Eji1700's recent activity
-
Comment on The one-and-done pen? in ~hobbies
-
Comment on Outsourcing plus local AI will soon become more economical vs frontier labs in ~tech
Eji1700 Link ParentYou forget a few. Okay I didn't get what I want in the budgeted spend. How much more to achieve our goal? Okay we found one issue, how many more are we likely to find, and what's a reasonable...question 1 would be: "what are we burning tokens on?" and question 2 is: "how can I tell that money is well spent?" Especially if it's standard practice to burn several hundred thousand tokens across concurrent instances.
You forget a few.
- Okay I didn't get what I want in the budgeted spend. How much more to achieve our goal?
- Okay we found one issue, how many more are we likely to find, and what's a reasonable amount of time to throw at that?
This is easier with humans because you can talk to them and say "okay how close are we?". They'll be wrong some % of the time, Hofstadter's law and all that, but AI is basically a black hole in these areas. The potential spend is infinite and the potential return is unknown.
At any moment you can "hyperscale" up to $1k/$10k/$100k for possibly no or minimal return.
It's less of an issue for hobby and small projects where the AI can at least "finish" looking at the code base, or it's at least obvious when it's gone off the rails (huh this maybe 5 minute project for a human in on minute 10 of running shell scrips and searches looking at documentation...), but boy oh boy do I think there's going to be some fantastic stories in the near future of companies blowing themselves up.
It's all the dangers of cloud with even less certain gains.
-
Comment on The one-and-done pen? in ~hobbies
Eji1700 LinkI have a few fountain pens that i've been happy with. Personally I started going towards the kinds that take the standard cartridge so that I can quickly refill if needed, and if I want to be...I have a few fountain pens that i've been happy with.
Personally I started going towards the kinds that take the standard cartridge so that I can quickly refill if needed, and if I want to be fancy then I can put in a converter. I have one of these which I just have hooked on my wallet with a small notebook in it and I love it.
There's a lot of options in that world.
-
Comment on US FBI says Google engineer used internal search data to win $1.2M on Polymarket in ~tech
Eji1700 Link ParentFor what it's worth I hate that the concept even gets lipservice as it shuts down reasonable conversation so fast. Fair request but it's a bit hard for me to point to anything public and also...I promise I'm not trying to sea lion
For what it's worth I hate that the concept even gets lipservice as it shuts down reasonable conversation so fast.
do you have a decent source on this?
Fair request but it's a bit hard for me to point to anything public and also digestible. I'd say this is a decent starting point as it refutes the idea that it's wisdom of the crowds so much as an informed few, and offers interesting data on how much the market/public actually benefits from things like insider knowledge vs the insider just cashing out.
For something with a bit more glance value I think this is decent, especially if you consider the presentation of the data.
First, for most of these categories when considering how bets are "traded" its nearly impossible to go below 50%, so 58% accuracy is hard to call better than coin flipping.
You get the literal opposite of what I'm claiming with:
Accuracy metrics have improved over time as user sophistication increased and liquidity deepened. Early 2023 markets showed 67% accuracy, rising to 76% by late 2024. This improvement correlates with growing institutional participation and professional trader adoption.
However that's from 23 to 24, and the next paragraph:
User behavior analysis reveals skilled traders concentrating in higher-stakes markets, creating accuracy clustering effects. Markets attracting professional attention demonstrate significantly better calibration than retail-dominated prediction events.
has part of my claim which is that "retail dominated events" under perform, and I think there's decent evidence that 25 and 26 are more and more retail dominated (as one would expect with every single commercial).
It's hard to get to the underlying data of all of this without doing your own api calls because unfortunately "fuck sources" has become the standard, and I very well could be wrong, but I believe especially this year the "new clothes" are going to start showing on some of this.
I'll say I could very well be wrong, but I
-
Comment on US FBI says Google engineer used internal search data to win $1.2M on Polymarket in ~tech
Eji1700 Link ParentI mean....what polymarket does really shouldn't be legally viable under the law, and only is because the federal government is literally in bed with them. And even then, insider trading is its own...Seems legally viable but very much murky.
I mean....what polymarket does really shouldn't be legally viable under the law, and only is because the federal government is literally in bed with them.
And even then, insider trading is its own thing separate of if you're a prediction market or not. Google could not, in theory, take action on its own bets. Polymarket gets a % of your bet, they do not bet on the outcome, so they just want more action and don't care about the outcome (in theory. There's arguments otherwise but not a lot of proof and that's a whole other quagmire).
So if google were to do as you suggest, and allowed to, what the person did would still be insider trading.
-
Comment on US FBI says Google engineer used internal search data to win $1.2M on Polymarket in ~tech
Eji1700 Link ParentPrediction markets have a worse hit rate than sports betting for the "bettor" The value/accuracy of predictions markets has worsened as they've gotten the average gambler involved.Prediction markets are a lot more useful then sports betting which is mostly entertainment. Prediction markets are also small potatoes compared to the harm that sports betting are causing.
- Prediction markets have a worse hit rate than sports betting for the "bettor"
- The value/accuracy of predictions markets has worsened as they've gotten the average gambler involved.
-
Comment on US FBI says Google engineer used internal search data to win $1.2M on Polymarket in ~tech
Eji1700 Link ParentThis whole discussion is difficult because there's multiple overlapping nuances and yet, yes what they're doing would be wildly illegal by any standard law. So in short, while yes they WANT that,...This whole discussion is difficult because there's multiple overlapping nuances and yet, yes what they're doing would be wildly illegal by any standard law.
So in short, while yes they WANT that, that doesn't mean it's legal.
The more "savvy" reading is "This isn't your race to rig",which means if you insider trade and aren't connected, expect to go to jail.
-
Comment on Who else is as excited as I am for the Backrooms movie tomorrow? in ~movies
Eji1700 (edited )Link ParentSpoilers for cube Quentin, the cop who starts out as the "lead the team because i know what i'm doing" trope slowly goes mad/becomes evil (more evil?) and starts killing people. At first secretly...Spoilers for cube
Quentin, the cop who starts out as the "lead the team because i know what i'm doing" trope slowly goes mad/becomes evil (more evil?) and starts killing people. At first secretly (oops they didn't make it) and then straight up hunting them down and trying to kill them while they flee the cube. I think he kills more characters than the cube itself, or if not it's close.To be clear there is sort of a "who knows why they did this, and only the savant escapes" ending, but the back half is mostly "run from lunatic".
Cube 2 actually does this as well.
-
Comment on Do you prefer to 100% games, or to move on to new experiences? in ~games
Eji1700 LinkI play until I stop having fun. That is rarely, if ever, to 100%, especially in an age of arbitrary achievements and content bloat. Obviously there's also just real life time sinks and what not,...I play until I stop having fun. That is rarely, if ever, to 100%, especially in an age of arbitrary achievements and content bloat.
Obviously there's also just real life time sinks and what not, but even if you cleared my calendar and gave me weeks I'd prefer to finish a variety of games rather than "100%" 2 or whatever.
Also worth noting that for genres like fighters, the game to me can't be 100%. I literally never touched story mode in SF V, but I put 350 hours in getting platinum in ranked. I have a DISGUSTING amount of hours in Dota plus 1500 in Dota 2.
Games with arbitrary completions don't excite me
-
Comment on Who else is as excited as I am for the Backrooms movie tomorrow? in ~movies
Eji1700 LinkI'm hopeful it: Manages to handle horror in a way that isn't standard for films. A LOT of horror starts with a neat premise, but falls into "evil person chases others in confined space" because...I'm hopeful it:
-
Manages to handle horror in a way that isn't standard for films. A LOT of horror starts with a neat premise, but falls into "evil person chases others in confined space" because they don't know how to deliver on that kind of horror. SAW 2, Cube, Event Horizon off the top of my head. Something like this shouldn't, ideally, become that.
-
If it does that, I hope it does well, so that we can start to get these kinds of films consistently, or at least move to a new standard template.
-
-
Comment on Who else is as excited as I am for the Backrooms movie tomorrow? in ~movies
Eji1700 Link ParentBasically yeah. I don't think it's SCP, or if it is, that came later as these things obviously share DNA. Just an idea that started somewhere and spread on the internet, like Slender Man....It's based on creepypasta?
Basically yeah. I don't think it's SCP, or if it is, that came later as these things obviously share DNA.
Just an idea that started somewhere and spread on the internet, like Slender Man. Hopefully with better delivery.
-
Comment on An open letter to the University of California Regents requesting that standardized testing be re-introduced into admissions, >200 UC Professors signatures in ~humanities
Eji1700 Link ParentYou may very well be right. It's been quite awhile since i've looked at the data so I only glanced the charts. I'll have to dig into later then when I have time. Thanks for tolerating me here.You may very well be right. It's been quite awhile since i've looked at the data so I only glanced the charts. I'll have to dig into later then when I have time. Thanks for tolerating me here.
-
Comment on An open letter to the University of California Regents requesting that standardized testing be re-introduced into admissions, >200 UC Professors signatures in ~humanities
Eji1700 Link ParentThat doesn't much tie with what i've seen over the years. Just some quick searching gives me: https://nces.ed.gov/fastfacts/display.asp?id=1 or perhaps less reliably...That doesn't much tie with what i've seen over the years. Just some quick searching gives me:
https://nces.ed.gov/fastfacts/display.asp?id=1
or perhaps less reliably
https://worldtop20.org/worldbesteducationsystem/
But these tie with what I recall, which is that while the US "competes" in things like math, that's from a "how many good math students do we produce" not "% of total students who are decent at math", and that's caused by an atrocious skewing of the data depending on if you live in a rich education area like some CA/NY suburb, or not like....well most of the rest of the country.
Not saying the US public system doesn't have issues, obviously, but it's not like the US system doesn't produce meaningful results.
I mean, if it did, then why do we push things like the SAT/Bachelors/Masters so much as "proof of knowledge/ability to do work". The US GED is mostly worthless in comparison to other countries.
-
Comment on If you let AI do your writing, I will come to your house and kill you in ~tech
Eji1700 Link ParentA big red flag to the whole AI industry for me is the lack of a "seed token/ID" or something that you can stamp your result with to regenerate the same response. Something like "this response was...A big red flag to the whole AI industry for me is the lack of a "seed token/ID" or something that you can stamp your result with to regenerate the same response.
Something like "this response was generated and is logged as UUID blah" with a link to regenerate on X version of Y AI.
There's a lot of technical challenge i'm handwaving here, but it's very much possible to store the resulting state to help clean up confusion on these discussions (oh his linked key is 4 years ago, of course it's garbage vs yeah that was 2 days ago at a bad time...concerning), but naturally it's not in demand and not to their benefit.
-
Comment on If you let AI do your writing, I will come to your house and kill you in ~tech
Eji1700 LinkI mean, this just reinforces what most people already know. A lot of top level subjective competitions are bias/corruption displays more than anything. It doesn't mean that some of these people...One prize-winning AI-generated short story is bad news. Three prize-winning AI-generated short stories are extremely bad news.
I mean, this just reinforces what most people already know. A lot of top level subjective competitions are bias/corruption displays more than anything. It doesn't mean that some of these people didn't deserve accolades, but when you're picking from 20 great writers, at what point do you say "There is no sane/objective way to actually do this". Not that I think it's okay, but just that i'm not remotely shocked judges didn't catch AI writing or have much different to say. I'm sure on some level that judge would've written a similar response for any piece they picked.
-
Comment on Corporations can vote in some Delaware elections, judge says in ~society
Eji1700 Link ParentEhhh there's a nuance here being missed (not that it isn't mostly right) As Elon demonstrated with twitter, Delaware has been the "thunderdome" of major megacorp decisions. It's dumb but it...It makes sense for Delaware, as a state whose outsized power comes from its incredibly corporation-friendly set of laws, to go all in on corporate power
Ehhh there's a nuance here being missed (not that it isn't mostly right)
As Elon demonstrated with twitter, Delaware has been the "thunderdome" of major megacorp decisions. It's dumb but it reminds of a joke where one lawyer bribes a judge for $5k and the other for $10k, the judge gives $5k back to the second one and says we're going to decide this fair and square.
When you've got two entities with the funds to drag things out forever, Delaware was an option to actually get a ruling and a result, which is one of the more healthy things their corporate sweetheart nature provides.
I see this ruling/activity as a switch away from the reputation they've built, perhaps in response to Elon/Trump's success with "fuck the rules" and "i'll take my ball and go home to Texas" behavior.
I guess to be clear, I think this is much worse than it appears, as even people like me who might have said Delaware serves a useful point, even if going past it, are going to say this is a sign the rules just don't matter anymore.
-
Comment on An open letter to the University of California Regents requesting that standardized testing be re-introduced into admissions, >200 UC Professors signatures in ~humanities
Eji1700 LinkThe very short version of my take on all of this: A % of our issues with the college system stem from the issues with our public school system. Other countries have public school systems that...The very short version of my take on all of this:
A % of our issues with the college system stem from the issues with our public school system. Other countries have public school systems that produce meaningful results so you're less reliant on "must go to college"/"must take tests to even get into college". The US spends a ton on education and it's mostly in keeping admin and parents happy at a major cost to the children.
As such I feel that you can change whatever you want higher than that, but you're not addressing the root cause.
-
Comment on What games have you been playing, and what's your opinion on them? in ~games
Eji1700 Link ParentThere’s an entire mechanic for going back in the past and a ton of plot bread crumbs for tomoe. The very clearly had some plan ala dark souls 1 but I’m guessing the lackluster sales killed it.There’s an entire mechanic for going back in the past and a ton of plot bread crumbs for tomoe. The very clearly had some plan ala dark souls 1 but I’m guessing the lackluster sales killed it.
-
Comment on What games have you been playing, and what's your opinion on them? in ~games
Eji1700 Link ParentDid you get the sword saint route or all of them or just another? We'll mail you your "how the fuck did this not get DLC" membership number within 5-7 business days.- Did you get the sword saint route or all of them or just another?
- We'll mail you your "how the fuck did this not get DLC" membership number within 5-7 business days.
-
Comment on What games have you been playing, and what's your opinion on them? in ~games
Eji1700 LinkMechanicus 2. I like it, but so far it's hard/not clicking. There's a lot different from 1 (which i'm either fine with or think is objectively better, although I know that's not popular), but man...Mechanicus 2.
I like it, but so far it's hard/not clicking. There's a lot different from 1 (which i'm either fine with or think is objectively better, although I know that's not popular), but man can missions (which are mad up of 1-3 skirmishes with some choices inbetween) swing from "okay it's fun finding dumb combos to ruin my enemies" to "annnnnnnd we're screwed" FAST.
Think it's mostly a skill issue thing as i'm only on normal right now, but I don't want to spoil it yet by just watching someone better play.
Currently running admech as well because while I think i'll like the necron army better, I like absurdity (even if toned down) of the admech more.
Ehhhhhhhhh. I've had mixed results with the clip. I like the pen, but the clip being friction fit means it can slide off pretty easily in a pocket/bag/whatever. It's fine but its not really good.