Ada 🏳️‍⚧️

you might not like it, but this is what the ideal symbol looks like

My job is quite literally to put the words in the correct order so the thinking sand does what my client wants.

Free*

* $2500

I also don't think it should matter if the code was written by a human or generated by an LLM, Ouija board, dice roll, whatever.

If the security or correctness matters to you, review it yourself! The source is open!

I am still waiting to see examples of these "agent bugs" or whatever people call them. I've seen multiple people make the claim that LLMs can somehow generate code that looks correct to any human reviewer, but contains massive security vulnerabilities.

No one has ever provided proor for this claim.

"[..] loves LLMs more than BIPOC women [..]" is forever going to exemplify Mastodon for me.

Hi Alice!

All the things I dislike are connected by the fact that people who do/like/support them are Bad People.

Here's a tip for writing good alt text: imagine you are the CEO of McDonalds eating a burger.

I switched back to linux few months ago, after daily driving Windows for years. I do not regret doing it, linux has come a long way since I last used it. Not perfect yet, but that's mostly my own fault for using Arch and Cosmic.

maybe I am the imposter 😔

Path traced picture of the Sponza GLTF scene. Instead of a typical render it should the calculated irradiance for the surfaces, so the materials look diffuse white with some colored bounce light on them. The image is slightly noisy, but that shouldn't matter for NIV training. 1024x1024 tile, 1 073 741 824 samples total. 12 bounces per sample. 28 seconds on my RTX 3060 Ti.

I'm cooking? Need FLIP so I can compare this against Cycles and/or Mitsuba3. bevy_mod_niv next weekend 😊

management wants to see the Big Picture 😔

And more importantly, at least to me, it can give me the correct terms so I can look up things myself!

replacing it with AI

"It felt better before I sampled it", a phrase I used to describe my failure to put a thought into words

I think therefore I am not wrong

I am no expert, but it is always funny how confident people who don't work in the field are. Meanwhile every researcher I have spoken to just says that they don't really know what's going on inside the "AI". In fact they also tell me we don't really know what is going on inside our brains either!

I guess I am not an expert then 😔 However, Sonnet 4.5 did invent a "Symbolic Constant", a compile-time constant whose value can only be resolved during runtime.

Scaling law implies the existence of the scaling police

Same vibes as the Suno CEO saying he invented "multiplayer music". I just want to know how these people grew up to have their dreams so twisted.

Sponsor @orhun on GitHub Sponsors Hey! I'm Orhun, a Rust enthusiast with a strong passion for all things open source. I cook Ratatui. If you like my work consider sponsoring me! 🐁

if your company uses @ratatui.rs
here's how you can keep the rats supplied with cheese 🧀
github.com/sponsors/orh...

GLaDOS is a LLM

No, kate! No further action!!

replaced DNS with AI

"The HTTP 402 Payment Required client error response status code is a nonstandard response status code reserved for future use."

Not sure how this is more correct than a 429

yes but someone has to create a little chaos 😌

fd 35; I'm just eyeballing it

fd 10

Like most problems with AI this too is not a problem with AI it's a problem with the mad gold rush around AI.

I have my issues with LLM and generative AI, but the really big problems are very much human behaviour, exploitation and public funding and regulatory capture by big capital.