HNHacker News
TopNewBestAskShowJobs

imrehg

1,431 karma · joined December 16, 2009

Senior Machine Learning Engineer at Faculty https://faculty.ai

Contact: gergely@imreh.net

Based at http://gergely.imreh.net

Taipei Hackerspace co-founder https://taipeihack.org

[ my public key: https://keybase.io/imrehg; my proof: https://keybase.io/imrehg/sigs/Vqc30FUDkAzeR6b4jTTQown3Qjned_Xju7DqvvT6DfU ]

submissionscomments
imrehg··on We still don't have a more precise value for "Big G"
Yup, that's exactly it:

- get the gravitational constant with these two known masses

- then can deduct the mass of the unknown Earth by its interaction with other masses (say the "g" gravitational acceleration value)

- then from the mass and the otherwise measured size of Earth the density pops out

More details in good ol' Wikipedia: https://en.wikipedia.org/wiki/Cavendish_experiment#Derivatio...

imrehg··on Give Up GitHub – Software Freedom Conservancy
I haven't really seen many alternatives besides what's listed, and looking for it (FOSSin my personal development, closed source at $WORK).

Imho the best options there are

- GitLab (if you want someone else to host your things, and be full featured),

- cloud provider-based repos (GCP, AWS has git hosting, if e.g. you are already using them, but it's a subset of features, needing to tie in other services to be a full replacement), or

- go down the self-hosting route if you have the capacity. GitLab is pretty easy to self-host, and there's forgejo, both mentioned in the doc.

Either way you are looking at paying for hosting, one way or another, which for commercial projects should be a baseline. The question is what to pay for, versus what the team should do itself if they can.

imrehg··on Paged Out Issue #7 [pdf]
Reminds me of PoC||GTFO.

Loads of fun in this and in that. And does indeed make me sometimes question if I know anything about programming (in a good way:)

imrehg··on I drank every cocktail
Yes, the site seems to save the checked cocktails into Local Storage (you can click some, and in your browser's inspection tools you can check. Eg. in Firefox > Inspect > Storage > Local Storage, there's a key with "cocktail-tracker").

I've checked and closing/reopening works (of course locally only, no incognito tabs, etc...)

imrehg··on Threads is nearing X's daily app users, new data shows
News TV channels in Taiwan that I usually watch, very often use videos from Threads for local news reporting (stuff sent in by the public). X-originated ones pretty much disappeared for the same use. This doesn't account for millions of users (by a long shot), but definitely a noticeable shift.
imrehg··on I maintain a 17 year old ThinkPad
I still have my Thinkpad X201 from 2011 running, and I use it as my personal machine (have an M1 MacBook from/for work).

Of course, had to replace the hard drive once or twice, replaced the whole motherboard once[0], and even though it's 64-bit, the CPU arch (Westmere) lacks some instructions that make some things non-functional (MongoDB, some Steam games don't start), and I had to limit the CPU frequency so it doesn't go into thermal shutdown. Nonetheless it's a joy to use still, and I boot it up with pleasure every time...

Thinking when will I pull the trigger on a Framework, though at least I don't feel the pressure too much just yet. :)

[0]: https://gergely.imreh.net/blog/2022/07/an-open-heart-motherb...

imrehg··on When imperfect systems are good: Bluesky's lossy timelines
This feels like an edge case.

The "reasonable limit" is likely set based on experimentation, and thus on how much people post on average and the load it generates (so the real number is unlikely to be exactly "2000", IMHO).

If you follow a lot of people, how likely it is that their posting pattern is so different from the average? The more people you follow, the less likely that is.

So while you can end up in such situation in theory, it would need to be a very unusual (and rare) case.

imrehg··on Should more of us be moving to live near friends?
Where's that Alain Bertaud quote from? I'd be interested in listening/reading more about his stuff.
imrehg··on Examples of Great URL Design (2023)
I found Notion's URL schema interesting as well. They have to contend with renames of pages, reorganisation of the hierarchy and all that. So they have something like:

    notion.so/:account/Current-Name-of-Page-:pageid 
where the name changes if the page is renamed, but the redirect works, as the page ID is unchanged. In fact, one can just use

    notion.so/:account/:pageid 
and gets redirected to the right page, or even

    notion.so/:account/Anything-else-:pageid
works too...

This is very handy in my use cases, when various Notion data is extracted into another tool, reassembled, and then needed to have a link to the original page. I don't need to worry about the page's name, or how that name gets converted into the URL, or any race conditions....

The page hierarchy is then just within the navigaton, not in the URL, so moved pages continue to work too (even if this looks like a flatter hierarchy than it really is).

I'm sure there are plenty of drawbacks, but I've found it an interesting, pragmatic solution.

imrehg··on Twilio confirms data breach after hackers leak 33M Authy user phone numbers
Besides all the other advice of using the password manager as a 2FA store as well, on the stand-alone side there is Aegis. I have good experience with it, and allows better interoperability than Authy as well.
imrehg··on Using Stockfish to identify ideal squares
Very interesting, especially when compared to shogi (Japanese chess), where captured pieces can be dropped in anywhere on the board. So for shogi players this "ideal square" calculation can be even more natural and more flexible as well: besides the "getting existing pieces from A to B", the "drop on B" is a lot simpler. No wonder that piece exchanges (so there is something in the hand to drop) are basic feature of the gameplay.

(Source: being a fan of shogi but very very very early in my learning journey, so experts would likely describe this differently.)

imrehg··on Intel announces the Aurora supercomputer has broken the exascale barrier
Looking at the two specs, interesting to see how Frontier (the first, running AMD CPUs) has much better power efficiency than Aurora (the second, running Intel), 18.89 kW/PFLOPS vs 38.24 kW/PFLOPS respectively... Good advertisement for AMD? :)
imrehg··on XZ backdoor: "It's RCE, not auth bypass, and gated/unreplayable."
That's a very good question. Likely depends on the circumstances. I don't quite know any ways of using untrusted sources safely. Maybe something where you can use temporary credentials (say 2FA), or the the likes of using AWS's EC2 Instance Connect, but there's always a problem of _something_ has to be on an untrusted location, I guess?

Having some emergency access certs in a password manager might be a good backup (and rotating it after using it on an untrusted source?).

The best way is, however, removing the need in emergencies to access a machine (e.g. more of the "cattle vs pets" way of thinking). But that's hard for sure.

imrehg··on 7.4 earthquake in Taiwan, 34km depth
Whenever I look at those reports, I keep wondering how many automated systems have to be there in place to generate all that. All the waveform records, the intensity maps, etc..., they should be all auto-generated, and likely verified by humans afterwards? Would be super curious of the IT setup and deployment of such things.

This also likely feeds into the automatic warning systems (sent to mobile phones to warn of an incming earthquake, tsunami, or something else), which is likely going to be discussed afterwards, as loads of people didn't get a warning. (As opposted to recent Chinese satellite launch where _everyone_ got the overly scary rocket alert.)

Edit: now they are saying their calculation has to project a minimum "peak ground acceleration" (PGA) of 25 (what units?) to have an alert, and a lot of the places didn't hit that, in part due to underestimating the intensity at the epicentre. I guess they will be revising this criteria, as this was overly conservative on the "less noise" side, while people are likely more forgiving in reverse (getting an alert when they didn't need one).

imrehg··on XZ backdoor: "It's RCE, not auth bypass, and gated/unreplayable."
On that last point, I wouldn't pass around the certificate to log in from multiple sources, rather each source would have its own certificate. That is easy & cheap to do (especially with ed25519 certs).
imrehg··on TOTP Codes in the Terminal
They are discontinuing Authy Desktop in 3 weeks (March 19th), brought forward (!) from August. https://help.twilio.com/articles/19753631228315
imrehg··on Ollama releases Python and JavaScript Libraries
This should be nice to be easier to integrate with things like Vanna.ai, that was on HN recently.

There a bunch of methods need to be implemented to work, but then usual OpenAI buts can be switched out to anything else, e.g. see the code stub in https://vanna.ai/docs/bigquery-other-llm-vannadb.html

Looking forward to more remixes for other tools too.

imrehg··on Who makes the most reliable new cars?
This sounded interesting, so looked it up, some more details in https://www.vanmag.com/city/people/vancouver-taxi-driver-wor...

It looks like it wasn't the 1+M km version that they bought back, but the chap was an early adopter of Prius in that taxi role, and they bought back (exchanged?) his first one from 2001:

> For those working on the next generation Prius 2004 model, Grant’s taxi held invaluable information about wear and tear.They struck a deal: Grant sent his 2001 model back to Japan for testing and had a brand new 2003 model delivered.

Still excellent story, and good on him!

imrehg··on Bandcamp support is faltering – maybe you should download your music now
I'd like to avoid having to convert it all the time (e.g. MacOS' music player doesn't seem to support that? But gonna have to check it). And the sizes are just enormous.

But yeah, as storage space & bandwidth are less and less an issue, if FLAC is fully supported across the board, the scale should be tipped. My worry comes from we not quite being there yet.

imrehg··on Bandcamp support is faltering – maybe you should download your music now
I'm always uncertain what formats to use if I were to download my digital library from Bandcamp. It seems like at least FLAC, OGG, AAC versions of everything would be pretty mandatory, given how different music players' support is, plus the for archiving. During the day I need to switch between MacOS, Linux, Android, and that's an ethernal technical sudoku...

Any other versions one should take themselves (say one of the MP3 variants), that have any tangible benefits?

imrehg··on Somafm
I'm a listener for heaven knows how long, should be close to 20, writing this while being tuned into Space Station. My PhD thesis' epigraph was from a song that I listened way to many times on Secret Agent or Groove Salad. One of my first IoT project was deploying a SomaFM player remotely to a Raspberry Pi. Back in the day 2nd hand CDs from Amazon, now I regularly search on Bandcamp for something I just heard on one of the channels, and voila, have a larger album collection from there than ever before. Such strong influence on my musical taste.

Thank you Rusty! https://defcon.social/@SomaFMrusty

I think my t-shirt is getting worn out, it's time to get some more swag.

imrehg··on Unix Time reaches 1.7 billion
Calling in from year 112 in Taiwan currently, hello, hello.
imrehg··on The laptop that won't die
I have a back-in-the-day (2011) maxed out Thinkpad X201, that was somewhere around $1.5k (whooping max 8GB RAM, etc... :)

In the intervening 12 years:

- I've replaced the disk at least 2x (going to SSD, and replacing a failed one with a current 1TB this year)

- Replaced the cooling fan (1x? 2x? I don't remember)

- Replaced the motherboard (that was fun: https://gergely.imreh.net/blog/2022/07/an-open-heart-motherb...)

- Tried to upgrade the wifi card for newer standards (that's a fail, as it seems to be locked down in the BIOS what cards the machine can use)

- Managed to set a max frequency to the CPU so now I don't have a thermal shutdown, it works rock solid and still reasonable performance.

It sounds like a lot of pain, but it's still my favourite device to work on ever since, ArchLinux humming away with XFCE. Not sure what my next machine will be, but this current one will give a run for its money...

imrehg··on LocalPilot: Open-source GitHub Copilot on your MacBook
Maybe it depends on the work people are doing. I am doing a lot of data engineering where I can be often tabbing through while adjusting bits that are off, or doing Python portion by portion, and depending on the complexity it does go from "finish this string I'm writing" to few lines worth of code.

I'm often surprised that how quickly the model discovers reasonable patterns (even across files as well, that's often necessary to be correct).

With the diffs I find that by the time I describe things in a way it has a chance of working, I might have just written the whole thing myself. Especially as the diffs are often need correctly. With tabs that correction is part of a fast feedback look, with diffs it's so far a slower loop and just more awkward.

Of course, with changes to workflows one or the other can shine and if there's an interface that's faster than typing the instructions out that might just supercharge things for the diff/chat type too.

imrehg··on LocalPilot: Open-source GitHub Copilot on your MacBook
It looks like that needs an NVidia GPU which I don't have, or a cloud service, which is out of scope for me.

The autocomplete + chat only seem to work on their own model, all the other models currently one or the other.

Nonetheles it does look interesting, cheers for the suggestion!

imrehg··on LocalPilot: Open-source GitHub Copilot on your MacBook
I haven't tried that particular one yet, I might not have enough RAM to run it, though, but gonna give it a try. It seems like one of the larger code-focused local models. Cheers for the suggestion!
imrehg··on LocalPilot: Open-source GitHub Copilot on your MacBook
I guess similar to ollama (recently discussed: https://news.ycombinator.com/item?id=36802582) which also has support for code-focused models (see: https://ollama.ai/library).

I tried pretty much all of them with Continue in VSCode, and it's a bit hit and miss, but the main difference is the way the workflows work (Copilot is mostly line completion, Continue is mostly chat or patches). So the main value add here for me would be a more Copilot-like workflow (which seems to align better with the day-to-day experience I has so far).

imrehg··on Show HN: See library availabilities for your Goodreads want-to-read list
This is amazing initiative, thanks for doing it! I like Libby is an amazing "best kept secret". Also Goodreads used such as this "database" is exactly as it should (at least partially) be used. If I could also connect it with my local _physical_ libraries as well + book stores, etc... it would make discovery so much easier - and I had even bigger piles of books to read than now. ;) So cheers for this!

On the user experience feedback:

- I keep getting "failed to load books", though some stuff did succeed (with libraries I don't care about, not my actual ones), so I wonder where the issue is (The query is https://projecttbr.com/?goodreadsProfile=3669238-gergely&lib... for reference; I'm guessing the libraries are giving the grief)

- "loading" is good, but it would be good to have some kind of indication of progress (is it loading my want to read list? is it going through checking the books? what is it doing?)

- it's nice to see that the settings are saved in the query parameters so one can just bookmark the page and don't have to fill in things again. This could be made more explicit, though, as it might not be obvious or guessable to people

- choosing the ebook type got me stumped a bit, even though I do use Libby for ages and e-readers of all kinds: "which one would I need to choose to be able to read in the Android app?" etc... This might not be this page's issue, but some clarity and UX help could be welcome!

imrehg··on AI's $200B Question
If only there was a non-X-thread version of this that was readable (as the threads are now hidden for people not logged in).

Of course, the author is free to choose to publish his views in any way they see fit. Except this way it gets a "meh" from me at most (and even that only because of the intro in this post here), since it's functionally hidden, as good as if it didn't exist.

imrehg··on Grouping digits in SQL
Here's a more up-to-date link to [1] as it is now published as RFC-9413: https://www.rfc-editor.org/rfc/rfc9413.html#name-virtuous-in...
← PreviousPage 2 of 18Next →