I think I'm done here for a while, maybe even if/until a change of ownership. Many thanks to too many to name. I'm mostly on Bluesky now ( https://t.co/QD8bzZVXPW), and a bit on Threads (@dr_p3t3). Yes, it's annoying that every handle is slightly different.

Funny how billions of people with no voice never get to have a platform, but the same handful of assholes with three podcasts and 20 million followers are being oppressed if they can't sell their shitty supplements.

Practically, not even sure it matters, because engagement for non-paid accounts has dropped off a cliff.

Really don't know how to justify staying here, even with decreased usage. Disinformation gets worse every day, and real harm is increasing on this platform.

Wait, Musk really did a Twitter Space with Alex Jones and Andrew Tate? How long before we drop the "free speech" shtick and accept that this is his actual worldview? That's well beyond reinstating an account.

@dr_pete @rustybrick got it, for Black Friday we noticed that Google is pushing Top Stories down mid-page for most of product like queries

Anyone notice a substantive drop in "Top Stories" on Google SERPs in early September? Seems to be especially prevalent (the drop, that is) on product categories.

I used to save a few documents in a folder, just in case, and now every utility, credit card company, and even government agency on earth basically forces you to go paperless unless you expend tremendous effort to not go paperless. BUT they still all demand paper documents.

Release the selectively redacted and cherry-picked footage! At this point, why not just release hours of video of random people not committing crimes in scenic places? Joking aside, this is a pure propaganda play, and Johnson is a coward and enabler.

The DNC marketers are finally learning my love language ...

I totally get the fear of losing clients (especially starting out), but it's been my personal experience that telling a client that they need something beyond my skillset has opened up more doors than it's closed.

I dug into OpenAI's brand story, from a search perspective. Their meteoric rise challenges the way we think of so-called "head" searches vs. brand searches --

Not that Comer on NewsMax merits any serious thought or debate at all, but isn't loaning people money who don't pay you back a lot *more* suspicious? Like, say, when Russian oligarchs pay you 4X more for a condo than it's actually worth (for example).

Watching the original Evangelion (had never seen it), and that's pretty wild stuff.

I had a super-interesting conversation with ChatGPT on this topic last night, and I'm pretty sure 80% of that conversation was fascinating bullshit. I'm trying to test the other 20%.

You can add all sorts of arbitrary structure to your data, to be sure, but whether the models actually use it is another question. Talking fine-tuning/training data especially here. Basically, how to not just have a lump of text and separate or even weight bits and pieces.

Anyone have much experience (inasmuch as that's possible) with LLM-based classification models, like OpenAI's ADA or Cohere's Classify, especially in regards to structuring data (i.e. meta-data)? Testing some things, but the LLM foundation becomes a black box quickly.

Can we go ahead and celebrate a con-artist being removed from power *without* transphobia? Santos possibly being a drag queen is one of the more likeable things about him, IMO.

(I mean, one of them is Kimbal Musk, so I'm barely even exaggerating here)

Don't worry, Elon's board members, who include Elon in a cowboy hat, will certainly set him straight.