HNHacker News
TopNewBestAskShowJobs

bjourne

12,624 karma · joined July 19, 2012

Email is bjourne att gmail dott com
submissionscomments
bjourne··on What garbage collection actually costs
> I disagree with this. The sentence implies that this work is done in order to make the compiler happy, where my experience is that it forces the programmer to actually get it right.

Haskellers say the exact same thing. :P Personally, I'd much rather get shit done and don't appreciate tools "forcing" me.

bjourne··on What garbage collection actually costs
Props for using a correct nomenclature. Reference counting is a "kind of automatic garbage collection. Tracing garbage collection is also a kind of automatic garbage collection.
bjourne··on The brain may be about to have its Ozempic moment
Ok, but this has nothing to do with contrastive learning. Your argument applies to all kinds of self-supervised learning methods.
bjourne··on The brain may be about to have its Ozempic moment
While there is much to suggest that sleep is important for learning, why contrastive learning in particular? It seems to me that the "deactive IO" argument could apply to any learning technique. Similar to how railroad maintenance is carried out at night because it is too unsafe to do during the day when trains are whizzing by.
bjourne··on London Underground begins scanning passengers' faces
> Or in another anecdote

No, man, anecdotes are more often than not pure misinformation. You live in a country where car theft has absolutely plummeted in the last 30 years. Look it up, I guarantee it. 100 %. I'm almost certain it's the same for bike theft too.

bjourne··on London Underground begins scanning passengers' faces
You only need 3-5 fps SD video and with motion sensing you only record when there is a customer at the self-checkout. Technically, it is absolutely doable.
bjourne··on Branchless Rust: Making a Filter 4x Faster by Removing an If
Keep in mind that this is a synthetic benchmark. The task is to remove outliers and for that the roughly 99% correct guesses of the branch predictor is perfectly fine.
bjourne··on Branchless Rust: Making a Filter 4x Faster by Removing an If
Because what makes stream compaction challenging is the loop carried dependency: the location you write to in a given iteration depends on the locations you wrote to in previous iterations. By first creating a lookup table of source -> destination locations you remove the dependency. Then you can apply extremely efficient parallel methods.
bjourne··on Branchless Rust: Making a Filter 4x Faster by Removing an If
This problem is called stream compaction and there is a wealth of research on it. The best methods use prefix scan. They first efficiently compute the index in the output array of each element that satisfies the predicate and then they gather them in one linear operation.

Also, I can tell that you are a good writer. You didn't need the LLM to "polish" your text.

bjourne··on Eight Myths on Software Engineering and GenAI
> Despite this decade-old research, many organizations still rely on lines of code as a measure of developer productivity.

No, they don't! It's easy to dispel myths when the myths are built on straw men. Dumb article.

bjourne··on Show HN: Isopolis – Isometric pixel map of SF
Wow, very impressive! But where is the bulldozer tool? I have some landscaping to do.
bjourne··on Wind and solar overtake fossil fuels in Germany for the first time
Do you know of any major German industry for which energy costs amounts to even 10% of the labor costs? It's a rhetorical question intended to engage the common sense pathways of your cognitive system.
bjourne··on AI Mania: From Tulips to Tokens
Agreed. It has nothing to do with either tulips nor mania. Just some random musings.
bjourne··on How is the Bun rewrite in Rust going?
THE SOFTWARE IS PROVIDED “AS IS”, WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.

I guess they put it in capitals for a reason.

bjourne··on Will it be possible to do LLM Pool Training?
Datacenter-scale GPGPU compute is extremely efficient compared to what you would achieve with a network of independently operated distributed heterogeneous devices spread across the world. Just imagine the overhead for ensuring that a node doesn't poison the gradients. And that's just one very small part of the problem.
bjourne··on Cruller: Bun's Zig Runtime, Continued on Zig 0.16
Yeah, this is one thing you should absolutely never, ever do as a responsible developer. It makes future debugging much more difficult (destroys git blame) and also is a huge dump on previous contributors who don't get credit for their work in the fork. If solenopsys is serious about the fork he should redo it. Throwing away history willy-nilly is such a giant red flag that I think he will have trouble attracting any contributors.
bjourne··on Ask HN: How to make AI better at writing prose?
Hard to say since you are not presenting any concrete examples.
bjourne··on Ask HN: Why are US tech workers so pro open borders?
You need to present some evidence of your claims that US tech workers does not benefit from immigration. If you can't there is no point wasting time with you since you're then a "true believer".
bjourne··on Kimi K3 Is Competitive with Fable; Kimi K3 and Fable Is SoTA
Speed and expenses are non-factors for arena.ai.
bjourne··on Kimi K3 Is Competitive with Fable; Kimi K3 and Fable Is SoTA
I'm skeptical. According to arena.ai, Fable 5 dominates almost every category: https://arena.ai/leaderboard Kimi K3 has an edge in WebDev but struggles to reach top 10 in many other categories.
bjourne··on Judge approves $1.5B Anthropic settlement for pirated books used to train Claude
> I sincerely don't understand what the point of these laws are, when the cost of flagrant violations is no more than a slap on the wrist -- these really meager sums that serve as nothing more than something to point at and say "Look, we did something!"

To create a moat around wealth generation. After all, that is the main purpose of all legal systems---to keep the wealthy wealthy and the poor poor. In this case, the settlement is chump change for Anthropic, but ensures that no upstart will be able to compete with them since they will get reamed on copyright charges. It's no different from Google Image search. They can make a product out of republishing others' images. You cannot do it.

bjourne··on Ask HN: Are Rust migrations delivering any benefits?
First you need to understand why C is a deeply flawed language. Google is your friend.
bjourne··on How we measured AI writing across arXiv, and where the measurement breaks
There is a fine line between using an LLM to clean up grammar and spelling and using it for text generation.
bjourne··on Ask HN: Would you rent M3 Ultra from me?
Nope.
bjourne··on What AI did to stackoverflow in a graph
> 1. Even if the pleasantries/signature were edited out of your question, is that so bad?

It pissed off znpy so bad that many years later they still recall on HN how irritated that made them! Now you could argue that letting znpy's pleasantries stay would have cumulatively pissed off more people in the long wrong. But I very seriously doubt it would.

bjourne··on What AI did to stackoverflow in a graph
It wasn't only LLMs it was also the downright user-hostile attitude prevalent on many sites on the SE network: https://ibb.co/WWgwNBpX

Yeah, bro, I'm not a statistics professor so I can't provide you with the "details or clarity" you need. I tried my best and if that's not enough, fuck it. Same story on history.SE: https://history.stackexchange.com/questions/60710/how-did-hi... If you're gonna require "evidence of prior research" I won't bother. The bots are way friendlier and just as knowledgeable.

The golden goose were users asking questions, not the anal-retentive content curators. SO let the latter drive away the former.

bjourne··on Ask HN: I feel like I've lost my identity due to AI
Bruh, software development isn't like boxing where your body reaches its pinnacle at 25. Self taught at < 30 you probably don't even know how little you know. You can either have your tantrum about the bots or you can continue honing your coding skills. For example, by getting in touch with premier free software projects. Many of them reject LLM-generated commits outright. For very good reasons.
bjourne··on $100 AI Music Video: Claude Fable 5 vs. GPT-5.6 Sol
I'm not claiming that no one can make a better music video for $100. I'm claiming that a random member of the HN peanut gallery can't make a better music video for $100. :)
bjourne··on $100 AI Music Video: Claude Fable 5 vs. GPT-5.6 Sol
"or i can just not do it", because you can't! To much pride to admit that the bot beats your ass every day of the week ten times over when it comes to making music videos so the cop out is "nah, man, I just don't wanna."
bjourne··on $100 AI Music Video: Claude Fable 5 vs. GPT-5.6 Sol
Eh, I'd like to see YOU make a better music video for $100 or less!
← PreviousPage 4 of 34Next →