Pelicans on a bicycle(simonwillison.net)
72 points by colejohnson66 | 51 comments
First screen, captured .
Looks similar
You might not need WebSockets(hntrl.io)
390 points by hntrl | 247 comments
Blog post by Hunter Lovell arguing that WebSockets are often unnecessary and simpler HTTP-based alternatives can work for real-time web apps.
OpenSSH Ships on Every Mac, Linux Server and Windows. Its Creator Trusts No One(zbruceli.org)
4 points by zbruceli | 0 comments
Long-form blog profile of OpenSSH creator Theo de Raadt, covering his NetBSD exit, lost DARPA grant, and security-focused philosophy.
Related stories
Are AI labs pelicanmaxxing?(dylancastillo.co)
632 points by dcastm | 237 comments
Blog post investigating whether AI labs are optimizing their LLMs for Simon Willison's 'SVG of a pelican riding a bicycle' informal benchmark.
Why do LLMs freak out over the seahorse emoji?(vgel.me)
618 points by nyxt | 312 comments
Blog post by vgel examining why large language models produce confused, looping output when asked whether a seahorse emoji exists.
LLMs use a surprisingly simple mechanism to retrieve some stored knowledge(news.mit.edu)
336 points by CharlesW | 121 comments
MIT News article on research showing LLMs often retrieve stored facts via simple linear functions, and a technique to probe what a model knows about subjects.
DuckDB Doesn't Need Data to Be a Database(nikolasgoebel.com)
353 points by tosh | 115 comments
A blog post by Nikolas Göbel discussing how DuckDB can act as a database without storing data itself, querying external data sources directly.
Do your own writing(alexhwoods.com)
690 points by karimf | 222 comments
Blog post by Alex Woods arguing against using LLMs to write documents and essays, since writing is a way to think and build trust.
Use Prolog to improve LLM's reasoning(shchegrikovich.substack.com)
343 points by shchegrikovich | 122 comments
Substack post arguing that having LLMs generate Prolog code can improve their reasoning, given the limits of autoregressive transformer generation.