What would be the worse non-semantic alternative? Regex?
LLMs are the holy grail for getting beyond string matches. I would hope one was used to solve such a problem, otherwise that would be a poor product, right?
I did not investigate the product, but my point here is irony. The correct solution to implement the TFA product is to use an LLM.
73% of users vote prior to reading TFA, according to this research. (I am sometimes guilty of this myself)
We live in a world being dimished by confirmation bias, but this isn't a new thing. Those who wrote/approved the headlines always had more power than those who wrote the articles.
> In the present work, we introduce and make available a new dataset containing the activity logs that recorded all activity for 309 Reddit users for one year.
However, tracking over a year might make the subject forget about the whole thing, and act naturally. As far as HN vs Reddit, not much difference really. I mean that as more props to reddit users than anything against HNers.
Here is my naive take on sovereignty, and how everything should work in the new "USA decided to kill its own dominance, and attack its allies" world. The world is now balkanized, let's live in that reality.
1. Almost every country has amazing universities with software tracks. A big issue is that universities often don't prepare their students for the real work, aka making and supporting products.
2. Governments should greatly favor products created by the students of their own universities.
The goal of every country should be to foster a sovereign software flywheel. Anything else seems pretty darn silly.
I ve had exactly same idea for many years now, but apparently that's not as obvisous to others
Developing new software? Universities! Maintaining/migrating old software? Universities! IT counseling and advise? You won't believe it ... Unive okey i stop here you got the point
>The goal of every country should be to foster a sovereign software flywheel
The very simple economic problem with this is that autarky does not increase aggregate output. Saying "I will do this myself", always requires the qualification "at the expense of what else that I'm not doing?"
The adaption to a reality of a balkanized world for small countries is, like Singapore does successfully, triangulate between large countries, specialize what you're good at, be pragmatic and flexible and strategically neutral which makes big powers compete for you without drawing hostility, rather than trying to become 'sovereign', which makes you poor and a target.
Don’t forget context. Basically I have 2 billion input and 1 million output. Every prompt you do, sends back the whole thing again and again. Let’s say you have 500k context used, you send 10 messages is 5 million. 100 messages 50 million.
Use 5 threats is 250 million.
But how is it even possible (bad harness?), or wise, to send 500k or 1M tokens per call? Regarding cache, how are you not hitting the 1hr cache? Also, start new chats early and often!
I have been "agentic coding" since Sonnet 3.5 and after this paper came out, it became my bible:
I personally don't count cached hits as $used... Neither in my harnesses, nor in the LLM-enabled apps I create. A cached token cannot be counted 1:1 as to a non-cached token, that would be silly.
Wait... when some Claude 5x/20x users say they are getting "$2000 of tokens for $100," does the 2k value include cached tokens, counted at the same $/token either way?
We cannot be this dumb as a community, can we? I must be wrong/misunderstanding..
I'm a fairly moderate user, never hit any kind of usage limits, but I used 44 million cache create tokens and 1.5 billion cache read tokens, which ccusage estimates would have cost $990, and calculates the different categories separately.
One one hand, it's political news, which I abhor on HN. On the other hand, international travel out of SF provides underlying support for the tech industry and without it, would cause deep headaches. On balance then, yeah, I don't mind this on HN.
LLMs are the holy grail for getting beyond string matches. I would hope one was used to solve such a problem, otherwise that would be a poor product, right?
I did not investigate the product, but my point here is irony. The correct solution to implement the TFA product is to use an LLM.
reply