Hacker News: noosphr

New comment by noosphr in "Coalton is an efficient, statically typed Lisp with ideas from Haskell and OCaml"

noosphr — Fri, 29 May 2026 10:28:40 +0000

Ah you should have just said you use arch.

But it's a very weak programmer that needs a test suite. Real ones prove the consistency of their programs with goose feather and vellum.

New comment by noosphr in "Coalton is an efficient, statically typed Lisp with ideas from Haskell and OCaml"

noosphr — Fri, 29 May 2026 06:47:50 +0000

If you get deeply excited about config files, sure, go right ahead and set up the dev env without help.

New comment by noosphr in "Coalton is an efficient, statically typed Lisp with ideas from Haskell and OCaml"

noosphr — Fri, 29 May 2026 00:35:52 +0000

Ask an llm how to setup.

This is one area where they excel at with no caveats.

New comment by noosphr in "I think Anthropic and OpenAI have found product-market fit"

noosphr — Wed, 27 May 2026 21:48:28 +0000

>Without all that might be $175/user/month but you're not shipping apps with just vi and bare gcc.

You're right, Linus uses Emacs.

New comment by noosphr in "Erin Brockovich made a map to track data centers around the country"

noosphr — Wed, 27 May 2026 15:01:54 +0000

Violence is defined by the World Health Organization in the WRVH as “the intentional use of physical force or power, threatened or actual, against oneself, another person, or against a group or community, that either results in or has a high likelihood of resulting in injury, death, psychological harm, maldevelopment or deprivation”.

Which definition are you using?

New comment by noosphr in "Erin Brockovich made a map to track data centers around the country"

noosphr — Wed, 27 May 2026 02:18:56 +0000

And yet we don't let people dump lead in the drinking water any more.

New comment by noosphr in "Erin Brockovich made a map to track data centers around the country"

noosphr — Wed, 27 May 2026 02:17:52 +0000

Killing you slowly is very much violence.

New comment by noosphr in "Memory has grown to nearly two-thirds of AI chip component costs"

noosphr — Mon, 25 May 2026 03:47:02 +0000

You should work in the Central African Republic coltan mines if you think anyone has a leg to stand on.

New comment by noosphr in "Memory has grown to nearly two-thirds of AI chip component costs"

noosphr — Mon, 25 May 2026 03:25:50 +0000

Help us Xi Jinping, you're our only hope.

New comment by noosphr in "Constraint Decay: The Fragility of LLM Agents in Back End Code Generation"

noosphr — Sun, 24 May 2026 22:41:50 +0000

They may be better than humans at reasoning but they are substantially worse than the first generation logic programs from the 1950s.

New comment by noosphr in "Constraint Decay: The Fragility of LLM Agents in Back End Code Generation"

noosphr — Sun, 24 May 2026 15:51:07 +0000

To this day frontier models think that A and not B means A and B when the sentence gets pushed far enough back in their context window. The context length that model can reason over without obvious errors is much smaller than the advertised context. Between a 1/4th to a 1/20th what is advertised on the tin.

New comment by noosphr in "My I3-Emacs Integration"

noosphr — Sun, 24 May 2026 01:15:46 +0000

There can never be too many modifier keys:

https://xcancel.com/octonion/status/1341113219142828039

New comment by noosphr in "Making Deep Learning Go Brrrr from First Principles (2022)"

noosphr — Sat, 23 May 2026 22:13:18 +0000

Maybe. Just means that the conventional wisdom was wrong and substantially over training can be a good thing. No one I knew at the time suspected that, including the people who wrote the paper.

New comment by noosphr in "Making deep learning go brrrr from first principles (2022)"

noosphr — Sat, 23 May 2026 12:38:23 +0000

>We show that a variety of modern deep learning tasks exhibit a "double-descent" phenomenon where, as we increase model size, performance first gets worse and then gets better.

New comment by noosphr in "Making Deep Learning Go Brrrr from First Principles (2022)"

noosphr — Sat, 23 May 2026 12:24:23 +0000

>For example, getting good performance on a dataset with deep learning also involves a lot of guesswork. But, if your training loss is way lower than your test loss, you're in the "overfitting" regime, and you're wasting your time if you try to increase the capacity of your model.

https://arxiv.org/abs/1912.02292

New comment by noosphr in "sp.h: Fixing C by giving it a high quality, ultra portable standard library"

noosphr — Sat, 23 May 2026 05:47:10 +0000

We support extreme portability for sufficiently large values of two.

New comment by noosphr in "AI has a multiplying effect on existing technical skills"

noosphr — Fri, 22 May 2026 19:56:43 +0000

The if in there is doing a lot of heavy lifting.

New comment by noosphr in "AI has a multiplying effect on existing technical skills"

noosphr — Fri, 22 May 2026 19:56:04 +0000

This isn't an indictment of how good AI is but how poor our tools are. We had gui makers in 1996 that made slop which allowed you to iterate in real time. They didn't need a datacenter worth of compute and a nuclear reactors worth of power to run.

New comment by noosphr in "Deepfakes Tore a High School Apart"

noosphr — Fri, 22 May 2026 19:48:23 +0000

That is a problem with US society. One that deep fakes can solve by allowing plaslsible deniability for any image or video.

New comment by noosphr in "Deepfakes Tore a High School Apart"

noosphr — Fri, 22 May 2026 13:07:28 +0000

And to a good Stalinist it was the dignity of the USSR.

Something millions died to protect in the great patriotic war.

Again, symbolic violence is worse than real violence to a lot of people.