Tim Rossiter
445 posts


$10 to whoever comes up with a better name than "watch match" that also has a good domain that I can buy
watch-match.trainface.workers.dev
also give feedback! let's make this the best way to avoid talking to your spouse

English

After Dylan and Lennon the best lyrics have been in hip hop imho. @Nas
youtu.be/GdmtQDSZS5I?si…

YouTube
English

@meredithlistens I've got a question for you about a possible photo credit. Can you send me a DM?
English

@joehagansays The moderator should turn to her side and say, "I happen to have Victor Orban right here."
English

@joehagansays I'm not saying he's not an idiot. Just how I read it.
English

@joehagansays I just took him to be genuinely surprised. He's probably in a bubble where he hears that pronunciation. Maybe the origins are racist but he's not aware. I do have a very liberal friend in California who pronounces her name the same way also.
English
Tim Rossiter retweetledi
Tim Rossiter retweetledi

@seanonolennon @DarlingStella88 @MeghansMole I like the new record. Your singing has never been better.
English

@DarlingStella88 @MeghansMole Hey that’s not fair! The only thing I’m missing is a good singing voice, a handsome face, and the ability to write great songs and I’d be there!
English

The Sussex squad can’t even take a joke from @seanonolennon
Prince Harry fans are brain dead morons, just like himself


English

@seanonolennon Is this not tempered by the fact that you have the coolest parents in the world?
English

What a superb advert. Polarisation needs to end.
Robert F. Kennedy Jr@RobertKennedyJr
My first campaign ad! What do you guys think — run it during the Super Bowl? #Kennedy24
English

@lopp No reasoning at all? My experience is that what passes for reasoning appears to fall out of their pattern matching. Perhaps much of ours does as well.
English

Large language models are fundamentally bullshit generators in the sense that LLMs have no reasoning capability. But pattern recognition isn't a worthless attribute: LLM output quality seems to be correlated with how open-ended of a prompt you use to query them.
For transcribing audio, translating between languages, or converting code between programming languages they do quite well because you're asking a question that is a 1:1 mapping.
For writing new code, they also still work pretty well because most code can be assembled from boilerplate building blocks.
The more subjective / creative / open-ended your queries, the more unpredictable (and untrustworthy) the results will be.
English

@JeffryGonzalez Nonetheless, if I dropped out it would be because I decided that Twitter has been changing my worldview in a negative way and hence my quality of life.
English

@JeffryGonzalez This one threw me. Maybe I don't use Twitter the same way, but most of the stuff I see here is what I explicitly choose to read. I don't find myself surrounded by more or fewer bad people.
English



