HNHacker News
TopNewBestAskShowJobs

coherentpony

2,363 karma · joined March 30, 2013

submissionscomments
coherentpony··on Delaunay Mesh Generation (2012)
> While true, it also produces a topology that makes you want to throw your PC out the window.

Can you elaborate?

coherentpony··on Memstop: Use LD_PRELOAD to delay process execution when low on memory
You are the person that made the initial claim. The burden of proof is on you, not someone else.
coherentpony··on Async Queue – One of my favorite programming interview questions
I have tried the first two things in that list. I haven't tried the third. I would need to think about how to do that in a way that is general enough that it can be re-used across candidates with different skill sets. I like the idea, though. Thanks for sharing it.
coherentpony··on Async Queue – One of my favorite programming interview questions
> It just doesn't feel like this is right.

I know the feeling.

The author says this is one of their favourite interview questions. I stop to wonder what the others are.

When I'm interviewing a candidate, I'm trying to assess really a few things: 1) the capability of the person I'm interviewing to have a technical conversation with another human being; 2) how this person thinks when they are presented with a problem they have to solve; and 3) can this person be trusted with important work?

For 1) and 2), coding interviews and the types of artificially constructed and unrealistic scenarios really aren't helpful, in my experience. I care a lot less about the person being able to solve one specific problem I hand them and I care a lot more about the person being able to handle a much more realistic scenario of being hand handed an ill-defined thing and picking it apart. Those conversations are typically much more open-ended; the goal is to hear how the person approaches the problem, what assumptions they make about the problem, and what follow-ups are needed once they realise at least one of their assumptions is wrong.

This is a really hard thing to do. For example, I imagine (but do not know) that when a medical practice hires a doctor for a certain role, there is an expectation that they already know how the human body works. For an ER doctor, you might care more about how well that person can prioritise and triage patients based on their initial symptoms. And you might also care about how that person handles an escalation when a patient presents not too awfully but is in fact seriously ill. For a GP, it's probably more important for a practice to care more about healthcare philosophy and patient care approaches rather than the prioritisation thing I mentioned above. I'm spit-balling here, but the point is these two situations are both hiring doctors. You care less about what the person knows because there is a tacit assumption that they know what they need to know; you're not giving the candidate a trial surgery or differential diagnosis (maybe... again I'm not a doctor so I don't actually know what I'm talking about here).

If I'm hiring a software engineer or performance engineer, I am trying to figure out how you approach a software design problem or a performance problem. I am not trying to figure out if you can design an async queue in a single-threaded client. This problem doesn't even generalise well to a real situation. It would be like asking a doctor to assume that a patient has no allergies.

Item number 3) is "Can this person be trusted with important work?" and this is basically impossible to determine from an interview. It's also impossible to determine from a CV. The only way to find out is to hire them and give them important work. CVs will say that a candidate was responsible for X, Y and Z. They never say what their contribution was, or whether or not that contribution was a group effort or solo. The only way to find out, is to hire. And I've hired candidates that I thought could be trusted and I was wrong. It sucks. You course-correct them and figure out how to get them to a place where they can be trusted.

Hiring is hard. It's a massive risk. Interviews only give you a partial picture. When you get a good hire it's such a blessing and reduces my anxiety. When you hire a candidate you thought would be good and turns out to be an absolute pain to manage it causes me many sleepless nights.

coherentpony··on Quarkdown: A modern Markdown-based typesetting system
> Converting complex HTML pages with navigation, ads, and JavaScript into LLM-friendly plain text is both difficult and imprecise.

Oh my god. It just occurred to me that LLMs may have a better experience “browsing the internet” than humans do.

That is so tragically depressing.

coherentpony··on Show HN: AI Baby Monitor – local Video-LLM that beeps when safety rules break
I did an in-person class. I highly recommend it.
coherentpony··on Accents in latent spaces: How AI hears accent strength in English
> Real-time accent feedback is something language learners have never had throughout all of human history, until now.

Do you have a source for this? It doesn't seem plausible to me, but I'm not an expert.

coherentpony··on Doom GPU Flame Graphs
Ah yes. Ok.

Yes, that makes perfect sense. Thanks.

coherentpony··on Doom GPU Flame Graphs
> Correct. Which means that every 20ms pixel slices two or three frames. Which is a really really bad way to profile!

If 20 ms is a reasonable frame time for a modern game, why is it an unreasonable thing to profile?

I understand other, shorter, frame times may be interesting to profile too. My point is that if you want to understand a reasonable or realistic workload, then it should also be reasonable to profile that workload.

coherentpony··on Doom GPU Flame Graphs
> Ain’t no game using 20ms frames!

A frame every 20ms equates to 50 frames per second. Doesn't seem too unreasonable for a modern game.

60 frames per second would be one frame every ~16 ms.

coherentpony··on Office is too slow, so Microsoft is making it load at Windows startup
It shouldn't be "an idea" at all. Profile the application, find the hotspots, understand what the performance limiter is, and fix it.

Granted, this is all Hard Work. I understand that. But it's the right thing to do.

coherentpony··on The Gang of Four is wrong and you don't understand delegation (2012)
Yes, I meant the RDMA verbs API.

Thanks for the insight.

coherentpony··on The Gang of Four is wrong and you don't understand delegation (2012)
> everyone knew the ways in which that was a bad idea

The state machine pattern is used currently in verbs. That was the state of the art for many decades before higher level interfaces appeared.

Why was it a bad idea? Can you elaborate?

coherentpony··on 4o Image Generation
> we’ve built our most advanced image generator yet into GPT‑4o. The result—image generation that is not only beautiful, but useful.

Sorry, but how are these useful? None of the examples demonstrate any use beyond being cool to look at.

The article vaguely mentions 'providing inspiration' as possible definition of 'useful'. I suppose.

coherentpony··on Hubble Captures Vivid Auroras in Jupiter's Atmosphere (2016)
Here's a timelapse of the aurora: https://science.nasa.gov/asset/hubble/hubble-time-lapse-of-a...
coherentpony··on Accelerating scientific breakthroughs with an AI co-scientist
It bothers me that the word 'hallucinate' is used to describe when the output of a machine learning model is wrong.

In other fields, when models are wrong, the discussion is around 'errors'. How large the errors are, their structural nature, possible bounds, and so forth. But when it's AI it's a 'hallucination'. Almost as if the thing is feeling a bit poorly and just needs to rest and take some fever-reducer before being correct again.

It bothers me. Probably more than it should, but it does.

coherentpony··on Kansas tuberculosis outbreak is America's largest recorded since the 1950s
It likely will never be an option. You could consider spoiling your ballot instead? I have no idea if spoilt ballots are counted in America, though.
coherentpony··on Every board game rulebook is awful [pdf]
> I need a how-to right now and all I get is an explanation.

This isn't a reasonable expectation. Your current state of documentation may be very different from some-other-software's current state of documentation. There may (or may not) be commonalities across those states, but assuming the most conservative situation leaves you with no commonality and the author's only option is to write the explanation for you. From there, you have to think about what transformations you need to apply to your current state to get it to the desired state.

Contrast this with game rulebooks. There is a clear commonality: situations where none of the players know any of the rules. Therefore, the rulebook can easily be written assuming no knowledge of the current rules of the game. Players that know the rules of the game can either a) go make everyone coffee and avoid polluting the learning phase with information that stretches the patience of the folks reading the rulebook; or b) claim to the players that don't know the rules that the rulebook is useless and they can do a much better job explaining all the nuances much better than the person that designed the game.

coherentpony··on How CERN serves 1EB of data via FUSE [video]
> Responses referenced the contributions the development of the infrastructure supporting the basic research itself have made, which is fine, but not the benefits of high energy physics discoveries.

I was one of those responders.

There were two very deliberate reasons I specifically avoided talking about particle physics:

1. I interpreted the tone of the original question to be extremely highly cynical of any scientific contribution particle physics has made, so I instead went for 'consequential' things. More excitement around education, outreach, and other adjacent aspects that are beneficial to humans. I did this to potentially avoid, "How has discovering a new Boson made my rent cheaper?" types of arguments that are only ever made in bad faith, but have been made to me a disheartening number of times in my career; and

2. I am scientist and I have collaborators and colleagues at CERN, but I'm not a particle physicist and so I didn't feel adequately qualified to highlight them. I was expecting someone with more expertise would jump in and simply do a better job than I ever could.

If I interpreted the tone of your question incorrectly, please understand that it wasn't an intentional sleight on you, and simply an artefact of a) plain text being an incredibly poor medium for communicating nuance; b) a defensive measure that I have had the displeasure of dealing with in the past. And if you were genuinely curious, that's wonderful, and I'm sorry that I didn't offer you more grace in my response.

coherentpony··on How CERN serves 1EB of data via FUSE [video]
https://en.wikipedia.org/wiki/CERN#Scientific_achievements

Here's a couple, in case you don't want to read the page:

- CERN pioneered the introduction of TCP/IP for its intranet, beginning in 1984

- CERN has developed a number of policies and official documents that enable and promote open science

- The CERN Science Gateway, opened in October 2023,[179] is CERN's latest facility for science outreach and education

I purposefully picked items that weren't directly particle physics related.

coherentpony··on Zen 5's 2-ahead branch predictor: how a 30 year old idea allows for new tricks
https://link.springer.com/article/10.1007/s10676-024-09775-5
coherentpony··on Higher Order Derivatives of Transforms
The first paragraph of the preface on page 7 in the book gives an explicit example of “higher order”: https://terrytao.files.wordpress.com/2011/03/higher-book.pdf
coherentpony··on How I got an O-1 visa as a software engineer
> Nice, just like in the good old days with serfs and feudal lords.

I can't tell if you're being sarcastic or not.

Serfs were largely agricultural workers bound under the feudal system to work on their lord's estate. Foreign workers aren't serfs. They're not bound to a particular estate, nor do they typically execute agricultural work. They have the right to collectively bargain their salary, working conditions, and benefits. They are free to resign their position and leave the country.

The O-1 visa is a temporary worker visa. The expectation is that folks on a temporary work visa work temporarily. That is, there is no intent to immigrate.

You are also not bound to a particular employer. You're only bound to that employer on that visa instance. You can, if you like, apply for another O-1 visa sponsored by a different employer.

There are, in some cases, restrictions on work even for folks that have the right to work and live in the US. For example, if company A contracts out work to company B, it is often not permitted for company A to offer workers at company B full-time positions at company A because of the existence of the contract. If someone freelances for company A they often can't also hold full-time positions at a competitor to company A. Is this feudalism? Of course not.

TL;DR: Freedom to work does not imply or mean, "I can do whatever work I want for whoever I want." There are rules, regulations and laws present for a reason. We can debate whether or not those reasons exist in good faith, but equating "I can only work for one specific employer on a temporary work visa" to serfdom is awfully disingenuous, in my opinion.

coherentpony··on Airlines will make $118B in extra fees
> I think sometimes we forget how fuzzy language is.

I'm willing to bet a reasonable sum of money that the folks that designed the dark pattern didn't forget how fuzzy language is. In my opinion, the entire spirit of the design of the dark pattern is in fact the complete opposite: to exploit the fact that language is fuzzy in an extremely targeted fashion with the express purpose of extracting as much revenue as is pragmatically possible from the customer.

coherentpony··on Airlines will make $118B in extra fees
No, I'm a software engineer.
coherentpony··on Airlines will make $118B in extra fees
> They're not lying to you

That's correct. But it's correct in the sense of "technically correct". It's a language-lawyer type of correct that nobody actually expects when communicating in plain English.

Yes, they're not lying to me. They're simply being intentionally and willfully misleading.

When I was 8 and my parents asked, "Have you brushed your teeth?" to which I responded, "Yes." What they're really asking is, "Did you brush _all_ of your teeth?" and not, "Did you brush at least two of your teeth?" I answered that latter question. And I thought I was the smartest person in the world when I did.

coherentpony··on ROCm is AMD's priority, executive says
> I am sure some AMD folks are on HN

The person you're responding to is an AMD employee.

coherentpony··on The Federal Helium reserve is for sale
Do you realise that doing nothing is also making a choice about the state of the future?
coherentpony··on The Federal Helium reserve is for sale
Well, I can't control the past. My options are the present, and the future, and the latter depends on the former.

So what are my options?

coherentpony··on MLIR For Beginners: A series of articles on the MLIR framework
https://mlir.llvm.org/
← PreviousPage 2 of 29Next →