wonnage

Sounds like Anthropic should close up shop then, those chumps are offering their product on the internet for any random loser to distill

this is the sort of brain rot thought that you have in a dorm room the day you are introduced to Econ

“bro like, what if we could price the sum total of human knowledge? That wouldn’t be that much, right?”

quite the leap from “authors rely on the work of others” to vacuuming up the sum total of digitized knowledge to tune some matrices

The hordes of acolytes mumbling “Jevon’s paradox, Jevon’s paradox” every time AI gets cheaper beg to differ. Every minute saved using AI is just an opportunity to use more AI

real nice leaps from “groups of agents are corporations” to “agents with a /goal are analogous to maximizing shareholder value”

Maybe you should try reading the readme it wrote you.

They’re useable if you’re trying to rank autosuggestions or something, not great at implying actual understanding

nothing completes a HN AI thread more than echelon jumping in from the top rail with a wild take

The Achilles heel of this class of library (including stylex, Linaria, etc.) is precedence. You either have to use a runtime and some complicated merging rules (e.g longhand overrides shorthand) or compile every possible combination of the styles at build time.

So if a program decided it owned .cache and deleted it every startup you’d be fine with it?

Don’t punish the people who do crimes because otherwise they won’t self-report that they are doing crimes? real galaxy brain shit

Ok agentic Ron Jeffries

https://ravimohan.blogspot.com/2007/04/learning-from-sudoku-...

The thin serifs not being slightly shifted to align weight-wise with the sans serif is triggering my OCD, but yeah Astra is miles ahead here

Yeah that’s why ea is like catnip for these folks, justify anything by some handy hallucination about how the outcome will help more than it hurts. Suffering is fungible

> Overall, the best-performing model was Claude Opus 5 on “reasoning” mode, which still made mistakes in 39 per cent of answers.

They indeed will be in about 20 years when the impending population collapse hits

Serving requests where every user has their own set of self-updating weights will absolutely murder whatever minimal margin the AI providers have today

Dan spent 7k words missing the forest for the trees

People had been calling a housing bubble since 2003. You quite literally need the market to do dumb things over a period of time before the bubble pops. Otherwise it’s not a bubble!

Actually making money on a bubble pop is another story and that does rely on timing. pointing at a history of failed market predictions as a gotcha is about as effective now as it would’ve been in 2003-2007 right up until the bubble popped

Right now it’s painfully obvious that there is a bubble, the circular financing is documented, the hype-driven valuations are real, it’s just a waiting game to see who holds the bag

Surely if you hoover up every book in existence to feed into an ai model you must be extracting more than 1.5B in value. If not then it’s not a viable business.

The whole point of transformers is that you can take “a tiny ML model” and just scale it up 100000x and then it tells Zuckerberg what to bake with his kid

You can pour a bunch of beads into a pachinko board and end up with a normal distribution. Do the system “know” they need to go there?

Then you feed a bunch of tokens into a GPU and end up with a distribution of possible next tokens…

I think you are just litigating the philosophical debate of Bayesian vs frequentist interpretations of probability. Because the weights really are just probabilities.

Bayesians say that the probabilities represent strength of belief, implying some subjective knowledge or information. It is necessarily subjective in that it requires priors, i.e information the predictor knew before making the prediction. In other words, the LLM has priors from training and is predicting tokens using real knowledge

Frequentists would say that probabilities are simply objective facts - e.g we all agree that the physical property of temperature follows from any molecules matching a particular energy distribution. You’re not predicting anything, there’s just some outcomes that are happening at the expected rate. In other words, the LLM is a stochastic parrot/next token predictor

This article presents some evidence for why Grep might work better but I don’t think it does a great job of explaining why it gets chosen - is it something that was intentionally reinforced during training or was it just because LSP is harder to train on because it’s usually hidden behind some IDE interface

There’s no reason why you couldn’t write a search tool that e.g combines LSP and grep. Or ast-grep, for that matter. It feels like one of those things we haven’t spent much time investigating because grep is good enough

Don’t forget making podcasts and telling you what to bake with your kids

They literally fearmonger for a permanent underclass, it’s definitely not decreasing

It’s all just vibes

Not sure if Blink is more locked down but Ring works alright with just Homebridge + HomePods

Those are rookie numbers, once Elon sends my agents to Mars they’ll be doing 40 extra minutes per sol

AI is somewhere down there with Covid and Ubisoft, I’m gonna go with yes

Do not recommend. There’s this stupid hard limit every day which takes like 8 hours to reset no matter how many tokens you used