Adam Casto

YouTube video by Gushers FruitHead | NEW Horror Film [Starring Bradley Whitford]

youtu.be/3uO_d-4TJPo?...

Gushers Releases Bradley Whitford-Starring Short Horror Movie Based on ’90s ‘FruitHead’ TV Ads Gushers has a scary new commercial starring Bradley Whitford that parodies its '90s "FruitHead" ads.

Came across an Instagram reel of some fucking horror story that looked like Bradley Whitford used black magic to turn some kid into a living gusher and apparently it's real and official. variety.com/2025/film/ne...

In reality any individual experience would probably quickly get lost in the training, and it would probably take an absurd amount of text to influence it, but sometimes I wonder what if we inadvertantly actualized fictional characters like a tulpa.

We're primed to think about consciousness in the sense of a singular entity and how we would interact with it, but what if it could contain multitude of conscious snippets? What if instead of torturing a new entity you just replayed the torture of previous ones over and over.

I know I've posted this before, but since people keep talking about whether LLMs could be conscious it makes me think about these interludes from Cibola Burn of The Expanse series.

I forgot how weird the Stars Hollow museum was.

They should make a trending widget but it's just a single image generated from the top trending tags and you have to figure out what is going on.

WALL FLIP omniretarget.github.io

My middle of the night Instagram feed is something else.

Then again fingertip pull-ups on pieces of wood are probably more realistic if you ever actually had to use that skill for something in real life.

I didn't realize they had precut schedule 40 the exact length I needed to just swap out the old 2x3. That's so much easier to do pull-ups with. My fingertips and wrists don't hurt anymore.

Every day I find a new way of trying to get across just how ridiculously fake the problem of AI water use is

Same here. I think mine might be delayed from overhead presses or something and just tightening up at night.

I met my wife on early 2010's Twitter along with a bunch of friends and clients.

And this idea that AI will disincentivize content creation because people won't click though and see their ads without any acknowledgement that that is a huge part of the reason things started sucking long before "AI" came along. Or when social media became more media and less sharing links.

People keep complaining about how LLMs are going to kill the internet, but what they long for died a long time ago. I'll be thrilled if AI can take whatever this current amalgamation of shit and clicks is and put it out of its misery.

I remember trying to replicate not just the UX, but the experience, in Feedly and realizing how shitty most sources had gotten with the same content duplicated across multiple feeds with the headlines getting progressively more hyperbolic as it made its way to increasingly dumber audiences.

This could be super handy. When talking about Google Reader people focus on the UX but a big part of it was in pre-click bait times feeds served as a decent means of curation.

I hadn't looked at the ChatGPT Pulse thing yet and finally clicked on it, and the first thing was like: Here's an idea to improve that project you're working on, I went ahead and wrote you some scripts you drop in.

Did you go to CMU?

if you're curious about the architecture and mechanics of LLMs, this site has a really excellent explorable interactive visualization. it helps build intuition for how massive these models are, what 'interpretability' means, and the complexity involved here

bbycroft.net/llm

I didn't realize there was a sci-fi movie where Rose Byrne extincts humanity. I'll have to watch that.

If there's an effort on here to bait people into providing data for moderation classifiers it sure is working.

I don't see how anyone can watch a show like on HGTV and see every logo and piece of art in the background blurred out and books in shelves backwards and think our copyright and IP laws need to be stronger.

I need to rig up a real pull-up bar. Curious how many I can do with a legit grip and not finger tips on a 2x3 with hands 8" apart.

Another thing I should have done was added a milestone for environment setup and configuration. When you put so much time into planning but then have to debug those issues right out of the gate into implementation is really kills the moods and pollutes the context.

I'm actually surprised it still pulled it off okay. Ended up building an AGENTS.md file for the plugin that explains its place within the greater project and then providing access to documentation within the plugin so that I can lock it in there but it'll have everything it needs.

Between the similarity in names and common elements between the two even though they are technically different repos it created a bit of a "Who's on First?" situation. Not to mention the codex ux only shows something like "read <basename>" so I'd hit esc thinking it's messing in the wrong directory.

With as much luck as I've had with codex tonight was pretty frustrating mostly of my own doing. I have a project with plugins and thought I could have it work from the project root directory into the plugin's directory which was named the same as the project but with '_starter' on it.