If the models are designed around it, and not resorting to compression to get to higher input token lengths, they don't 'fall off' as they get near the context window limit. When working with large codebases, exhausting or compressing the context actually causes more issues since the agent forgets what was in the other libraries and files. Google has realized this internally and were among the first to get to 2M token context length (internally then later released publicly).
I quit social media many years ago and to answer the question: No, I just watch Youtube. If I could stop watching Youtube, I'm totally sure I'd finally be able to read books again /s
The problem is the award delay. In Youtube, I get my "award" in 10 minutes max. Starting to enjoy a book requires 1-2 hours investment, and the award can be anything between 1 and 10 in a scale of 10 (while median being more like 7), and Youtube is 3-6 with a rare 9.
I read a lot of self-improvement books lately, or heard to be honest. They didn't help me start reading. Atomic Habits came close.
I have (diagnosed, yet untreated, because of side effects) ADHD though. So maybe not the typical experience. I also couldn't read much (or do any homework) as a child.
Currently trying to stop myself from starting with short videos.
Social media never really caught my interest, so quitting made zero difference to anything. I do want a fair bit of YouTube for periods of time, then almost nothing, and instead read a ton of books for a few months and then switch back.
I really want to swap out the YouTube part of more programming, but I find that I need at least an hour or so of quiet time before my brain sort of switches mode and can start enjoying it, so it's harder to get started on and disturbances quickly snowballs into not getting anything done.
Maybe you didn't quit really. "Social media" is a rather misleading term now. With TikTok success, almost all of them bet on short video format. So really Youtube is in the same group as Instagram, FB, TikTok, and Xwitter.
The only thing I need to stick to a good sci-fi book is:
- have it on my kindle
- actually have read a bit. a page is enough.
- phone out of the bedroom (the hardest)
> I quit social media many years ago and to answer the question: No, I just watch Youtube. If I could stop watching Youtube, I'm totally sure I'd finally be able to read books again /s
You /s, but when I quit the internet completely, I did become a voracious reader of books. I also spent hours practicing piano. And I went to bed on time.
Like you I also have diagnosed and untreated ADHD. But at this point it feels like it's a misdiagnosis and I'm simply incompatible with the internet.
You'll probably think that I'm sort of maniac or that I'm messing with you but honestly, I feel like I'm part of the internet. I mean it. I'd probably feel like a screw that fell of a large ship. The ship is unaffected, while I'd be collecting rust at the bottom of the ocean.
My jokes sound like reddit. I give HN reactions to new startup ideas. I review code like I'm in front of a large crowd from GitHub. I make meme references. I don't play games, I watch other people play -> less stress.
On the other hand, I want to read books! I want to practice the piano! (See, I bought this nice YAMAHA keyboard that's collecting dust).
I haven't laughed this much for a while :) I'm exploring the possibility for gemini to write me such jokes every day when I wake up - perhaps it can vibe code something itself.
I think it's not about own versus someone else's money.
Hardware is usually a small piece of the financial puzzle (unless you're building a billion dollar AI datacenter I guess) and even when the hardware price quadruples, it's still a small piece and delivery time is much more important than optimizing hardware costs.
As someone who abuses gemini regularly with a 90% full context, the model performance does degrade for sure but I wouldn't call it massively.
I can't show any evidence as I don't have such tests, but it's like coding normally vs coding after a beer or two.
For the massive effect, fill it 95% and we're talking vodka shots. 99%? A zombie who can code. But perhaps that's not fair when you have 1M token context size.
I could try to explain that most jobs are way more nuanced than just 'failing and deserving to be called a monkey' or 'not failing.' Or, I could just call you names for not seeing that, you could call me names back, and we can keep doing this forever.
Your argument is lacking nuance, declaring that the criticism being levied here must be a simple binary.
The specific error they are criticizing is extremely egregious, akin to builder declaring a house without a roof complete. “failing and deserving to be called a monkey” is a criticism being levied against a 0/100 level mistake, not a mere minor mistake as you are claiming.
While it might be desirable to use less colorful language, it is frankly challenging to express the sheer level of grossly incompetent organizational ineptitude on display here in a reviewed and delivered product actively causing negative customer impact for literal years which is trivially fixed and yet has been ignored.
Customers of Github should be infuriated that Github gleefully foists such utterly defective software upon them. It is hard to get that across in dispassionate writing.
> Your argument is lacking nuance, declaring that the criticism being levied here must be a simple binary.
That isn't my argument. I am arguing against the idea that there is an "objective" threshold of failure where, once crossed, it becomes acceptable to call people names.
> Customers of Github should be infuriated that Github gleefully foists such utterly defective software upon them. It is hard to get that across in dispassionate writing.
See, while it has its bugs, I don't see a major problem with GitHub as a software product (setting aside the monopoly concerns). I encourage passionate discussion, but calling people names doesn't communicate passion; it communicates impatience. It suggests you don't have the patience to actually make a case for something you're supposedly passionate about, so you're choosing a shorter, more aggressive form instead.
Is it silly though? With enough linguistic archeology I bet you can make this entire comment I'm writing right now extremely problematic and offensive. The linguistic treadmill means exactly that older terms change meaning back in time. They also change meaning FORWARD in time, meaning your inoffensive terms today will almost certainly be offensive in the future.
It's also the case that offense is language dependent, which is always funny when Americans hard ban certain words on chats and then Swedes can't use the Swedish word for "end" because it's spelled like a slur in English.
The whole terminology in IT could be turned upside down because it can be quite offensive if people ignore the context, so it is not limited to processes. There are utilities like "man", "finger", etc. that could come across as offensive too, to some, with no context-awareness.
Today it is "master" -> "main", tomorrow the whole IT terminology.
There are many PRs on GitHub with regarding to these, by the way.
... also what about pins? Slave and master pins! Must be about slavery, right? No, it is not, not at all.
In any case, who made the association of the git branch "master" to slavery? It is absurd. People need to take the context into account.
It is based on nothing. It is not intended to be offensive, and it is not intended to be about slavery. Similarly how master and slave pins are not either, or how blacklist and whitelist are not about race either!
I grant it's not nothing, but I think it's not enough of something to make changes over it. Thinking of a master record or similar is the natural reaction when you learn about the terminology, and most young people have never used bitkeeper, so unless you go out of your way to explain why this is "bad" most people won't even know, so what do you gain from it?
Of course, but they do not care about that. They made the association, and now they are being vocal about it. I am pretty sure most of us never made this association or attribution. I have never thought about slavery until they told me their own associations to it.
I am pretty sure master / slave pins were not intended to be offensive, nor attributed to slavery. Similarly with the git "master" branch.
rules.ini and rulesmd.ini are the 2 text files, in my life, I've spent the most time with.
I'd probably lose another week if I had easy access to RA2 modding. Or let's say "experimenting and watching the AI burn" not to disrespect the real modders.
reply