Vlad Erium 🇯🇵

新緑の伊勢神宮、四季と伝統が織りなす日本の真髄

When there's smoke, there's a chainsmoking cat. Delve into her uncensored life in Chainsmoker Cat, premiering July 2 on Netflix.

My Underwater Caustics Pro shader for Blender is now FREE for a limited time! 🌊🥰 This scene is also included in the files! It's in celebration of our store Grand Re-Opening 🥳 Go check it out at... smouse.studio ❤️

@antirez Those memory safety cost a lot 😞

The Fil-C story is a very interesting one, I talked about it in my YouTube channel back then, but what is very intriguing is that if Rust was *really* about safety, the same community would adopt Fil-C massively. Together with Rust, likely, not in substitution, but massively.

@parhelion318 Это что за хня?

Классный продукт у Sakana AI вышел

Vlave not great in PR for sure. They must set the Steam Machine price at 1/2~1/3 of its price. Sell a couple of pieces to real people. And immediately go out of stock due to high demand. Sell the rest on eBay as a scalper for a real price🤔

We've kept hearing how GLM-5.2 beats Opus 4.8, and are skeptical of benchmarks - so we tested them on a real bug from the Cline repo. While both models fixed the issue, GLM was the winner in terms of cost and code quality: - GLM used twice as many tokens (GLM 1.1m vs Opus 660K) but cost half as much (GLM $0.41 vs Opus $0.81) - Opus finished quicker - 1.6 min and 12 tool calls vs GLM 4.7 min and 28 tool calls - GLM cleaned up dead code and verified the build compiled before completing. Opus didn't - it left type errors that passed tests but broke the production build. Both runs used the same Cline harness prompting and tools, so it seems GLM is RL trained to spend more tokens verifying its work before completing. Impressive work by the @Zai_org team!

@SadlyItsBradley Drugs can be cheaper than Ai hallucinations🤔

Holy. Shit. People don’t realize how quickly we will be able to save full 3D models of things we see online, even without the host generating/modeling an object Just like right clicking or screenshotting a JPG and setting it as a desktop wallpaper. The future is spatial splats

Light pillars over rippled water: why their sides appear parallel Philip W. Kuchel arxiv.org/abs/2606.18547 Abstract: Streetlights shining across rippled water often produce tall, narrow reflections with strikingly parallel sides. These light pillars appear almost architectural, yet the water surface is neither vertical nor smooth. We develop a geometric optics model that explains the phenomenon using the specular reflection rule, projection geometry, and the physics and statistics of surface slopes. A rippled water surface can be viewed as an ensemble of small facets (tangent planes on waves) acting as tiny mirrors. As one looks farther across the water (coordinate x), the physical width of the region whose facets reflect the light into the eye (or camera/pinhole) increases, but the pinhole projection onto the image plane compresses this widening by a factor proportional to 1/x. The two effects cancel, producing a reflection whose image width remains constant. This paper appears to be the first to formally document this. (Deviations from the strict cancellation at both ends of the image are also explained.) The analysis clarifies how earlier qualitative treatments failed through a lack of identifying the role of projection geometry. The model also qualitatively explains the brightness variation along the pillar, and why extended objects such as buildings do not produce elongated reflections. The treatment is intended for scholars of introductory optics and for those interested in the physics and mathematics of everyday visual phenomena. We conclude by outlining how the same framework extends naturally to more complex viewing situations, such as the setting sun viewed from a cliff top over the sea.

@nixcraft That is not a Linux become user friendly, but LLM made such steps simple. 🤷‍♂️

Every one is a #Linux expert until they try to install #Nvidia drivers on a fresh Debian install 👩🏻‍💻

@thsottiaux @TechByTaraa Be careful, you already speaking with ai bots 😅

@TechByTaraa You're a Claude user

I'm a Claude user. Give me one reason to switch to Codex

Codex rate-limits have jumped 10x since last week. this is reported by many users on the Codex repo and backed by stats.

I integrated this into Copilot and ran some evals this afternoon. Results were neutral at best and in most cases resulted in MORE token consumption. These results are consistent with our prior evaluation of RTK. Compression ends up removing information that the agent needs, which triggers a re-read, ultimately resulting in larger costs and latency. Beware of things that sounds too good to be true.

@YeheLiu @polarization_yu My common workflow start in Pro to discuss a topic, write ideas to MD and pass that to a Codex-CLI

Not sure if anyone tried this. ChatGPT pro is not available in codex, and Xhigh is a bit behind in research. You can use pro to review the repo in GitHub (with optional instructions saved in md), create plan, and feed the shareable URL back to codex to execute.

Evening winter courtyard

@84Aleha Питерские колодцы. Круто :)

@ssh4net Да

More fun with profiling in Superluminal. Couple of days ago I had 32 badly utilized threads. Profiler showed that most of it was due to malloc&free (including hidden ones from std::vector and co). Now I have 10 threads doing more work than the 32 did before.

Response from my radiologist father who has read MRIs for decades: “This will fail miserably, will harm people and will lose them millions upon millions of dollars. If there were a way that I could short a private investment, I would do so here”

There are a lot of question of that even possible in reality. Couple centimeters of water are possible and already sometime used. 10cm already almost not. But fullbody with many decimeters from emitters to a body, will probably generate a data similar to a lensless sensors, that required deconvolution, and probably neural networks to be usable. Hallucinations in a infrasound scans is what medicine don’t need 😅

À gauche, les images révolutionnaires qu'ils annoncent. À droite un scanner banal.