meff

Everyone is reacting to the headline here, let's be real.

I can't. It's paywalled and all I can read is the abstract. PISA is a well known survey but I have no idea what the author did here.

It's also in the "Journal of Adolescence", has a single author, and was cited once in an Italian language work.

How can I trust paper I can't read with no citations?

Wait how can you prove correlation with 2 point? Is this a 2 point linear regression? What is R^2? Is the correlation statistically significant?

The domestic opposition exists within the Islamic Republic system and while it may oppose aspects of the ruling elite isn't really a fan of the US entering

The opposition largely exists as a diaspora with many Iranian dissenters based out of the US. LA is a big center of this (often dubbed "Tehrangeles".) The current Reza Pahlavi is in the DC area.

I mean having a debate about what switches an LLM provider should hand to the government is fine. Coercing and shitlisting a company to force them to do your bidding is not.

3 generations is the standard that India uses for a lot of its ideas of cultural continuity. If you can trace Indian citizenship lineage within 3 generations you qualify for an OCI card.

Everyone I talk to who works at those labs talks about how they're working 60+ hour weeks so it seems understandable. Personally not sure if that amount of money is worth 60+ hours of work but shrug.

Klein seemed to be pretty frustrated (as are many of us I guess) that legislation follows from harms and because the electorate isn't "feeling it" yet that there's little policy appetite for any solutions.

You can use captchas or other POW style strategies to make spam costly, so I'm curious if that's the future. Open networks gated by captcha or interstitials or micropayments.

I think Internet comments have been directionally trending in this direction for a long time now and LLM slop is the final nail in the coffin.

For at least 5 years now, Internet comments have been much less useful and more hostile than a group chat. LLM slop is eroding the little signal left.

Market manipulation can happen but there's an incentivized system of resolution which encourages each outcome holder to dispute if they feel unfair about it.

Ultimately there's no perfect way to get to the truth unless you get to it yourself. But prediction markets can be better than social media.

Behind a paywall but this seems to just degenerate to the average social media case of being wrong yet fishing for likes.

Resolution risk is a thing in betting markets but on Polymarket there's an oracle system where folks can dispute resolutions.

I think with a lot of the AI water use reporting we can see this in action. The veracity of claims is ultimately less important than the philosophical position one holds. Eventually you lose the plot with what are real facts and what are bubble opinions.

There's an added element of being "punished" for being wrong that's appealing in prediction markets.

A big issue with media bubbles is you can just say incorrect things that fit your bubble's opinions. Even if you're completely wrong you'll get all the likes from everyone in your bubble.

Yeah there certainly are good ways to use the network. It's just... a lot of work. I can use that energy to do something else that brings me more joy. I think for many that's neither Twitter nor Bluesky.

I joined the network when it was in referral beta and I really want it to succeed but the posts here just do not bring me joy. I can read other sites that have none of the problems of neither Twitter nor Bluesky. And if you're willing to put up with the Twitter algo, it can be beaten into shape.

I think the reality is users who want neither an algorithmic timeline nor Twitter's ideological bias and mobbing problem just stopped using both websites. If Bluesky is going to keep having these problems, I think it will only ever be a niche network.

The incumbent progressive data has too small data sizes to make any conclusions positive or negative. At 14 / 15, that's max likelihood p = 0.933 at a 95CI (even 90) that's indistinguishable from the moderate rate (95CI - 68%, 90CI - 72%)

Excuse me, I like to think of myself as a Big Decimal

You got this!

fil-c.org

Huh looks like a memory sanitizing runtime on top of various libcs. Looks cool.

I think Doom Emacs is the hot thing these days for emacs distros. I've been told that evil mode (which is default in both Doom and Spacemacs I think) is actually a faithful recreation of vi bindings. Tbh I've been using vanilla emacs since 2004 and with LSP and tree-sitter I see no reason to leave

Yeah I've abortively tried to switch out of emacs over the years, but it's so hard to switch. I guess editors are good examples of the Douglas Adams Tech Reaction meme.

外国人の視点で日本の旅行動画はいつもおもしろい

GitHub - meow-edit/meow: Yet another modal editing on Emacs / 猫态编辑 Yet another modal editing on Emacs / 猫态编辑. Contribute to meow-edit/meow development by creating an account on GitHub.

Come to the emacs side and you can even define your own modal editing mode 😉

github.com/meow-edit/me... is Helix / Kakoune inspired

Claude also sounds more like a STEM person and ChatGPT more like non-STEM.

ChatGPT seems to one-shot stuff occasionally faster than Claude but also goes down paths that I don't want it to go.

ChatGPT 5 vs Claude Sonnet 4.5 w/ Extended Thinking

Thought I'd try ChatGPT again after a long time of using Claude. ChatGPT just feels a lot less steerable. With Claude I know I can get it to search by asking it to. I can generally "guide" Claude down certain ways of thinking or certain idea paths. ChatGPT seems more opaque.

Too much social media posting is how people want the world to be rather than how it is. It's human to want the world to be a certain way but when you get a room of 1000 people who only want to talk about an ideal world and not the real world, you get a lot of the bad stuff of social media.

Yes arxiv.org/html/2310.06... RLHF decreases LLM output diversity. Optimization and safety pressures are a force working against the "usefulness" of LLMs.