On the flip of this. For those getting started at their first speaking things, don't freak out when you look over the audience and feel like zero people are registering that you are even using real words :) . I've been on stage a lot: speaking, improv, plays, stuff. And there are countless times you'll actually be able to make some eye contact with the audiecnce if there aren't too bright of a stage light, and you will just see stone cold expressions staring back at you. And you will wonder if words are still coming out of your mouth. :) It happens. The human connection you get with a one on one encounter that your body kind of expects when you talk. Them nodding, or any expression whatsoever often goes completely absent in a speakers setting.
That's why "open with a joke" is so useful. It's not really for them! It's for you!. If you can hear a laugh from your audience, YOU start to chill out. Your body is like "ah. the mic is actually on, and I'm not having a stroke".
Another useful tip, if the scenario allows, try to hang out by the front row before your talk and chat someone up. That's who you want to then make eye contact with often during your talk. Their expression will be or at least seem to be warmer to you. And you'll feel some more calm looking at them than the random heads everywhere else.
I used to have a colleague that would always nod to things in meetings, and make a humming/agreeing ("mhm") sound once in a while. I used to find it a bit weird, he probably didn't even notice it himself. But then I held a presentation once with him in the room, and started to love that little quirk of his. Could always just find him in the crowd and feel some encouragement.
As for telling speakers you liked their talks, another higher up in the company used to send me a DM and thank/praise me. My guess is he did it for everyone, but still it felt very validating as a junior, both to be seen and also encouraged. It's a bit like with kids, tell them they're good at something (even if not), and they'll try to be, heh.
But I've tried to take that with me, and send DMs to people, where it maybe normally wouldn't feel natural for me. But after a while it does. And it can be even for small things, like "I like how well prepared you were for that meeting, any questions they had you could answer". You'll bet that person is well prepared the next time we have a meeting as well.
I have a huge problem with WFH speaking that I otherwise don't in a regular setting - I work at a fully remote, cameras off company and it is very difficult for me to talk in a meeting for longer than a minute without getting that feeling. Just me, in my room, talking to myself. I have no idea if anyone is listening or what they are thinking.
The audience is usually my internal team but I've never once seen their faces in the years I've worked at the company which definitely makes it more unnatural. It's the strangest feeling and I still haven't gotten used to it, but I think you just nailed why I feel like that!
I don't know. I dislike camera asymmetry (mine on, others off) but when everyone's is off it's just like the phone conference calls of olden days. I like it because I can pace around, stand up, stare at the ceiling to allow me to concentrate.
Another trick is to arrive early and try to talk to as many people before hand as possible. They’re your allies during the talk—they’ll stick around and root for the person they met.
I've installed multiple of these skills. but it seems to quickly revert to crappy writing. i've tried the adhd. the brain fried at the end of day. the "Concise" output from the harness itself. the most useful fix to me has been yelling at it "Speak English!! No jargon!!!" it goes back to more concise talking for a little while.
Anyone have a theory why this even made it to this point? the switch was such obviously a bad idea. just corporate weirdness that no one there bothered to raise their hand and be like "uh, are we really doing this?" or was there a story here that anyone knows about?
Because the bounce rate of Hide My Email addresses being deliverable is going to rise over time, by design.
Whenever I start getting spam at an address that's been leaked, I deactivate it. I've done the same with my oldest gmail account, but the work required there is notably higher.
If they were moved to plain old icloud.com, I could absolutely see a bunch of companies starting to filter out all icloud.com email addresses to avoid private relay. Either just because they’re jerks or from bounce issues.
Keeping it on a subdomain fixes that problem, to some degree. If the user is named ffjvhtu57325cjdjvg501a2@icloud.com no one is going to think that’s a real address. It’s very obviously a private one. So it’s not like they were “camouflaged.“
It’s a little odd they’re switching the subdomain though.
> If they were moved to plain old icloud.com, I could absolutely see a bunch of companies starting to filter out all icloud.com email addresses to avoid private relay.
I couldn't. "Uses Apple products" is one of the more reliable signals of willingness and ability to spend money on stuff online.
Hide My Email addresses are not just a random string at icloud.com. They use plausibly-human names, probably generated by a language model. There’s no easy-to-check pattern.
They’re not switching the subdomain. They’re keeping it the same. That’s the news.
They’re switching the subdomain for the “Sign in with Apple” sign ups, which is not the same service.
> Hide My Email addresses are not just a random string at icloud.com. They use plausibly-human names, probably generated by a language model.
why would random selection from sets of predefined strings and joining them using a "." need any LLM involvement? Oh you need to check if it already taken... maybe for that? I'd use a database though...
These days some people would even generate GUIDs with some language model, I guess...
Like for my usage there are no bounce issues with the ~400 legitimate providers that I have Hide My Email addresses from. The only ones with bounce issues are the spammers who've acquired leaked addresses that I've deactivated.
If you merely deactivate your email address rather than closing your account or changing your notification settings or whatever, then the next time a legitimate service goes to send you a legitimate email that you asked for, it will bounce. In some sense this "shouldn't matter", as in a purely P2P system the only people who would notice are the sender and Apple -- and Apple knows what is going on, so should not penalize senders the way, say, Google would if they see you sending a ton of email to their domains and they all bounce -- but people tend to use services to help send email and centrally pool their reputation (such as mailchimp) and so these services themselves then watch the bounce rate of their individual customers and either charge them more or ban their access due to the bounce rate increase.
Is Apple really stupid enough that they BOUNCE emails after you deactivate, rather than just silent discard? What's the point of bouncing unwanted emails these days? It's not like these bounces go to humans who go "Oh, gee! This address must not work. Allow me to go and figure out how to contact him!" It's just a stream of full-on spam with completely fake return addresses, and crap from email campaign software.
Maybe hidemyemail allows easily creating many accounts on a site. And some big site didn't like it.
The "Sign-up via Apple" button and creating an iCloud email yourself have a slightly higher barrier than creating a new throwaway hidemyemail email (1 API call w/o captcha/phone verification or whatever).
We might find out later this year if some site starts blocking @icloud.com but keeps allowing @private.icloud.com.
Whoever operates the smtp servers for icloud.com complained loudly enough that their job was too hard because of all the traffic, but not loudly enough for someone smart enough to hear about it, until it was announced to the public.
Of course I echo the universal: Opus 5 sucks. But what also sucks is still using 4.8. Are you all seeing this? It's like the older models got dumber just before Fable and Opus 5 were coming out? I've heard the theory that it's because 4.8 is getting put on older hardware? And I imagine that just ratchets down reasoning time then possibly?
Sometimes I'm just trying to sus out if I'm truly seeing things these days or going a little nuts :)
I have noticed very serious degradation in performance from Anthropic. I've switched away from Opus 5, but 4.8 is still much worse than how it was before Fable came out. It's constantly making what seems like obvious mistakes... I point them out, it's constantly apologizing.
I don't know why...
Is it A/B testing?
Is it load shedding?
Is it because I'm in Canada?
Is it because I'm not on the Claude Max plan?
Is it because I'm not paying via API?
Is it because I'm not paying via Bedrock?
Is it because the U.S. is worried people are distilling?
Is it because the U.S. wants to keep the top capability to themselves?
I think open models are the future. Anthropic is killing their reputation so fast. If they don't come clean I think they're cooked.
This is yet another reason why I think local models will win in the future. They're almost certainly A/B testing all sorts of opaque stuff that people have no clue about, hence the various 'How's Claude doing this session?' popups.
So what you are paying for may vary on a day by day basis, which is quite undesirable, even if their main goal is simply to make a better model. When it comes to a tool, I'd rather have consistent mediocrity than instability.
I definitely agree. I honestly cannot do tasks that require even minor complexity. Opus 5 keeps forgetting things in context as well and coding conventions. Really cannot build with CC without Fable.
RIGHT!? The data suggests this isn't true but every fibre of my being is convinced that Opus 5 High was EXCELLENT at launch and has been lobotomised since then. My benchmark is Sol High. I've been using both consistently and either Sol High suddenly became MUCH more capable - and the data does not support that - or Opus 5 became much dumber. It's so bad I can't even use it anymore.
IMO Opus 5 wasn't that great at launch, but 4.8 was definitely downgraded just prior.
Agree that Sol is a great model. I find that it's improved a bit but I mostly attribute this to its eagerness to use the harness' memory features. (I'm using it in Hermes Agent, FWIW)
For me, Opus 5 mostly sucked because of its incomprehensible writing style. Having the system prompt focus on writing in terms that are easier to understand helped a bit
I'm not an expert by any means whatsoever but deploying models is not a straightforward task. There's a lot of levers to pull and I bet when models get "downgraded" to older hardware they do so WITHOUT the same stringent quality control of the output as they do when they release it.
I don't think it's something deliberately malicious like planned obsolescence but it's more like startup culture of "just make it fit in this sprint".
I attribute intent. These are for-profit companies with more leveraged capex than any other industry in history. They have enormous pressure to optimise their limited compute. Especially Anthropic, which is far more limited than OpenAI. Of course they're pulling those levers in the background to achieve "good enough." If they can reduce memory consumption by 20% and their metrics show a 4% loss of intelligence, they might very well pull that lever. These decisions compound.
Your explanation is probably the most likely and largest contributor. Anthropic states that Opsu 5 is a "pinned snapshot". They claim the weights and model configuration are not silently updated, BUT the surrounding serving infrastructure can change, including the request router, safety classifiers, and sampling logic. Anthropic has stated that if behaviour unexpectedly changes on a stable model ID, an infrastructure update is the most likely cause.
Further, "High" isn't a fixed amount of compute. Anthropic describes effort as a "behavioural signal" and not a token budget, with the model deciding how much thinking to do. So their "High" might be "Low" now, and we would never know.
Finally, I strongly suspect some quantisation or KV-cache compression is happening. Anthropic doesn't clearly delineate whether this would fall under the pinned weights and configuration, or the infrastructure, which almost certainly guarantees it's the latter. Forgetting earlier information, poor retrieval of details, contradicting previous conclusions, hallucination, degraded instruction-following, and losing the thread during complicated tasks are all symptoms of quantisation and compression.
I'm so confused how to feel about these articles. this one is clearly AI written. But blessed by a lawyer so we think the stuff in here is accurate with facts?
I've accidentally posted an AI article before that seemed just as this was, and the community flagged that to oblivion. Rightfully so I think?
Does Hacker News need a "report AI" and we can then hide the AI articles as a filter?
This is the Claudiest article I've seen in a long time. "Defensible view." "No correction, no note, no acknowledgment." "The honest statement about X."
So the next question is, how much of this is true at all, versus just being hallucinations.
I realize so much stuff comes and goes and comes and goes again. Sony's getting rid of physical media we just read this week. But I also think there's a new bubbling up trend of folks craving real stuff that's been curated. Especially in a real place that they can see other people at.
i also really wish, Apple would make it easier and maybe obvious to gate in the store instead. there's so many apps that look like they'll actually be free, but only after a complicated wizard do you end up with the "includes in-app purchase" label which really means: you have to purchase or get nothing. for allihat, i also am gating all functionality with a subscription (at least i don't have a sneaky dark pattern wizard), but i really wish i could just do this in the app store and not the app itself. would just be so much more obvious for users. (i think iOS apps do have this "quick subscribe" mode, but mac app store doesn't have this yet...)
That's why "open with a joke" is so useful. It's not really for them! It's for you!. If you can hear a laugh from your audience, YOU start to chill out. Your body is like "ah. the mic is actually on, and I'm not having a stroke".
Another useful tip, if the scenario allows, try to hang out by the front row before your talk and chat someone up. That's who you want to then make eye contact with often during your talk. Their expression will be or at least seem to be warmer to you. And you'll feel some more calm looking at them than the random heads everywhere else.
reply