> At some point they're going to drop the real prices in your lab
The real price for tokens at a specific level of intelligence is dropping rapidly. Between models improving and new hardware getting faster, I think prices will be going down, not up.
> Better to get a handle on it now.
I experiment with a lot of providers and models. I strongly disagree that anyone needs to start migrating now in preparation for prices maybe going up in the future.
It doesn’t make sense. Deal with a price increase when it happens, if it does.
Switching models and providers isn’t some huge lift. I can jump to a new model or even hardness and get a feel for it in a day or two. It’s not some huge undertaking that requires preparation and planning.
If prices go up, try something at that point in time. Don’t worry about it until then.
> Yeah, but you don't have to be such horrible people about it. At least give the person a few days to move their stuff.
That’s not how sanctions work. A company can’t knowingly service a customer under sanctions. If your account was termination for sanction related reasons, they have no choice but to shut it down as fast as possible once they know that fact.
They cannot continue to service it for a few days or even service transfer requests. It’s a hard stop.
I just went to zillow and typed in my zip and tried two different filters for single family homes priced between $300K-$550K:
Listings:
No HOA fee: 32
up to $500/month: 62
I'm not going to bother because this is an HN comment but I think you can imagine what happens to the results when I specify newer than say 1993 (when my house was built).
I have an HOA, but it seems defunct. Everything we do to the house, if visible, can be interpreted to mean I must get unanimous agreements, in writing, from all my neighbors.
It turns out I managed to land right between two asshole neighbors (constant trespassing, dogs, etc)
> because I’ve seen neighborhoods without them and they’re never really stunning.
Most neighborhoods I’ve lived in have not had an HOA. They’ve ranged from okay to great.
I can think of a few neighbor situations where an HOA with teeth would have been welcome. The most specific involved someone getting a big angry dog breed, neglecting it, and then leaving their gate unlocked so it got out. That one was eventually dealt with by the city, though.
Meanwhile the one HOA I lived under was pretty boring and friendly. Didn’t really bother anyone but maintained a nice common area. HOA board members were friendly and reasonable, but then again so were all the neighbors so there wasn’t much to deal with.
I think the nightmare situation is a busybody HOA that starts going after things that don’t matter, which would be hell.
> It's so bad Apple can just abort M6 Pro and Max and reuse the M5 generation for an extra year because why not, the most powerful alternatives are still playing catch-up with the base M4.
The rumor is that Apple decided to put the engineering effort into the M7 series to make a bigger leap.
Engineering resources are finite. Expecting a company to release every version of every chip at every generation is not optimal. Some times it's better to put the development effort into a bigger goal that takes longer.
I wouldn't call that bad. They're being smart about development strategies instead of forcing everything into arbitrary release goals to satsify marketing and shareholders.
That's what I want from companies who development my products. This is a good thing.
Apple's work is cut out for them; M7 and M8 desperately need a better GPU. These SOCs put up fantastically efficient single-core performance, but have some of the worst GPU compute efficiency you can buy. It also needs a better interconnect than RDMA, since Thunderbolt is snail-speed when networking several iGPUs together.
Apple Silicon is a mirror into the Mac's decision paralysis. Apple wants to unify the Mac power users and iPhone casuals onto one silicon backbone, but can't decide who gets the short end of the stick. Tim Cook especially couldn't figure out how to ship a Mac Pro while retaining "ecosystem" level control over academic and prosumer customers. Professionals and researchers had a stack they wanted to use, and Apple didn't support it.
Ternus will probably try to pivot Apple Silicon towards professionals, but he'll need to make software concessions to enter the true ARM HPC market. Ironically, that may entail Apple's rapprochement with Khronos to get the industry onboard with a CUDA stack alternative. Even a basic "inference hardware" GPGPU standard would put Apple Silicon back on the map for professional applications it has otherwise sat-out.
The professionals who use multiple gpus or outboard gear etc. Professionals used to love filling up the old Apple towers. Im guessing a MacOS workstation with several PCIE SSD cards and several GPU cards would be a welcome system for say somebody working on VFX on Marvel movie, and need to keep rendering things in the background without slowing down their workflow.
Professionals that see "rackmount" and don't stop at a one-unit rack. Industry-scale customers, not prosumer.
I don't think he intends to make a DGX Pod out of Macs, but the Studio is barely scalable. If we still had Mac Pro PCIe, then at least you could use that for interconnect. Thunderbolt RDMA is such a low pittance that Apple is basically only competing against novelty hardware like the DGX Spark.
This is a report from the developer forums where people are supposed to report these bugs.
The second response in the thread is an Apple engineer acknowledging the bug, assigning a number, and asking the person to retest when new builds are available.
If you only read the comments here you'd come away with a completely different impression of what's happening.
That's true, but I'm sorry, telling them to test it themselves every time a new release comes out because "I can't comment on known issues or on whether or when a fix might ship" is just stupid. It's a liability-reducing strategy benefiting only Apple, not a nice way to treat customers.
Also, not having a regression test that catches OpenGL being completely broken on many common hardware configurations seems pretty bad. Yes, OpenGL on macOS is deprecated, but guess what, deprecating it was also stupid. It was a legal decision made by Apple lawyers pursuing a particular litigation strategy vs Nvidia that caused Apple's withdrawal from Khronos group. It wasn't a sound technical decision.
Both of these things make it seem like the lawyers are in charge at Apple rather than the engineers.
TBH, I'm not saying "DTS Engineer" isn't an actual Apple engineer, but it feels very LLM-written. For something like that, I wouldn't want it to guarantee that a fix can or will be done in any sort of timeframe.
I hope it does get fixed, but Apple actually has a bad track-record with another deprecated API, AppleAL (OpenAL). They broke it in a very interesting way in Tahoe where mono-channel audio on Aarch64 (not through Rosetta) did not play and it was broken through all the betas. I believe it actually may still be broken today. The workarounds were either mono-to-faux-stereo channel duplication or switch to openal-soft.
OpenGL is an even bigger deal so I don't see a world where it doesn't get fixed, but I also thought exactly the same way with the previous issue. It seems that Apple does not have any older video games in their testing apparatus.
15.8.2 isn't released, but the 15.8.2 release candidate clearly is. I don't think think it's that crazy for people to be at least somewhat concerned that a fairly common important library got broken a bugfix release to a maintenance update that came out a month ago for a major version that came out two years ago without anyone internal testing it before publishing the RC.
They used to make these beta versions private so only ios/macos developers could test on them to report bugs to get them fixed before public release. They since opened it up to anyone who wants to install it which generates a lot of outrage from the peanut gallery when the beta has bugs.
The problem is that this is a well established pattern with this operating system:
- A bug or regression is identified in a beta or release candidate. People quickly come to defend Apple saying it will surely get fixed.
- The bug or regression is released. People quickly come to defend Apple saying it's not a big deal, there are workarounds, and it will probably get fixed.
- The bug or regression is now the status quo. Why are you rehashing old news? Nobody cares, this impacts only 1% of people, you're imagining things, thread locked, user banned, et cetera.
The time to complain loudly and broadly is _before_ the bug or regression becomes the new status quo for the platform.
Thanks for the context, the state of HN has been quite sad in recent years, now people (hopefully, and not bots) read the title and then go on random anecdotes, dumb generalisations and awkward tangents.
Especially on special internet forums, HN included, bots are generally paid for by professional influence campaigns so..
Yeah..
Whatever those AI's have to do to shape the conversation to get their message in, that's what they'll do. And that's going to sound pretty awkward to human ears.
An Apple engineer acknowledged the bug? That comment reeks strongly of Claude. I have a feeling that somebody did not use their brain and the response is an auto-response by Claude. Also things like attaching a sysdiagnose is useless and is well known to be hoops Apple requires bug reporters to jump through with no benefit to either party.
You also make it sound like the Apple engineer assigned the number but the number was auto-assigned when the bug report was filed. The same number was present in the first post so it could not have been assigned by the Apple engineer.
As far as I can see the Apple engineer did nothing at all. The attitude of that Apple engineer is worse than staying silent on this bug.
> Also things like attaching a sysdiagnose is useless and is well known to be hoops Apple requires bug reporters to jump through with no benefit to either party.
As a former Apple SWE that has read innumerable sysdiagnose reports, this statement is bullshit.
External bug reports very commonly have a variation of "THIS IS BROKEN YOU SUCK FIX THIS" as the main note. Even with a crash log attached that's not at all helpful. A sysdiagnose gives a state of the system the reporter has in front of them. You can't triage a bug with "FIX NOW" as the description.
I know I know, Apple has infinity dollars and every SWE should just magically fix bugs without needing to triage them first. Every external bug reports very is also a special snowflake that deserves concierge VIP serve and a dedicated SWE team to fix.
“Most bug reports are bad” and “most good bug reports aren’t handled well” can both be true.
As a macOS user, my most recent datapoint is a crash in Apple Music that I’ve submitted probably a few dozen crash reports for, along with steps that trivially and consistently reproduce the crash - from my perspective, the crash was introduced in macOS 26 last year, and has received zero attention since then.
Did you read the bug report? This is not your garden variety THIS IS BROKEN YOU SUCK FIX THIS bug. It is respectful, it has a minimal reproducer, and it is communicated clearly. There’s no reason to stall it using the sysdiagnose requirement.
And yes, this is not the first time I’ve seen Apple requiring sysdiagnose for an excellently written report. Even John Siracusa would get these sysdiagnose requests; it’s a stalling tactic being applied indiscriminately.
If Apple cannot distinguish between an angry nontechnical user that SHOUTS IN UPPERCASE and a developer that communicates professionally with an equal, that’s Apple’s problem.
You've convinced yourself that asking for a sysdiagnose is only ever used as a stalling tactic by Apple. There's not much I can say that will change your mind.
A minimal reproduction is great in a bug report but it's still not a complete in-situ snapshot of the machine.
I didn’t say sysdiagnose is only used as a stalling tactic. It may be useful if the reporter’s minimal reproducer doesn’t reproduce. But requiring it indiscriminately is definitely a stalling tactic.
People who commit fraud on this scale are simply those who start at this scale. Not always the super wealthy but often the super wealthy and sometimes those with social ties to the super wealthy.
> To even suggest that Elizabeth Holmes didn't really harm anyone because she in prison for investor fraud, and then have the nerve to call that distinction "nuance", is absurd.
I have been blown away by all of the efforts to rehabilitate Holme’s reputation and erase any of her wrongdoings.
I’m sure some of it is coming from directly seeded or funded efforts from Holmes doing PR work, but I think most of it across social media is actually organic. Elizabeth Holmes was so clearly caught out committing fraud that it’s triggering contrarian reactions among people who think they’re the ones who can see the real story behind every media narrative. Unlike other unpopular figures like Zuckerberg or Musk, most people had never heard of Elizabeth Holmes or Theranos until her fraud was revealed and became a major news story. When the media introduces a figure as a bad person, some people are so hardwired to be contrarian that they must take the opposite position.
The other difficulty with this story is that many people still don’t understand what she did wrong. This is common all over social media and even on Hacker News in this thread. I see a lot of arguments that she was only doing “fake it till you make it” like every other startup from people who don’t understand what that phrases means (it means you present yourself as a bigger, more established company to overcome resistance to being a small startup, not that you literally fake that your product works). I responded to another comment from someone who thought that because all blood tests have error rates, Theranos having problems was no different than any other blood testing company, which is a false equivalence that doesn’t even begin to compute if you know anything about the fraud they were doing.
Then there are layers of other social issues being injected into the controversy. I knew someone in person who grew suspicious that the only reason Elizabeth Holmes was prosecuted was because of her gender. They lived in a social media bubble where all CEOs were evil criminals doing frauds all the time and therefore the only reason Holmes could be singled out and treated differently was due to her gender. This led to a wild argument that going after her was unfair, actually, even acknowledging that she did the fraud. That was a conversation I exited from because there’s no overcoming that type of belief with logic.
The most interesting conspiracy theories I continue to see are the belief that Theranos technology actually was working and she was on the way to solving it all, but that the industry didn’t want to be disrupted and therefore they had her taken out. This one doesn’t even hold up to minimal research but it persists among those with deep conspiracy theory tendencies.
It’s wild to observe how often this plays out. Something about this cases triggers a lot of people’s desires to doubt and defend, and there’s a large audience for content that takes Holme’s side no matter how illogical or contrary to the facts.
> I see a lot of arguments that she was only doing “fake it till you make it” like every other startup from people who don’t understand what that phrases means (it means you present yourself as a bigger, more established company to overcome resistance to being a small startup, not that you literally fake that your product works)
Not disagreeing with the rest of your comment, but regarding this tangent: haven't there been a ton of (say) "AI" companies that quietly used humans in the background and gave the impression AI was doing the work, because their vision was that that was the future?
Not that I think that would excuse this case in any way regardless - just mentioning because of the drawn comparison.
A couple, yes, but I wouldn’t say it’s a ton of companies.
These stories were popular during the period when people didn’t understand how AI worked or what was possible. It was feasible to produce fake demos that relied on humans doing the work because few people understood how AI worked. If your demo took an input and produced an output 1 to 12 hours later with variable delay nobody questioned it because they didn’t know what AI looked like. They thought maybe that’s how long the computers had to think or that there was a single computer working through a backlog.
Now everyone has used ChatGPT and understands that actual AI doesn’t work like that, and if the task actually took 12 hours of crunching on tokens the bill would be an eye watering large number. So you can’t get away with faking your AI responses with a few humans doing the work.
There were also follow on stories where writers got confused about hybrid systems that had humans doing review and RLHF, thinking that any involvement of a human in the loop was equivalent of faking AI by having humans do the work. Most of the stories I saw toward the end of this trend were this style.
> They lived in a social media bubble where all CEOs were evil criminals doing frauds all the time and therefore the only reason Holmes could be singled out and treated differently was due to her gender.
They’re also unfamiliar with FDA-regulated spaces where it’s not unheard of for VPs and C-suite types to receive fines and jail time and bans from the industry.
All tests do have error rates and all doctors know this very closely because they look at results every day.
However saying “all tests have error rates, therefore all tests with error rates are equally bad” is a completely wrong takeaway. One of the things Theranos was doing was using traditional equipment to run the tests as a smokescreen for the fact that they didn’t have any working technology. So the results should have been similar to anyone else running them when that was the case.
But there is a problem that when a company is saying they’re doing one thing or using one technology but your actual tests could be run on unknown combinations of unknown lab hardware depending on the day and the whims of a CEO trying to defraud investors, you don’t know what kind of results you’re going to get.
Real labs have established controls and ship reference ranges for their test results because they control their processes. If Theranos was doing who knows what to keep the test results flowing to keep the illusion up for the primary purpose of satisfying investors, that’s not what you want to see out of your test lab.
> Google tells me the FIT test for cancer has a false negative rate of between 8% and 26%. That appears to be likely worse than the Theranos numbers you cited here.
You’re looking at the wrong thing. The Theranos number above is for tests they determined had to be voided because they weren’t run properly due to procedural errors they identified. Their own incompetence or fraud. And those are just the ones they admitted to!
The false negative rate for the test you Googled is a false negative rate, which is a simple fact that tests can’t reliably diagnose everything even when run properly. Tests look for markers in samples as a proxy for a condition, and those markers aren’t always present at levels that meet the thresholds even if the condition exists. Even when run absolutely perfectly, a test like that will have false negative rates.
The part you’re missing is that
the Theranos voided and invalidated tests were due to problems on top of the normal test variation rates because they were from Theranos doing things wrong, which is directly downstream of Holmes manipulating investors.
> therefore all tests with error rates are equally bad
I did not say that, nor even imply it.
Oh I don’t doubt there was funny business or fraud. What you’ve added is what the parent comment was missing; the reasons beyond the error rates.
> the test you Google is a false negative rate, which a simple fact that test can’t reliably diagnose everything even when run properly.
Yes I mentioned above it’s a false negative rate, right? And false negatives are potentially more dangerous for survival than false positives. Again, I was just reacting to the narrative about people getting scared by false positives, as if that was the primary problem. Your comment supports my claim that the error rate was not the primary problem, no?
> Theranos was doing was using traditional equipment to run the tests as a smokescreen for the fact that they didn’t have any working technology. So the results should have been similar to anyone else running them when that was the case.
> But there is a problem that when a company is saying they’re doing one thing or using one technology but your actual tests could be run on unknown combinations of unknown lab hardware depending on the day and the whims of a CEO trying to defraud investors, you don’t know what kind of results you’re going to get.
Exactly. They were using Siemens machines, but they often had insufficient blood because they were trying to maintain a pretense of "nano collection". Siemens' machines weren't even trying to pretend they could work like this, so Theranos kept trying to figure out ways to dilute or diffuse the samples they did have to get to a usable volume for the tests. But of course every experimentation there was destructive, so they had even less to use.
So we're not talking about error rates on properly calibrated machines, with sufficient blood that re-tests could be performed before needing another collection, we're talking about error rates on effectively misusing machines.
The real price for tokens at a specific level of intelligence is dropping rapidly. Between models improving and new hardware getting faster, I think prices will be going down, not up.
> Better to get a handle on it now.
I experiment with a lot of providers and models. I strongly disagree that anyone needs to start migrating now in preparation for prices maybe going up in the future.
It doesn’t make sense. Deal with a price increase when it happens, if it does.
Switching models and providers isn’t some huge lift. I can jump to a new model or even hardness and get a feel for it in a day or two. It’s not some huge undertaking that requires preparation and planning.
If prices go up, try something at that point in time. Don’t worry about it until then.
reply