Peter Berger
@peterb.bsky.social
effort + coffee = software. See also @[email protected]
Reposted by Peter Berger
one of the dumbest kinds of person you encounter on the internet is "I want a planned economy but I will murder you if you ever show me a graphs"
November 30, 2025 at 11:41 PM
one of the dumbest kinds of person you encounter on the internet is "I want a planned economy but I will murder you if you ever show me a graphs"
"Paid advertising results at the top of the page" at least makes sense! "Let's generate an LLM answer, usually wrong" is incomprehensible.
November 30, 2025 at 5:57 PM
"Paid advertising results at the top of the page" at least makes sense! "Let's generate an LLM answer, usually wrong" is incomprehensible.
The life hack (for the situations where the page exists) is to always use simple.wikipedia.org instead of en.wikipedia.org for your initial dive into a math topic.
Wikipedia
simple.wikipedia.org
November 30, 2025 at 3:27 PM
The life hack (for the situations where the page exists) is to always use simple.wikipedia.org instead of en.wikipedia.org for your initial dive into a math topic.
It feels to me like that's what Google has managed to accomplish, which is the exact opposite of their stated mission of "organize the world's information and make it universally accessible and useful.
What an incredible own-goal by Google. (7/7)
What an incredible own-goal by Google. (7/7)
November 30, 2025 at 3:10 PM
It feels to me like that's what Google has managed to accomplish, which is the exact opposite of their stated mission of "organize the world's information and make it universally accessible and useful.
What an incredible own-goal by Google. (7/7)
What an incredible own-goal by Google. (7/7)
The net result of this is the information environment becomes unstable. I remember someone describing the intent of Russian propaganda being not to make you believe false things, but to make it impossible for you to have confidence in any facts at all. (6/7)
November 30, 2025 at 3:10 PM
The net result of this is the information environment becomes unstable. I remember someone describing the intent of Russian propaganda being not to make you believe false things, but to make it impossible for you to have confidence in any facts at all. (6/7)
But if I'm using a search engine, the search engine doesn't know the context! So Google's decision to always give me a maybe-wrong answer first is infuriating, and I absolutely understand people's reaction to that to be "I'm gonna flip this table. LLMs all suck." (5/7)
November 30, 2025 at 3:10 PM
But if I'm using a search engine, the search engine doesn't know the context! So Google's decision to always give me a maybe-wrong answer first is infuriating, and I absolutely understand people's reaction to that to be "I'm gonna flip this table. LLMs all suck." (5/7)
(My fave real-world example of this is that every Wikipedia math page strives for complete accuracy, and thus ends up 100% unreadable by any actual Earthling. GIVE ME AN INACCURATE SUMMARY EXPLANATION FIRST before throwing 67,231 formulas at my head.) (4/7)
November 30, 2025 at 3:10 PM
(My fave real-world example of this is that every Wikipedia math page strives for complete accuracy, and thus ends up 100% unreadable by any actual Earthling. GIVE ME AN INACCURATE SUMMARY EXPLANATION FIRST before throwing 67,231 formulas at my head.) (4/7)
Or, sometimes, you want an answer that's approximately correct, but is simplified and without nuance. (3/7)
November 30, 2025 at 3:10 PM
Or, sometimes, you want an answer that's approximately correct, but is simplified and without nuance. (3/7)
LLM answers, in the right context, can be super-valuable. "But the machine will make stuff up!" I hear doomers replying. My rejoinder is: there are PLENTY of contexts where what you want is not a correct answer, but an answer that SOUNDS RIGHT, and the accuracy doesn't matter. (2/7)
November 30, 2025 at 3:10 PM
LLM answers, in the right context, can be super-valuable. "But the machine will make stuff up!" I hear doomers replying. My rejoinder is: there are PLENTY of contexts where what you want is not a correct answer, but an answer that SOUNDS RIGHT, and the accuracy doesn't matter. (2/7)
Indeed. I should be clear I'm not saying it WILL end up like the Concorde - there are real reasons to believe deploying compute is less fraught than deploying aviation equipment. But I'm tired of the "it's inevitable" arguments when we don't actually know what the costs will be.
November 29, 2025 at 6:56 PM
Indeed. I should be clear I'm not saying it WILL end up like the Concorde - there are real reasons to believe deploying compute is less fraught than deploying aviation equipment. But I'm tired of the "it's inevitable" arguments when we don't actually know what the costs will be.
Same same (complimentary).
November 29, 2025 at 4:54 PM
Same same (complimentary).
But the Concorde should serve as a reminder that in our lifetimes we've seen advances in technology that simply weren't financially viable to sell to anyone other than nation-states. (3/3)
November 29, 2025 at 2:32 PM
But the Concorde should serve as a reminder that in our lifetimes we've seen advances in technology that simply weren't financially viable to sell to anyone other than nation-states. (3/3)
I don't think we can tell yet whether AI at scale is something that can be profitable to operate in the consumer sphere; right now it's being subsidized by investors, and I don't think anyone can accurately predict the operating costs of the technology 20 years from now. (2/3)
November 29, 2025 at 2:32 PM
I don't think we can tell yet whether AI at scale is something that can be profitable to operate in the consumer sphere; right now it's being subsidized by investors, and I don't think anyone can accurately predict the operating costs of the technology 20 years from now. (2/3)
おめでとうございます！ Looking forward to playing it on Steam (will there by any chance be a Mac version?
November 28, 2025 at 2:30 PM
おめでとうございます！ Looking forward to playing it on Steam (will there by any chance be a Mac version?
Happy Swordsgiving to those who celebrate.
November 28, 2025 at 1:24 PM
Happy Swordsgiving to those who celebrate.
I think we can all agree the most important aspect of this is we once again beat out Cleveland.
November 27, 2025 at 4:20 PM
I think we can all agree the most important aspect of this is we once again beat out Cleveland.
Bonus tweet: 18 seconds of me being (a little) irritated at the way nice language features are sometimes denigrated as mere "syntactic sugar".
youtube.com/shorts/wpBMc...
youtube.com/shorts/wpBMc...
Syntactical Sugar: "We need sugar to live" #abstraction #haskell #coding #functionalprogramming
YouTube video by Tea Leaves
youtube.com
November 26, 2025 at 1:41 AM
Bonus tweet: 18 seconds of me being (a little) irritated at the way nice language features are sometimes denigrated as mere "syntactic sugar".
youtube.com/shorts/wpBMc...
youtube.com/shorts/wpBMc...
Anyway, that's why when I make silly errors in my Haskell (or whatever) coding streams, I leave them in the video. You make mistakes. So do I. So do all of us.
Welcome to software development. (5/5)
(And, yes, I'm thinking of the same jerk you are all thinking of.)
Welcome to software development. (5/5)
(And, yes, I'm thinking of the same jerk you are all thinking of.)
November 25, 2025 at 4:40 PM
Anyway, that's why when I make silly errors in my Haskell (or whatever) coding streams, I leave them in the video. You make mistakes. So do I. So do all of us.
Welcome to software development. (5/5)
(And, yes, I'm thinking of the same jerk you are all thinking of.)
Welcome to software development. (5/5)
(And, yes, I'm thinking of the same jerk you are all thinking of.)
"I'll just choose to not make mistakes" isn't just false, it's a toxic lie people spread to intentionally boost their brand. It's harmful to every other developer out there. And if you say this then you're a bad person and should feel ashamed of yourself. (4/5)
November 25, 2025 at 4:40 PM
"I'll just choose to not make mistakes" isn't just false, it's a toxic lie people spread to intentionally boost their brand. It's harmful to every other developer out there. And if you say this then you're a bad person and should feel ashamed of yourself. (4/5)