HN user

gkbrk

2,947 karma
Posts34
Comments673
View on HN
david-gilbertson.medium.com 5mo ago

Shrinking a language detection model to under 10 KB

gkbrk
1pts0
centralflows.github.io 9mo ago

Understanding Optimization in Deep Learning with Central Flows

gkbrk
2pts0
unsloth.ai 12mo ago

Unsloth – Dynamic 4-bit Quantization

gkbrk
2pts0
www.gibney.org 3y ago

A Syntax for Self-Tracking (2020)

gkbrk
81pts32
hg.sr.ht 4y ago

Garnet programming language – What if Rust was small?

gkbrk
2pts0
github.com 4y ago

Kyoto project is moving from GitHub to Sourcehut

gkbrk
211pts275
www.jamesstevenson.me 4y ago

Reverse Engineering the Google Play SafetyNet Attestation API

gkbrk
3pts0
www.gibney.de 5y ago

A Syntax for Self-Tracking (2020)

gkbrk
1pts0
metin.nextc.org 5y ago

Linux Kernel Overlapping Internet Fragment Bug

gkbrk
1pts0
www.gkbrk.com 6y ago

Plaintext Budgeting

gkbrk
2pts0
www.youtube.com 6y ago

Primordial Particle Systems

gkbrk
2pts0
adi.tilde.institute 6y ago

Make website – shell script to generate static websites

gkbrk
2pts0
twitter.com 6y ago

Twitter bots steal artwork for t-shirt companies

gkbrk
4pts1
adventures.michaelfbryan.com 6y ago

Audio Processing for Dummies

gkbrk
395pts47
drewdevault.com 6y ago

An old-school shell hack on a line printer

gkbrk
4pts0
codegolf.stackexchange.com 6y ago

Images with All Colors

gkbrk
4pts0
adventures.michaelfbryan.com 6y ago

Audio Processing for Dummies

gkbrk
2pts0
gkbrk.com 7y ago

Mastodon Bot in Common Lisp

gkbrk
5pts0
gkbrk.com 8y ago

Fetching ActivityPub Feeds

gkbrk
5pts0
mlvnt.com 8y ago

Raspberry Pi Headless Setup

gkbrk
3pts0
visual6502.org 8y ago

6502 Simulator with Transistor-Level Accuracy

gkbrk
4pts0
gkbrk.com 8y ago

Writing a Simple D-Bus Service in Python

gkbrk
1pts0
blog.zhpass.com 8y ago

MAY – Rust Stackful Coroutine Library

gkbrk
3pts0
jasonpunyon.com 8y ago

Fun with Random Numbers: Random Projection

gkbrk
3pts0
gkbrk.com 8y ago

Putting My Blog on IPFS

gkbrk
2pts0
codegolf.stackexchange.com 9y ago

Images with all colors

gkbrk
1pts0
gkbrk.com 9y ago

Unprotected Redis Instances in the Wild

gkbrk
1pts0
gkbrk.com 9y ago

Android Dialers Are Stealing Your Data

gkbrk
3pts0
gkbrk.com 9y ago

Generating graphs from my todo.txt

gkbrk
2pts0
gkbrk.com 9y ago

Numerical Domains of China

gkbrk
5pts2

The same server that's responsible for sanitizing garbage JS out of user content is also responsible for sending the Content-Security headers. Why would you trust it with one, but not the other? If it's buggy garbage it will also send the wrong headers.

CursorBench 3.1 21 days ago

It looks very natural, cheaper is better after all. Performance axis going up, and cheapness axis going up match each other.

Firefox has YouTube integration of course.

Here's a Firefox file [1] specifically for integrating YouTube videos into their picture-in-picture system. Your random video website won't get this treatment of course, need to be a popular one.

Here's a piece of C++ code [2] in the Firefox engine that specifically rewrites old YouTube embeds from their old HTML embed snippet to the new one. Again, your own video website will never be so deeply integrated into Firefox because it's not a top 10 website.

Firefox Readability mode makes pages more readable by removing useless stuff like videos. Unless it's a YouTube [3] or other top-N popular video website of course. YouTube videos are given special treatment because it's popular and having small integrations like this make the user experience better.

[1] https://github.com/mozilla-firefox/firefox/blob/1f43fe5ffadd...

[2] https://github.com/mozilla-firefox/firefox/blob/1f43fe5ffadd...

[3] https://github.com/mozilla-firefox/firefox/blob/1f43fe5ffadd...

The way users use Facebook and LLMs are so massively different, it almost seems like a bad faith argument to equate them.

Facebook is mostly scrolling the timeline and passive consumption. It doesn't benefit from being on the side because the content you interact with on Facebook is completely separate from the content on your other tabs.

In contrast, LLMs have ongoing conversations that the user can come back to, and each conversation might relate to multiple tabs that the user is working on. On top of that, it's a very common occurrence that the user has questions about, or a task to be done using the content of the current page. This makes LLM and chatbot integration much more useful than a Facebook integration.

Also if you have the Facebook Messenger installed, Firefox already gives you an integration to share things with your Facebook contacts.

Because people like chat bot sidebars.

My code editor has a built-in chat bot sidebar that I use every day. It's not a huge stretch that people who use chatbot sidebars in other applications would also want one in their browser.

ChatGPT is the #6 most popular website in the world, why wouldn't a browser want tighter integration with such a popular kind of service?

They don't pick the keywords uniformly randomly from a list of all keywords though. They think they randomly picked something that popped up in their mind, but those keywords are either

- stuff they saw online recently — ads or otherwise, which put the keywords in their mind

- or stuff they were already interested in recently

Not hard to imagine targeting algorithms picking up on either of these

That's not because 4o is good at things, that's because it's pretty much the most sycophantic model and people easily fall for a model incorrectly agreeing with them then a model correctly calling them out.

Not just companies, people too. It will have trouble getting into Linux distro repos. And a lot of devs/users avoid non-open-source projects especially if they went looking for solutions on Github.

Pandas 3.0 6 months ago

we know how those go

They get forked and stay open source? At least this is what happens to all the popular ones. You can't really un-open-source a project if users want to keep it open-source.