HNHacker News
TopNewBestAskShowJobs

carlmr

7,934 karma · joined September 1, 2017

submissionscomments
carlmr··on Show HN: I trained a 9M speech model to fix my Mandarin tones
他是 is tāshì which doesn't transform I think. Did you mean to write 你是 nǐshì? I think that transforms differently though. With the half 3rd tone only dropping.

The classical example is 4/4 不是. Which goes bùshì -> búshì.

Or 3/3 that becomes 2/3. E.g. 你好 nǐhǎo becoming níhǎo.

The 1/4 -> 2/4 transformation I think is specific to one. 一个 yīgè becomes yígè.

carlmr··on Any application that can be written in a system language, eventually will be
>In particular, an error on one line may force you to change a large part of your code.

There's a simple trick to avoid that, use `.clone()` more and use fewer references.

In C++ you would be probably copying around even more data unnecessarily before optimization. In Rust everything is move by default. A few clones here and there can obviate the need to think about lifetimes everywhere and put you roughly on par with normal C++.

You can still optimize later when you solved the problem.

carlmr··on Scientists find a way to regrow cartilage in mice and human tissue samples
>Year Of The Linux Desktop

After Win11 Microsoft really did all they could to get us there this year.

carlmr··on Show HN: Cachekit – High performance caching policies library in Rust
I think so, too, it's beautifully designed in many ways, but this seems like an oversight.
carlmr··on Show HN: Cachekit – High performance caching policies library in Rust
On your example, without reading into the implementation, I'm wondering if the comment is wrong, or if the comment is telling us about a hidden default, but then what does the 2 mean.

    // Create an LRU cache with a capacity of 100 entries
    let mut cache = LRUKCache::new(2);

Why 100? Why not 2?
carlmr··on Ozempic is changing the foods Americans buy
Thank you, this is what I constantly say. For population statistics, BMI is nigh perfect, since it's much easier to gather than more accurate data points, and the number of exceptions are super small.

I know very fit people that still fall well in the BMI 20-25 range. Most around 23. You have to be very focused on natural bodybuilding for years if you want to become an outlier on BMI.

Or some combination with being super short or super tall. But this again affects a tiny minority.

carlmr··on Ozempic is changing the foods Americans buy
Nobody did, please don't derail the conversation. I did even acknowledge that Germans aren't that slim.
carlmr··on Ozempic is changing the foods Americans buy
I got some similar responses on my post above. HN discussion quality is really suffering recently. Wondering how many LLMs are involved here at this point.
carlmr··on Ozempic is changing the foods Americans buy
If you don't read the main part of my reply, why do you even reply with snarky nastiness.
carlmr··on Ozempic is changing the foods Americans buy
The question is how you calculate or define excess. For one, excess calories aren't 100% stored as fat. We're not that efficient.

Additionally the fatter you are the more calories you use at rest. So there's a point where if you consistently eat too much you'll stop gaining weight.

The biggest source of error here will be the calories in the garbage bin though. I wouldn't be surprised if a lot of this is stored in the waste disposal, not the waist disposal.

carlmr··on Ozempic is changing the foods Americans buy
I didn't dismiss any benefits. But so many people needing an anti-addiction drug to get off addictive foods IMO should raise alarm bells.

The food producers need to be sanctioned. It's unsustainable for a whole nation to be on these expensive drugs.

If you think an outside perspective is irrelevant I'm also not sure why you don't just move on and not comment.

carlmr··on Ozempic is changing the foods Americans buy
We have quite a few obese people. Not as many as the US, but not few either.
carlmr··on Ozempic is changing the foods Americans buy
There may be, but I haven't seen anyone lose significant weight either since this started.
carlmr··on Ozempic is changing the foods Americans buy
>The share of U.S. households reporting at least one user rose from about 11% in late 2023 to more than 16% by mid-2024.

I was wondering how you could get such a high impact overall. But it seems one in 6 households are on GLP-1 drugs in the US.

In my friend circle in Germany I don't even know one single person on this stuff.

It's insane to me that so many people need these to get off the processed foods killing them in the US.

carlmr··on Don't fall into the anti-AI hype
>It’s not readily apparent at first blush the LLM is doing this, giving all the answers.

Now I'm wondering if I'm prompting wrong. I usually get one answer. Maybe a few options but rarely the whole picture.

I do like the super search engine view though. I often know what I want, but e.g. work with a language or library I'm not super familiar with. So then I ask how do I do x in this setting. It's really great for getting an initial idea here.

Then it gives me maybe one or two options, but they're verbose or add unneeded complexity. Then I start probing asking if this could be done another way, or if there's a simpler solution to this.

Then I ask what are the trade-offs between solutions. Etc.

It's maybe a mix of search engine and rubber ducking.

Agents are, like for OP, a complete failure for me though. Still can't get them to not run off into a completely strange direction, leaving a minefield of subtle coding errors and spaghetti behind.

carlmr··on Alien: Braun Aromaster KF 20 Coffee Makers (2012)
They have, kind of. Check out Gaggiuino. Sadly not really open source (Gen3) but you can get a cheap $500 Gaggia Classic and add fine grained control with an Arduino and display: https://gaggiuino.github.io

Then there's the really open source Rancilio PID Clever coffee project:

https://github.com/rancilio-pid/clevercoffee

Different machine, similar approach, a bit more DIY since you can't buy a kit.

carlmr··on Start your meetings at 5 minutes past
Additionally, most meetings are worthless drivel from an IC perspective. The off-topic/drilldown is usually when ICs actually discuss topics relevant to them and get into a level of detail on issues that actually helps further the project.
carlmr··on Alien: Braun Aromaster KF 20 Coffee Makers (2012)
>With the cost reduction opportunities of mass production, I don't see why they should cost more than $500.

I think it's more of a supply and demand thing. People spend a lot on their hobby. And it's not as much mass produced as some other items.

carlmr··on Maybe comments should explain 'what' (2017)
The within settlement window might still be a good addition to the method name. Since that is something I can at least Google.
carlmr··on Was it a billion dollar mistake?
>It is generally desirable to keep invalid objects properly typed

I think that's the bigger issue in Java. `int` can't be used in generics, so now you use the boxed type `Integer` which needs to be checked on each use, which depending on the usage is a lot of work.

If Java had proper generics without this workaround, we wouldn't have nearly as many issues.

carlmr··on Children and Helical Time
>I find that the years that I spent on art (playing around, learning new things, not taking other peoples' orders) lasted longer than the ones I spent doing software development for money.

>I personally don't find the logarithmic experience theory convincing.

I think this tracks though. When you do art or other things you can explore different things. Doing the same thing for 40-60h per week is just not such a varied experience.

carlmr··on Kidnapped by Deutsche Bahn
>Higher incomes have historically been a symptom of productivity

If I go to Zürich I get a burger for 30Fr that I can get in Southern Germany for 15€ and in Berlin for 8€. That is with roughly the same quality.

I'd say past productivity leads to network effects and investments in one area that boost local salaries and decouples them quite strongly from current productivity.

My previous company had a per-dollar extremely unproductive location in silicon valley. The people there weren't at fault. You don't magically become more productive because you live next to SF.

carlmr··on Kidnapped by Deutsche Bahn
>invest more relative to their passenger load.

For Switzerland does this account for the almost double salaries or only absolute spending?

If you spend 1€ in Switzerland I imagine you get much less work output than for 1€ in Germany.

carlmr··on Experts explore new mushroom which causes fairytale-like hallucinations
>so we farm incredible quantities of tobacco to extract its poisons.

Evolutionarily the best survival strategy.

carlmr··on Seven diabetes patients die due to undisclosed bug in Abbott's glucose monitors
>the possibility of it being software, hardware or QC

There's a certain overlap here. It's not completely orthogonal. Having worked on safety critical systems before a lot of effort is put into detecting hardware errors in the software. E.g. random bit flips, ALU hardware issues, RAM writability issues, hash check of the loaded software being ok, plausibility check with (partually) redundant sensors.

You can detect a lot of hardware/QC issues on the software level. While it's still a hardware issue, better software can sometimes at least detect it

carlmr··on Over 40% of deceased drivers in vehicle crashes test positive for THC: Study
Asa foetida is way worse than skunk
carlmr··on Over 40% of deceased drivers in vehicle crashes test positive for THC: Study
That's fine as well. I just don't want to punish you for life for small infractions every once in a while. Humans make mistakes.
carlmr··on Over 40% of deceased drivers in vehicle crashes test positive for THC: Study
>it turns out that they got dozens of speed camera tickets per year

To me the answer is quite simple for any of these. Treat repeated small infractions like bigger and bigger infractions. E.g. double the cost every iteration if it happens within a specific time frame.

Ok, you speed once? $100. Twice $200. Thrice $400. And so on. We only reset if you don’t reoffend for any speeding in 5 years. If you want to speed 20 times in 5 years, ok, go ahead. You pay $52,428,800.

Bonus points for making it start at something relative to your salary. People will stop at some point out of self-preservation.

If you don’t believe high fines work, drive from Switzerland to Germany. In Germany the Swiss have no problem speeding, because the fines are laughable. While south of the border they behave very nicely on the street.

You could extend this to other crimes. Google and Microsoft happily pay fines, since it’s cheaper than what they make from breaking anti-trust regulations. If you doubled it on each infraction they would at some time start feeling the pain.

carlmr··on Pricing Changes for GitHub Actions
>Gitea scales really badly with large repos in my experience.

Isn't it written in this super scaling language that everybody says scales super well?

What is the problem with it?

carlmr··on When did the job market get so rude?
I've also been hiring before and if my colleague told me he had a list of people he didn't want hired because they didn't write him back, I'd laugh my ass off and continue on with my day.
← PreviousPage 3 of 34Next →