The building complex is always called “the Barbican”, but the surrounding map area and its tube station are named just “Barbican”. Also, the arts theatre place within the Barbican seems to be officially named “Barbican Centre” (but people always say “going to see X at the Barbican”).
That makes sense. I still find hard to imagine saying something like "I live in Barbican" if I didn't live in _The_ Barbican. But going to "Barbican" if traveling by tube would be obvious.
Interesting! I’m wondering, does caching the model state mean the tokens are no longer directly visible to the model? i.e. if you asked it to print out the input tokens perfectly (assuming there’s no security layer blocking this, and assuming it has no ‘tool’ available to pull in the input tokens), could it do it?
The model state encodes the past tokens (in some lossy way that the model has chosen for itself). You can ask it to try and, assuming its attention is well-trained, it will probably do a pretty good job. Being able to refer to what is in its context window is an important part of being able to predict the next token, after all.
Theres no difference between feeding an LLM a prompt and feeding it half the prompt, saving the state, restoring the state and feeding it other half of the prompt.
Ie. The data processed by the LLM is prompt P.
P can be composed of any number of segments.
Any number of segments can be cached, as long as all preceeding segments are cached.
The final input is P, regardless.
So; tldr; yes? Anything you can do with a prompt you can do, becasue its just a prompt.
These are great points. But I would debate the 100x point a little. And I think there are some cases where ignoring fast taps is clearly preferable.
I’m specifically thinking about phone notifications that slide in from the top – ie, from an app other than the one you’re using.
So we have two options: ignore taps on these notification banners for ~200ms after the slide-down (risking a ‘failed tap’) or don’t (risking a ‘mis-tap’).
I’d argue these are in different leagues of annoyingness, at least for notification banners, so their relative frequency difference is somewhat beside the point. A ‘failed tap’ is an annoying moment of friction - you have to wait and tap it again, which is jarring. Whereas a ‘mis-tap’ can sometimes force you to drop what you were doing and switch contexts - eg because you have now cleared the notification which would have served as a to-do, or because you’ve now marked someone’s message as read and risk appearing rude if you don’t reply immediately. Or sometimes even worse things than that.
So I would argue that even if it’s 100x less common, an mis-tap can be 1000x worse of an experience. (Take these numbers with a pinch of salt, obviously.)
Also, I’d argue a ‘failed tap’ in a power user workflow is not actually something that gets repeated that many times, as in those situations the user gets to learn (after a few jarring halts) to wait a beat before tapping.
All that said, this is all just theory, and if Apple actually implemented this for iOS notifications then it’s always possible I might change my view after trying it! In practice, I have added these post-rendering interactivity periods to UI elements myself a few times, and have found it always needs to be finely tuned to each case. UI is hard, as you say.
> So we have two options: ignore taps on these notification banners for ~200ms after the slide-down (risking a ‘failed tap’) or don’t (risking a ‘mis-tap’).
Yeah, notifications are an interesting corner case where by their nature you can probably assume a user isn't anticipating one and it might be worth ignoring input for a bit.
> Also, I’d argue a ‘failed tap’ in a power user workflow is not actually something that gets repeated that many times, as in those situations the user gets to learn (after a few jarring halts) to wait a beat before tapping.
You'd be surprised. Some users (and most software types are definitely in this camp) will learn the input delay and wait so they optimize their effort and minimize the number of taps.
But there are many other people on this planet who will just whale on the device until it does what they want. These are the same people who push every elevator and street crossing button twenty times.
I don't view notifications as a corner case. I think two factors are key:
1. Can the user predict the UI change? This is close to the static vs dynamic idea, but doesn't matter if the UI changes. If the user can learn to predict how the UI changes, processing the tap makes more sense. This allows (power) users to be fast. You usually don't know that a notification is about to be displayed, so this doesn't apply.
2. Is the action reversible? If a checkbox appears, undoing the misclick is trivial. Dismissing a potentially important notification with no history, deleting a file etc. should maybe block interactions for a moment to force the user to reconsider.
Often even better is to offer undo (if possible). It allows to fast track the happy path while you can still recover from errors.
For that reason, it's wonderful when games provide a log of actions and/or recent dialogue, so you can easily see what you missed. That kind of functionality seems less common outside games.
I am 99% sure the NY Times Games app on Android is blocking input until fully rendered on its 'home' screen where all the games are listed, and it drives me nuts. I tap on the element I want and nothing happens, I have to tap again. Maybe some kind of overlay or spinner would help signal that it's not accepting input would help? Arg.
It’s not about the tree, nor the joy and happiness it brought to many (as very few people knew of this tree, compared to how many are upset). I think what people really don’t like is the deliberate attempt to upset other people (even if it’s not them) for fun.
maybe if they didn't try to protect specific (or arbitrary) trees by law then idiots (most people) wouldn't try to demonstrate how stupid the law was by cutting the tree down. oh, and you're right, i'm not trying to hide the fact that i'm with the people not the trees - in general and in this specific case.
cutting an annoying random tree down in your annoying random neighbor's yard is just as bad as this event. this is the foundation of western civilization. prioritizing one tree over another leads to bad stuff. many have thought this through and many have argued the opposite (as you) and have sought opportunity in dismantling this principle. it's not a new debate, despite what this HN clickbait makes it seem to be.
Is defacing the Mona Lisa the same as defacing a train car, in your view?
If there is any difference, would it be related to the value of the object? Would you say a random tree in a backyard has the same "value" as an iconic tree whose destruction is causing widespread outrage?
I'm not sure what your argument is, but as far as I'm aware this tree didn't have specific, out of the ordinary, protection. I believe it may have even been on private land so the offense is exactly the same as cutting down a neighbors tree. It also happened to damage a protected heritage site, but those are hardly unique in the UK...
it had no specific protection. it was just a tree in law. but they have a silly way of 'working out' how much it was valued and that is apparently almost a quarter of a million pounds!
Presidents, royalty, maybe celebrities, there are plenty of people that deserve extra protection, there are plenty of examples of animal species with small populations that deserve specific protections, and there are chairs that have more protection than others too, the Vitsœ 620 Chair is an example, and a court recognised the chair's design as a "personal, original creation of highly aesthetic value".
It really isn’t. The comment argues that the proposed solution is unworkable and would have adverse consequences, not that it would only partially work.
Unreliable workers cause timeline problems. They can't build the roof first or do the wiring before the walls are put up. If some guy quits, gets fired, injured, it can throw off other dependant timelines. Maybe now that Bob got fired for showing up drunk and hurting someone, framing is going to take 2 more weeks, and now the drywaller contractors window got fucked up, they drop the job, and now you have to find another drywaller contractor on short notice. Maybe you can't until 6 months, so now your plasters are gone too, etc.
Perhaps it depends on the individual, but I never found it possible.
The news just made me sad, sad and angry most of the time, it's just a stream of 24/7 misery and if there's not enough misery going on locally the news will find misery from around the world to fill the run time.
What helped me is to realize: Sadness and anger come from within, not from the outside. Nobody can "make" you mad. They will do what they do, and it's up to us to decide if and how to emotionally respond to it. We are not amoebas that simply respond to stimulus. We have agency over our own thoughts and feelings. This is something I try to teach my kid, and I think it's also helped her deal with others who she would previously say "made her mad."
I think "deciding whether to emotionally respond" to something... isn't emotion?
Emotion is something you feel, not something you decide to allow yourself to feel.
Like, if I hear about someone being raped or murdered, how am I not going to have an emotional reaction of sadness or anger to that? And ultimately what use was that emotion? I cannot prevent the event happening, it has already happened, I am just a voyeur to someone else's tragedy.
Most of the news is like that. It's events that have already happened, that I can do nothing about but I'm vaguely meant to be up to date with because.... reasons? Some vague concept that everyone is meant to have an inch deep understanding of current events so they've got something to gossip about?
> Emotion is something you feel, not something you decide to allow yourself to feel.
Recognizing your emotions when you are making a decision is key. The emotions you feel will largely be outside your control but you can catch a thought you disagree with when you have it and wonder what triggered that thought. If the trigger was an emotion, you can wonder what triggered the emotion. Ask "five whys" (google it if you don't know what I mean). You have more control over this than you seem to think; you will just have to practice exercising it.
The simplest way to control your inner life is to not let whatever miserable output in. In other words turn it off.
It’s really entitled (by whom? who knows) to say that people have control over their inner lives as a response to the News being misery-inducing (according to them). Yeah. So turn it off. You don’t own the outside world your attention.
No. It doesn't matter, because just like with fire, with new virusses it is well known that these will happen regardless of people having bad intent.
You can't prevent someone with mental problems setting a tree on fire to smoke out the deamons in their head. It's a huge distraction and waste of energy that could be spent on the things I wrote about earlier.
reply