the time people don't know if that old code is useful or not.

BTW tinygrad shows that tens of thousands of lines are enough, it's skipping even most of the AMD kernel drivers and talks to the hardware directly.

xiphias2··on Make a "Magic Eye" image using Excel
Thanks for the link, I saw the Tokyo parallel view beautiful crispy for the first time, now I can't do it again, I guess my eyes became tired and need some time
xiphias2··on Ask HN: How can I grow as an engineer without good seniors to learn from?
,,I am not producing the software to the highest standards of the industry since we don't have any rigorous cross-checking''

Standards can slow you down, which is fine in a big tech company, but probably not good for building value for your company.

It's better to think more about what may break in your code base that could be emberassing for your company and build integration / othere tests for those important cases.

Otherwise generally simple code is good code, keep building and have fun in life. Lot of those ,,standards'' just overcomplicate things and may hurt your company.

xiphias2··on Starlink Direct to Cell
While it's true, SpaceX will raise money by making Starlink (the less important / easier business) public and keep SpaceX (the real business that makes it possible that Starlink depends on) private.
xiphias2··on Playground Wisdom: Threads Beat Async/Await
It's because implementing this is not that easy: there are differences between the implementation of coroutines and await that makes it tricky (especially waiting for both CPU tasks and network events).

For Python I loved this talk by David Beazley:

https://www.youtube.com/watch?v=MCs5OvhV9S4&t=2510s

He's implementing async/await from coroutines ground up by live coding on the stage

xiphias2··on SEC chair Gary Gensler will step down
It will be much more fun and he is more than capable of doing it, he just has burned too many bridges.
xiphias2··on Bluesky Exceeds Threads DAUs in the US
We can all agree of helping others, making the world a better place and being deeply opposed to war being important values. The hard part where it's important / interesting to discuss is what achieves it the best and what takes precedent over what.

I loved listening to Lex Friedman's imterview with Javier Milei as he was describing how Argentina went to a state of almost hyperinflation by the government helping every poor person with overtacation and overregulation.

I believe that the main reason for wars (and the only way it can be achieved) is money printing (elespecially creating bonds).

Most people from US would never agree in sending money to Ukraine (or at least not everybody) if the amout was explicitly written out on their tax statement. Also I believe the governments shouldn't take money away from people to redistribute how they want (just the minimal amout that they need for keeping up a working society).

xiphias2··on Bluesky Exceeds Threads DAUs in the US
I use these threads as meditiation: have a deeper understanding of how people different from me think and accepting that it's part of the world.

Nowdays I'm much more scared than angry about socialist thinking getting us all closer to WW3 though.

xiphias2··on Call it like it is: TSLA rise is not due to performance, but hope for corruption
Electrec itself got corrupted a few years ago. It was far the best web page for news about electric vehicles going into details, but it was probably bought and now it's just ragebait.
xiphias2··on TSMC cannot make 2nm chips abroad now: MOEA
He understands it, he has much more tech people around him than 8 years ago (his second largest donor just purchased 100k GPUs).
xiphias2··on Principles for product velocity
Sure, it's the speed of adding a new incremental simple feature to the existing code base.
xiphias2··on Principles for product velocity
I love The Rise of Worse is better, and this sounds just like that:

https://www.dreamsongs.com/RiseOfWorseIsBetter.html

It doesn't explain _why_ simple implementation is more important than having a simple interface, but it just makes an observation that simple implementation usually wins.

In my experiments simple implementation won for velocity, because quite often I need to change small parts in all the implementation no matter how modular / reusable code I'm trying to write.

As my velocity increases testing becomes more important as well (especially integration testing), as there are more features interacting in a small amount of code.

xiphias2··on And now we see what is possible
It did work for giving power to Elon. Many people think that it helped a lot in getting Trump reelected.
xiphias2··on Rewrite It in Rails
It's not really sad. Optional static typing has advantages and disadvantages and there are enough dynamic programming languages that have it (JS, Python for example).

I think the problem with static typing is that it changes the whole ecosystem too much to be considered just optional, even if it's backwards compatible (like in Python).

Just as an example I'm using lambdas in fasthtml in Python that don't allow type specifications for adding routes and fasthtml keeps on spitting out warnings that I didn't specify the type and doesn't handle variables in lambda functions correctly.

xiphias2··on Moderates Should Vote for Harris
Sure, right now the democrat policy is letting migrants in so that they vote for democrats.

As a Hungarian I see my own country going down economically because Hungarians outside my country got voting rights and benefits because of Orban without reqirements for taxation, and now we are stuck with a dictator.

Of course US has a much more stable system, but the leftist strategy there seems to me like a parallel to what Orban was doing on the right.

xiphias2··on Moderates Should Vote for Harris
This endorsement is the weakest I have seen. Paul Graham wanted to enorse the left no matter what because of his silicon valley connections, and then created an exaplanation afterwards (it would be suprising for me if he ever voted republican in the last 20 years).

In my experience the best way to stop banana republic from happening is to always give power to the party with less power (which was democrat last time, republic this time), as both parties have a danger of trying to accumulate too much power nowdays.

xiphias2··on 35% Faster Than the Filesystem (2017)
Yes, it's especially good for that.

There are lots of small things missing from it (like boolwan support), but it's used as an archival format and backwards compatibility for the archive format is prioritized over new features.

xiphias2··on Why "founder mode" is a rebranding for micromanaging, top-down leaders
The article is garbage. I just checked 1 example on a bit longer timescale: Tesla 5Y is 1100% up, it's amazing stock performance.

Actually there isn't even 2Y option in Google search, so the author should explain why he chose that time horizon (my guess is just to make a negative article).

xiphias2··on The Rise of Worse Is Better (1991)
There are some differences between your examples in my opinion:

Rust started as an experiment of Mozilla team replacing C++ with something that helps them compete with Chrome in developing safe multi-threaded code more efficiently. It took a lot of experiments to get to the current type system, most of which gives real advantages by using affine types, but the compiler is at this point clearly over-engineered for the desired type system (and there are already ideas on how to improve on it). It's still too late to restart, as it looks like it takes 20 years to productionize something like Rust.

As for React I believe it's I believe an over-engineered architecture from the start for most web programming tasks (and for companies / programmers that don't have separate frontend and backend teams), but the low interest rates + AWS/Vercel pushed them on all newcomers (and most programmers are new programmers, as the number of programmers grew exponentially).

HTMX and Rails 8 are experiments in the opposite direction (moving back to the servers, nobuild, noSAAS), but I believe there's lot of space to further simplify the web programming stack.

xiphias2··on The Brutalist Programming Manifesto
This is a mix of amazing and really bad advice. And the problem with it is that it is a series of advices / philosophy instead of explanation.

Reading the original Rise of Worse is better article ( https://www.jwz.org/doc/worse-is-better.html ) gave me a huge appreciation of programming implementariin simplicity, and improved my pace of coding a lot. It shows a real example of why C won over Lisp (a much cleaner interface) 50 years ago.

The funny part is that right now I'm programming in something that may be closer to Lisp (Python), but happy to rewrite parts of it in C in the future when it gets practical.

The amount of features I can add in a small time just because I always pick the smaller implementation surface (while still moving forward) is amazing.

Regarding dependencies: I may want to move to lower level libraries in the future when my project is in a more advanced state, but again they are great as training wheels (lower level would help with customization, but I'm waiting until I _need_ it).

xiphias2··on Type-erased generic functions for C: A modest non-proposal
It's a cool idea, simple, I like it much more than the _Var and _Type stuff, but I'm not sure if it's 100% backwards compatible:

If I pass (int) and (void ) at different places, they will convert to void *, but won't be accepted in the new code.

Still, it would be great to have an implementation where it can be an optional error, and see what it finds.

xiphias2··on OpenAI to become for-profit company
It's quite simple: the talent pool that had already enough money that they quit their well paying job at a for profit company in part because they wanted to continue working at a non-profit high impact.

As OpenAI found its product-market fit, the early visionaries are not needed anymore (although I'm sure the people working there are still amazing)

xiphias2··on Twitter/X will let people you've blocked see your posts
It looks like a reverse of a trend I have seen of people understading the details of selectively hiding information from others and using it for manipulation.

,,Show to group of people except X'' is manipulative behaviour (instead of kicking out the person from the group, or in this case not sharing publicly what's not public).

On instagram I have seen a lot of these things, for example posts, stories, archived stories all have different very strange behaviours that only people to take the time to understand (more manipulative people) use, which makes people more manipulative in itself.

xiphias2··on Show HN: Bullshit Remover
Love the project, too bad it didn't get more votes
xiphias2··on Safe C++
This looks like a huge improvement from the original Circle language that didn't look like C++ at all. I love it and would love to have it in C++.
xiphias2··on What I gave up to become an engineering manager
The best manager I had never gave up on building things and getting deeply technical: he just spent half of his time asking around people if everything is good, and whenever he saw that somebody is stuck, sat down together and helped debugging the problem (and maybe helped finding people getting it unstuck).

About meetings: he put some ,,meeting''s in the calendar entry, becuase PMs loved taking his time unnecessarily (he was still available for meetings, just not all the time).

As a high level example Elon Musk talks a lot about his managing style at Starbase in his interviews with Everyday Astronout, and it looks similar: he always goes to the place which is most critical for the product and looks at the technical issues.

xiphias2··on Inductive or deductive? Rethinking the fundamental reasoning abilities of LLMs
Transformers are amazing pattern matchers and terrible use of GPUs for reasoning, which is mostly search + execution of highly non-linear programs (lambda calculus).

I love seeing Victor Taelin experimenting with parallizing these programs (with HVM and other experiments with proof languages), but it's sometimes a bit sad how much time researchers take in making papers about existing things instead of trying to improve the state-of-the art in something that's most probably missing from the current models.

xiphias2··on Procreate defies AI trend, pledges "no generative AI" in its illustration app
When a CEO hates a technology it's probably because it leapedfrogged his app and he's already seeing decreasing revenue growth.
xiphias2··on Ask HN: What books should I read to improve as a software engineer?
Look at Casey Murrey's Clean Code horrible performance youtube video:

https://www.youtube.com/watch?v=tD5NrevFtbU

I also watched clean code videos and used it in practice, and why it _felt_ right my code size blew up and actually it got harder to understand by other people instead of easier.

Nowdays my main guidelines are getting a balance between less lines of code (while still readable), readable variable names and better performance.

Also I may get hate here for it, but I learned a lot from watching George Hotz programming tinygrad (available on YouTube).

xiphias2··on Dutch DPA fines Uber €290M because of transfers of drivers’ data to the US
Or maybe all companies will learn to leave EU behind in innovation.

Even though the rules are great, I'm just not sure if it will be good or bad long term for EU.

← PreviousPage 13 of 34Next →