rolleiflex

Similar, but slightly different story for me. I ended up buying it as an enthusiast ‘Apple-grade’ product where UX was there to do something I would be able to do on my own. Then they got high on their own supply and started to believe they can be as restrictive and up charging as Apple, forgetting that they’re still a product for primarily fairly technical people.

Also, for all server needs I’m running a Raspberry Pi at a single digit fraction of the ongoing power use of my Synology, and it just no longer makes sense to have this weird rare platform as my base when I could just be running things on Debian and systemd.

More philosophically, life got busy, and I no longer have the mental capacity and willingness to maintain something like a Synology. The only large content I back up are my family’s photos and I just pay Apple for iCloud monthly, I consider that to be money well spent.

A problem I had with CityMapper is that at the time I last used it, it used the distance units of the city you’re in, with no ability to change it. For example, if you’re in New York, the distances will be in feet, and feet only.

Since I already know the public transport of my my own city and that I reach out for CityMapper when I’m travelling, it’s a jarring omission. I was incredulous enough to check with support, and sure enough they confirmed as of last year at least it is indeed the case that the units cannot be changed.

Turkish also retains the hard C in some forms, the city of Kayseri in Asia Minor is also from Caesarea. However some of that has been eradicated by the more recent influence of French. An example of that is Julius Caesar, which is Jül Sezar.

I’m the maintainer of this project. While it is good to see it posted here, I should note that Aether is on a hiatus for now.

I might eventually get back to it, and I’ve been working on it since 2013 so it is a long lived work for me, but I currently have zero capacity to support it in any meaningful way.

That said, if there are any serious would-be maintainers interested I would be happy to review code and eventually distribute commit rights though.

Nothing. It's technology for the love of it.

I'm sure there are potential uses but training your own LLM would probably be more meaningfully useful versus running someone else's trained model, which is what this is.

I'm following the instructions on the post from the original owner of the repository involved here. It's at https://til.simonwillison.net/llms/llama-7b-m2 and it is much simpler. (no affiliation with author)

I'm currently running the 65B model just fine. It is a rather surreal experience, a ghost in my shell indeed.

As an aside, I'm seeing an interesting behaviour on the `-t` threads flag. I originally expected that this was similar to `make -j` flag where it controls the number of parallel threads but the total computation done would be the same. What I'm seeing is that this seems to change the fidelity of the output. At `-t 8` it has the fastest output presumably since that is the number of performance cores my M2 Max has. But up to `-t 12` the output fidelity increases, even though the output drastically slows down. I have 8 perf and 4 efficiency cores, so that makes superficial sense. At `-t 13` onwards, the performance exponentially decreases to the point that I effectively no longer have output.

IIRC Based on the latest numbers 70+% of Turks do hold their savings in foreign currency. So it’s accessible and widely used. It’s probably one of the reasons Turkey is weathering the storm better than expected, though it’s still a pretty bad situation. This corroborates with my personal experience as well.

This is through USBC which also charges the laptop.

I have a 32Gb M1 Pro as a work laptop and connecting a retina 4K external display is not a fun experience for me either. It was very surprising to see it fail to maintain 60fps in the OS’ own animations.

‘ One of Phantom’s bestsellers is a repeater crackling palm break called “Shaggadelic Mojo,” whose flower-strewn box also features a lava lamp, a beaded curtain and a disco ball. It dates back to the late nineties, when Austin Powers was cool.’

Is this a fly trap to catch old people? I thought lava lamps were mostly a 60s/70s thing.

While optical Thunderbolt cables do exist, they are a niche product. Most Thunderbolt cables in existence is comprised of good old metal wires.

https://www.corning.com/optical-cables-by-corning/worldwide/...

Here’s something better I’ve found to improve on it: you can get a $30 massage gun from Amazon with one of those round soft heads, gently put it to the back of your head where doing the manual exercise above hits with your fingers and run it at the slowest setting. It’s exactly the same thing, and it’s very effective for shutting off tinnitus. I do this when mine gets bad.

I’ve also noticed my tinnitus is inversely proportional to the last night’s sleep quality for me, so the article bears out on an anecdotal standpoint.

Same in Turkey, and I presume through most of continental Europe. Most stop signs were removed in late 90s and converted into either implicit yields (i.e. nothing, ‘don’t be stupid’ rule) or traffic lights. Right now they’re trialling allowing California style allowed right turn at red lights.

Just wanted to say this is probably the best launch HN description I’ve ever seen, and I’ve been here for the better part of the decade. Thank you for going into the real meat (ahem) of it and not being scared of writing at length.

Your cause, if successful, will hopefully indirectly result in orders of magnitude reduction in total animal suffering, so best of luck and godspeed.

Zoom does have product merits of its own, for example, I have hearing loss and uniquely sensitive to voice quality — Zoom has by far the best sound quality across all apps I tried. It remains to be seen if these kinds of advantages will make it stick, because on the other end of the balance lies a giant security risk due to China ties.

I'm very confused, the article says:

'The MTA had good reason to believe the program would succeed. Just a few years prior, it had dropped more than 1,000 Redbird trains in the ocean. They remain on the ocean floor to this day, in part because they were made of carbon steel, which helps prevent corrosion.

By comparison, Brightliners were made of stainless steel. When the subway cars debuted in 1964, they were a mechanical and aesthetic innovation. The stainless steel made the train cars lighter on the tracks, but this worked against them underwater.'

Few questions here. First, isn't carbon steel ... steel? Steel is primarily iron and carbon, so my understanding is that carbon steel is mostly just a marketing term to have at least an adjective of some sort, just like 'aircraft grade aluminium', which is in fact one of the cheapest, bulk types of aluminium (hence use in bulk in aircraft)

Second, how does regular steel fare better than stainless steel in a corrosive environment? The article says the stainless steel started to corrode from the welds, which is fair enough, but wouldn't regular steel just corrode wholesale?

Incidentally, to your tips on what you guys are doing to hire, I would add 'posting insightful commentary on Hacker News', ha — likely this post alone has gotten you a couple new interested folks.

Not the author, don't know him, no connection – but I caught a glimpse somewhere on the blog that he is currently unemployed. As an open-source maintainer, I feel like it is my duty to plug him so he can perhaps get some patrons or donations, the blog is great. https://www.bookandsword.com/support/

He is also not a software engineer or in tech in any meaningful way, so his 'tip to total income' ratio is probably off the chart compared to, say, me.

As an aside, now that everybody is asking for 'tips' of some sort, it is getting quite difficult to figure out for whom these tips are essential (i.e. him) and for whom they are just gratuities. I wish I had a good answer for this.

I've noticed that the court documents issued by civil courts in Turkey have electronic signatures with signed hashes for each of the signatories (judge, clerk and all else) in every document. To make people not freak out, they seem to have also added a PNG image of a slightly smeared generic wet-ink looking signature above the hash so it looks real on first sight. But if you look closely the signatures are all the same, and the signature says e-imza (e-signature) in cursive. Heh.

Another cool thing, the whole document itself does have a hash where you can go to the website of the ministry of justice and input the hash to verify the document. It was unexpectedly neat.

I understand, however, consider that being good for patent policies between multinational Fortune 500s does not in any way imply that it is good for anything else. In the case of two F500s with their respective lawyer armies, they are evenly matched, they're equals. Thus a licence can work even if murky because either side can make it very painful for the other to make it not work. It's a repeated prisoner's dilemma case.

For me, as the creator of the hypothetical software in question, I am vastly, vastly disadvantaged — not only am I not a lawyer, I do not want to hire one because if I hired one for every sale I would end up bankrupt. Not only that, it is possible that the software the big company uses might be the last one they would ever get from me, if they so choose, so they have no incentive to play nice.

I probably do not need to quote you the Athenians response to the Melians, you get the idea. I would recommend, if you like to get any real adoption of your licenses with the developer community, to think more like the Melians and less like the Athenians.

Hi Kyle, thanks for responding to criticism. I think you might have slightly misunderstood my main objection and comment. Your machinery is all fine and well but it requires negotiation, murky legal definitions (even if it is clearer than the usual FRAND clauses) that potentially need an adversarial process to resolve.

As an open-source developer, I do not want to sue people. When I mean holes in your license, I do not mean actual loopholes, I mean soft, bruised spots in the flesh that a skilled lawyer (which any company qualifying for the big company section of this license will have) can softly press to incur excruciating pain. I understand that you are a lawyer thus are much more comfortable with ambiguity and with the legal process, I am not — and I suspect many of us OSS devs are not either.

I cannot talk for the developer ecosystem, but I can talk as myself, as a potential customer of your license and the primary person bringing value to the table as the developer of the aforesaid software: I value clarity — any sort of discovery process where I have to hire a lawyer and work through what is fair because the license did not define it well is just such a nonstarter I don't know what else to say.

Ah, my apologies, it wasn't obvious from the second site how to get to the actual paper. I think this is fair to say there are ways that attempt to be objective in defining what is fair licensing cost, but I'd have liked to see one of these methods explicitly mentioned in the license so as to not have this question left open as a landmine.

FRAND defines fair in a very specific way in terms of pricing. For example the definition of F (fair) in FRAND is exemplified as not requiring purchase of other, unwanted licenses as a condition to the purchase of the particular license the customer wants to buy. However that does not exactly seem to be the use in this license because here it seems like fair would also carry the meaning of 'not too expensive', as I interpret the author's explanation:

' If you need a big-company license, reach out for a big-company license, and either don’t get a response, or get a clearly unfair, unreasonable, or discriminatory proposal, this is your fallback.'

'Too expensive' far as I know is not a part of FRAND, the fair in FRAND means something that is subtly different — though I am not knowledgeable enough to conclusively say FRAND includes this author's particular meaning of fair. To my best reading, it seems like it does not.

I like Mr Mitchell, and I am glad that he is attempting to do something that so many people shy away from for obvious reasons. As a developer of a fairly popular open-source tool (Aether), I have an interest in following the software licensing discussion in detail, though I am not a lawyer.

My general impression is, if you excuse my flippance, this license has holes so big I could drive the fully unfurled James Webb telescope through it. I don't mean to be dismissive, so here's an example:

`... indefinitely, if the licensor or their legal successor does not offer a fair commercial license for the software within 32 days of written request'

This is the escape hatch condition inserted for the safety of big companies that stop qualifying for the small-business section of the license. Except ... what is fair? More importantly, are you willing to spend six years in court arguing what 'fair' means? Because that is how you end up arguing what is fair in court for six years.

To be fair (ha), the license tries to firm up the term somewhat by defining the term later on as:

A fair price is a fair market price for a fair commercial license. If the licensor advertises a price or price structure for generally available fair commercial licenses, and more than one customer not affiliated with the licensor has paid that price in the past year, that is fair.

Great, but what happens if the software has not been purchased before? How is 'more than one customer not affiliated with the licensor' going to be resolved? What does 'affiliated' mean and how broad we are talking about here? Unknown, until there is a software product that uses this license, gets very popular, and then we get to see the answers in court, through the poor developer dragged through hell.

Yes, they have the Radeon Open Compute project (ROCM) but they seem to be intent on following the tensorflow from a few stable versions behind. Additionally, and likely this is not something they can do anything about, but if you are using a Linux VM instead of running it natively, ROCM does not work. I had attempted to do this to get some use of the AMD GPU that was in my Mac — no dice.

The holy grail would be a direct replacement backend that could be fed into TF, like CUDA.

I do not think rights in discrimination suits ever expire, for reasons you have cited exactly — the discriminated are usually in a disadvantaged position, don’t have the resources, time or the willingness. If an attorney told you your rights expire, was it your attorney? As in — were you paying them or was someone else doing the paying? Until you hear an attorney you pay say this, you probably have non-expiring rights.

You should contact Department of Justice’s Civil Rights division. If what you say is true and you have this on paper, you should be able to get some actual justice on this. You can start filing a report here: https://civilrights.justice.gov/

I would implore you to do this — we have unfortunately allowed this kind of thing to happen in our industry for far too long. My condolences.

Thank you! Corrected and credited.

We had, in fact, three — we called the one in Libya Trablusgarp, Trablus of the West, and he one in Lebanon Trablusşam, Trablus of the Levant, and we call the only one that we have left in Anatolia, today's Turkey, Tirebolu. Just like hashes, given a large enough empire, name collisions are inevitable.

Edit: Corrected one of the names. Thanks @egiboy.

Heads up, ‘There is a lot to try before going that route’ is a medical opinion that is hard to notice. Exactly what the treatment plan would be and in what order is entirely up to the doctor and sometimes (often) the best thing that works is the actual drugs.

I’m just conscious of this because I’ve seen this argument used to justify delaying actual treatment in favour of ‘alternative medicine’.