Pinned Tweet
Replying to @teleoflexuous
If you prefer a reasonable format here's the same thing as an article substack.com/inbox/post/9308…. For all of you who can read only by promising themselves it's only one more tweet, here goes:
weren't these people supposed to have heard of Girard?
Kith CEO @RonnieFieg says social media algorithms have flattened taste:
“Everyone now wants the same thing. You used to be forced to have your own opinion.”
His influence came from riding across Queens, watching completely different cultures get on the train, and walking into stores before the internet had already told everyone what to like.
“Every other stop would be a different culture hopping on the train.”
“What I got to see became my world. That was my influence. I got to see so many different styles, people dressed in so many different ways, and that’s when the sense of individuality was at an all-time high.”
“Before you were influenced by your algorithm, which now, a lot of people want the same things because it’s based off what they see.”
“Back then, when you went shopping, you were forced to have an opinion because you wouldn’t have seen everything.”
“Today, people will walk into a store and know about all the products before they walk in. They’ve already seen someone wearing it, a photo of it, a lookbook of it, a campaign of it.”
“Back then you were just forced to have your own opinion.”
This video is larger than Cloudflare's 512 MB cache, so it can't be played through. More donations are needed to cover a larger cache. Donate
on a resume, does 'claudish' go into programming languages or foreign languages?
designer hating you when as a customer you had too strong and detailed opinions was load bearing in the same way having to manually apply to jobs and read resumes was
why does anthropic want me to use claude code so much? I don't mind but I don't get it, they own their cowork harness exactly as much
HOW DID HE KNOW
nitter.cf/Aella_Girl/status/2095…
Replying to @thoth_iv
PLEASE HOLD ALL THE STATS ARE BORKED RIGHT NOW they should be fixed in a few hours
Anthropic learning right now why OpenAI is giving resets could've been avoided by hiring a single person with WoW experience in product.
the moment I switched from GLM 5.3 to 5.3 turbo in conversation it started thinking in Chinese. I think that's cute
getting suspicious that difference between Codex and Claude writing comes from sample bias of audience. different flavours of bad, arguably Claude's more easily visible
I was somewhat surprised it was possible to sell removal of free tasks every 5 hours as 'removing limits', but alas
looks like Codex limits may have been quietly nerfed.
back in June, SemiAnalysis tested the ChatGPT plans and estimated their max API-equivalent usage at roughly:
$20 Plus → ~$700/month
$100 Pro 5x → ~$3,500/month
$200 Pro 20x → ~$14,000/month
now a Plus user tracking with NerfTrack says their weekly limit dropped from around $160 → $80, basically a 50% cut.
one Pro 5x user also shared numbers showing $674 → ~$157/week, which is around a 77% drop.
a lot of people have been noticing their Codex quota disappearing much faster lately too.
hopefully this is just a bug, right @thsottiaux ? right??😭
I know I'm the first person to read axes' labels here
but why does 'calculations (...) per $1000' go up as you add humans? do you get to feed them less as you develop policing, what is the angle here?
The argument that “AI structurally centralizes power” because it’s currently compute hungry ignores 125 years of super exponential growth in compute price-performance.
Improvement in algorithms and continued hardware efficiency gains means there is no reason to assume AGI-level capabilities will always require a data center to run.
Scaling laws are not laws of physics. They’re simply empirical relationships observed for particular architectures, objectives, datasets etc. Change any one of those factors and you get a different scaling curve.
If the brain is any guide, true AGI will likely be very efficient. In fact, current scaling laws might a bug and not a feature. It shows how inefficiency of the ML algorithms we discovered so far.