HNHacker News
TopNewBestAskShowJobs

chaxor

1,725 karma · joined August 29, 2020

submissionscomments
chaxor··on “Most notorious” illegal shadow library (LibGen) sued by textbook publishers
Researchers often write book chapters in the same day along side their research papers, so it's not like these arethat different. Effectively they're similar to review articles. Still mostly publicly funded.
chaxor··on Macron wants social networks to detect and stop VPN users
Pretty much every company, and many homes run vpns. I guess those should also be added to the list.
chaxor··on The Evolution of Vi and Vim
Mapping escape key to another key is always useful
chaxor··on Stolen van Gogh painting returned in an IKEA bag
That top image looks like a generated image.

Those hands. Man those hands are wildly big.

Camera angles are powerful.

chaxor··on How far back in time can I take my website's design?
I think it's perfect fwiw
chaxor··on Solar and batteries are going to win, and our thinking needs to adjust
Agreed that you have to squint a certain way. Specifically to see the very small part of the timeline over human history which is relevant. The period where the rate of change in global temperatures has been significantly altered is very small when zooming out. That certainly doesn't mean it's not real.

https://xkcd.com/1732/

chaxor··on When deployments are easy, code becomes simpler
Good point, this system is much more useful for scientists and ML researchers, etc. wherein the code may generate datasets over several weeks or months of compute time, but produce some output that should be directly tied to the code.

You're right though. Systems in the social media space dealing with users appending to a database typically don't have this requirement to be easily reproducible, so the local snapshots are the right choice.

chaxor··on When deployments are easy, code becomes simpler
I don't understand your statement. If a git clone of a specific commit also pulls the data that is generated from that commit hash via IPFS simultaneously, then you can always have the data at whatever state you choose. How might this have a negative effect on conserving the data? Do you mean that there are additional structures required to manage keys if the data is encrypted?
chaxor··on Opposites don't actually attract: meta-analysis study

    (\ _/)
   (o_o )
< "/s" < \

There ya go, you forgot that

chaxor··on ZSA Voyager: Low profile split keyboard
This is the keyboard everyone wants and needs. Right here.

The trouble is, I think most would be willing to maybe 'snap together' the keys and such, but not find a local PCB factory and have them create the boards and such.

Also, the cost should be only ~$100 or less if the device requires DIY.

I tried to quickly find information on how to create this board, or how to buy it directly from somewhere, or just how much it may cost to DIY, and I didn't easily see that type of information.

Ultimately, this is the best keyboard in this thread for this class (very portable, configurable, split kb). This is even better than the Voyager due to the added thumb keys.

chaxor··on Chrome now tracks users and shares a “topic” list with advertisers
The fix sounds easy to me.

Switch to a better browser.

chaxor··on Opposites don't actually attract: meta-analysis study
Or people choose partners that are more similar to their socioeconomic status from the start.

See how you can make up any story and make it sound plausible? This is why so many people don't trust studies in psychology or economics.

chaxor··on NSO group iPhone zero-click, zero-day exploit captured in the wild
The pixel is decent if using graphene is. Not sure if any system is good by default. Apple fans think their defaults are somehow more private or secure, mostly due to marketing.
chaxor··on NSO group iPhone zero-click, zero-day exploit captured in the wild
Firefox led a hand in making Rust, so I imagine if there is a browser that can make a more secure browsing experience, it would be Firefox, by making media decoders in Rust.
chaxor··on 25 Gbit/s at home, part 1
Dynamic dns is a thing?
chaxor··on 25 Gbit/s at home, part 1
You absolutely should not ever get hassled for using a product that you already paid for, that's incredibly stupid. Unfortunately the US just treats anything that isn't a corporation like trash.
chaxor··on ZSA Voyager: Low profile split keyboard
Ah, yes I know so many devs that struggle to type G. Shift+g is so hard these days.
chaxor··on US broadband grant rules shut out small ISPs and municipalities, advocates say
"640k ought to be enough for anybody"
chaxor··on When deployments are easy, code becomes simpler
If a git commit hash was directly tied to a data hash at that state (IPFS), that would be trivial.
chaxor··on ZSA Voyager: Low profile split keyboard
I really want to, but it needs to be very cheap to go through the effort of making it yourself.

A very easy 'snap together' like keyboard bought in different components should at least knock the price down to ~$100

chaxor··on ZSA Voyager: Low profile split keyboard
What on earth are you talking about?

Of course it has F1-12, it's just in a better place. It's completely programmable, so you can choose something that makes sense to you.

chaxor··on HuggingFace Training Cluster as a Service
So you can buy one H100 for our own server (if you can find it), or this.
chaxor··on Do we think about vector storage wrong?
Wow!

Not the point of the article, but it is striking to see how mastodon is way better in terms of usability/aesthetic and general UX than Twitter or 'x'.

I've been mildly interested in mastodon for quite some time from a political perspective (remove large support of big companies where possible), but I didn't realize it was just sooo much better to use.

chaxor··on A currently maintained fork of SSHFS
People have a hard time orienting if coming from js, where a 'stable' project means refactoring the entire code base to keep up with the dependencies refactoring their apis every 2 months.
chaxor··on ZFS for Dummies
I wish zfs had a 'tag' type of feature like Apple style fs. I have thought about making an object store to get some of the metadata on the fs level, rather than having to make some cobbles together solution of an SQLite DB of metadata with the files or something, but many of the object store solutions seem to be for fairly large files, rather than a few thousand pictures of family and such. Unsure what a good solution using zfs would be.
chaxor··on Writing a C compiler in 500 lines of Python
It would be great if there were more emphasis on some of the aspects that made academia so great many years ago, but unfortunately the scene has changed dramatically.

My perspective on the value dwindling comes from several graduate degrees and teaching at some of these 'prestigious' universities in the northeast.

It is of course just my singular experience, but the handful of research institutions I have worked at actually provided more of a typical 'academic' atmosphere than the universities, unfortunately. I still really want to believe that universities provide the most innovative research environment, but the incentive and operational structures don't really make it as optimal.

The reason I mention the new generation viewing its value as dwindling is that the cost isn't really 'worth it' and many of the best educational resources are actually free online now.

I'm an academic, so the fact that the investment doesn't pay off doesn't really bother me (academics typically don't care about money), but many people do want some financial advantage from education, and that isn't as present anymore.

chaxor··on Medical-evidence giant Cochrane battles funding cuts and closures
Much of the scientific literature in biology is working off of hypotheses that don't have a huge amount of evidence. There are a large number of studies that present a molecular mechanism which, if studied in depth, will not always be consistent with what is presented in the literature. This is often dealt with by appending several conditions that must be met, such that the combinatorial space grows large enough and the only study that is done that meets the requirements is the one being presented. In this way, almost anything can be stated to fit some shiny new hypothesis.

This is effectively the science based method, since it can be tracked to a, perhaps, 'plausible' mechanism (although often now lean towards more 'intriguing' than parsimonious, especially if published somewhere like Nature)

Evidence based medicine has other faults. One subfield is meta-analysis, which looks for very solid observations that are reproducible, but often lacks the complicated modeling that may be necessary to see any highly conditional effects. So the findings tend to be real, but boring and often "well-known". It is however, a good way to check if the "well-known" observations in the literature are reproducible.

chaxor··on Writing a C compiler in 500 lines of Python
You don't need a university education to do those things, just some curiosity.

The function of the university in the near future will probably just be to have like-minded curious people to discuss ideas with, and to get a better grasp of what problems need to be solved (specifically scientific ideas, rather than just applying engineering).

The prestige element (specifically of certain universities over others, perhaps not university over high school) is dwindling, and hopefully will be abolished with this new generation.

chaxor··on Recursively summarizing enables long-term dialogue memory in LLMs
It absolutely FAILS for the simple problem of 1+1 in what I like to call 'bubble math'.

1+1=1

Or actually, 1+1=1 and 1+1=2, with some probability for each outcome.

Because bubbles can be put together and either merge into one, or stay as two bubbles with a shared wall.

Obviously this can be extended and formalized, but hopefully it also displays that mathematics isn't even guaranteed to provide the same answer for 1+1, since it depends on the context and rules you set up (mod, etc).

I should also mention that GPT-4 does quite astoundingly good at this type of problem wherein new rules are made up on the fly. So in-context learning is powerful, and the idea that it 'just regurgitates training data' for simple problems is quite false.

chaxor··on Recursively summarizing enables long-term dialogue memory in LLMs
Why are people so intent on incorrectly asserting these models are Markov chains? It makes sense to use the analogy as an educational tool for exposition, but it more often seems that many use it as a way to minimize the notion that these models could ever possibly be useful for anyone. Is this just simply to make it more intuitive for others that it's a sequence model? Because it seems about as helpful as 'email is just bits' when everyone and their grandma knows about the relation between transformers, GAT, and circulant matrices.
← PreviousPage 6 of 31Next →