activescott's Notes

More than half (55%) said they have a favorable view of the new mayor, while 75% said they believe Mamdani is working hard.

Also, at least 60% of NYC residents see Mamdani as:

A good leader Fulfilling campaign promises Working to represent all New Yorkers Understanding the city's problems Doing more to unite the city than divide it Despite all that good news for the mayor, less than half of New Yorkers approve of his job performance thus far. But if the majority holds all the aforementioned beliefs, why is his overall approval rating floundering at less than half?

With the state budget unresolved, and Mamdani so far unable to deliver his tax on the rich, many of his bigger campaign promises like free buses and affordable housing remain up in the air too.

Mamdani's approval rating is lower than that of Eric Adams (61%) and Bill de Blasio (49%) at the same time during their administrations.

Similar to IDF's current Gaza model, the defense establishment intends to destroy villages near the Israel-Lebanon border, and establish permanent military outposts in the area between the border and the Litani River

The defense establishment's perception is that all of these villages are used by Hezbollah for activities against Israel, and therefore they must be completely destroyed to prevent Hezbollah operatives from returning to the area.

Defense Minister Israel Katz said on Tuesday that all houses in Lebanese villages near the Israeli border will be demolished "like in Rafah and Beit Hanoun," referring to areas in the Gaza Strip where the IDF carried out widespread demolitions of homes during the war.

more than 600,000 Lebanese residents who have been evacuated will not be allowed to return to south Lebanon "until the security and safety of northern residents are guaranteed." The defense minister said that following the IDF's operation in Lebanon, Israeli troops will continue to be stationed in a "security zone" inside Lebanese territory to defend against anti-tank missiles and to maintain security control of the area south of the Litani River.

SANTA CRUZ CARBON CHAMELEON (C): FULL REVIEW

Italian ‌Prime Minister Giorgia ‌Meloni said on Tuesday that her government has halted the automatic renewal of a military Memorandum of Understanding (MoU) with Israel, as Italy has become increasingly critical of Israel’s wars and killing of civilians.

According to Haaretz, the MoU includes the exchange of military equipment and cooperation on military research, though an Italian source told the Israeli newspaper that Meloni’s announcement reflected a policy already in place.

The source said that Italy halted military cooperation with Israel shortly after October 7, 2023, though according to reports from last year, Italy suspended weapons exports to Israel but was still fulfilling previous arms deals. Meloni has been under significant pressure from her opposition and from Italian citizens to cut ties with Israel. Italian Foreign Minister Antonio Tajani visited Lebanon on Monday, where he slammed Israel’s “unacceptable attacks by Israel against the civilian population” and said it called for the avoidance of “another escalation like the one in Gaza.” When asked if Italy would take a similar step toward the US, the source speaking to Haaretz said that Italy was still assessing the impact on Iranian civilians in the US bombing campaign in Iran, which started with the bombing of an elementary school, which slaughtered more than 100 children.

From taxes to the environment to public broadcasting like PBS and NPR, the Senate has recently passed record levels of legislation and confirmed record numbers of nominations with senators representing less than half the people.

Using historical data, GovTrack found 56 examples of Senate votes on legislation that passed with senators representing a “population minority.” 26 of those 56 examples, nearly half, have occurred since President Donald Trump’s current term began.

Several of the second Trump administration’s most prominent members were confirmed by Senate “population minority” votes – including RFK Jr. (Secretary of Health and Human Services), Pete Hegseth, Pam Bondi, Kash Patel, and Tulsi Gabbard.

The oldest example GovTrack found of a “population minority” Senate vote is actually  famous: Clarence Thomas’s 1991 Supreme Court nomination by President George H. W. Bush. The Senate approved Thomas with 52% support, but 49% of the population.

He still serves on the Court today.

GovTrack found three other “population minority” Senate confirmations for Supreme Court justices, totalling four: Thomas plus Trump’s three first-term nominees. All four still serve on the Court.

Currently, Republicans hold a Senate majority: 53 to 47. However, based on the Census Bureau’s current estimates, it’s actually the other way around by population: Democratic senators represent a 53% majority of the states’ population, versus Republicans with 47%.

How is this possible? Because while the U.S. House is apportioned based on population, with larger states receiving more representatives, the U.S. Senate guarantees each state two senators regardless of size.

This was baked into the American system from the beginning, creating what political scientists call a “counter-majoritarian” institution.

In 2025, according to Census Bureau estimates, the most populous state (California) had about 67x the population as the least populous: Wyoming. Today, a Senate voting majority could be cobbled together from senators representing just 17% of the population.

But that’s actually been the same for a while. Going back to 1900, a Senate voting majority could be cobbled together with senators representing 16% to 20% of the population.

Instead, small states may be more politically aligned than they used to be and are voting together more often as a bloc.

Senators have recently taken advantage of old rules, and also changed some rules, to use lower vote thresholds. This means votes are more often succeeding with less support.

Both parties contributed to this.

In 2013, under President Obama, Senate Democrats changed the threshold for most nominations from three-fifths to a simple majority. They left it at three-fifths for the Supreme Court, though.

Then in 2017 during Trump’s first term, Senate Republicans changed the threshold for the Supreme Court, too, to confirm Justice Gorsuch by a simple majority. (This rule applied to all subsequent justices, too.)

As for legislation, many of the recent “population minority” Senate votes used the Congressional Review Act of 1996, which lowered the usual Senate vote threshold from three-fifths to a simple majority for certain deregulation bills. The One Big Beautiful Act and the Rescissions Act were both voted on under other rules, which lower the vote threshold for certain spending-related bills.

So the three-fifths threshold is now gone for nominations and some types of legislation.

It might not stop there. Trump has called for the Senate to end the three-fifths threshold for all legislation, in order to enact certain Republican policies – particularly regarding election rules. If that happens, “population minority” Senate votes could become even more frequent.

Why does this usually benefit Republicans?

This discrepancy usually benefits the GOP, since they tend to represent smaller states.

This small-state Republican benefit also holds true at the presidential level. Indeed, two presidents in living memory won election despite losing the national popular vote, both Republicans: George W. Bush in 2000 and Donald Trump in 2016.

The Congressional Review Act, which makes it easier for Congress to deregulate – and the rules for rescissions bills, which makes it easier to cut funding – also are more aligned with Republican goals than Democratic goals.

But for better or for worse, it’s clear that the Senate is diverging from popular opinion far more than ever before, at least in recent memory. Even if one believes the Senate is, in fact, “right” while popular opinion is “wrong.”

vLLM offers an official Docker image for deployment. The image can be used to run OpenAI compatible server and is available on Docker Hub as vllm/vllm-openai.

llama.cpp is an Inference Engine that supports a wide-variety of models architectures and hardware platforms. It however does not support Batch Inference, making it less than ideal for more than one request at a time. It is mainly used with the GGUF quantization format, and the engines runs with an okay performance for single-request runs but not much else. The only time I would actually recommend using llama.cpp is when you do not have enough GPU Memory (VRAM) and need to offload some of the model weights to the CPU Memory (RAM).

Accompanied with AMD Epyc Milan 7713 CPU, I was able to get approximately 1 token per second solely through CPU offloading of DeepSeek v2.5 236B BF16 model, which might sound okay but it really is not. To illustrate why this is suboptimal, utilizing 8x GPUs of my 14x GPU AI Server , and with GPU-only offloading, my server could handle approximately 800 tokens per second while processing 50 asynchronous requests on Llama 3.1 70B BF16 through vLLM’s Batch Inference utilizing Tensor Parallelism.

Cool laptops and desktops for Linux.

You can use the OpenAI Compatible Provider package to use language model providers that implement the OpenAI API.

Below we focus on the general setup and provider instance creation. You can also write a custom provider package leveraging the OpenAI Compatible package.

We provide detailed documentation for the following OpenAI compatible providers:

LM Studio
NIM
Heroku
Clarifai

The general setup and provider instance creation is the same for all of these providers.

HLS.js is a JavaScript library that implements an HTTP Live Streaming client. It relies on HTML5 video and MediaSource Extensions for playback.

It works by transmuxing MPEG-2 Transport Stream and AAC/MP3 streams into ISO BMFF (MP4) fragments. Transmuxing is performed asynchronously using a Web Worker when available in the browser. HLS.js also supports HLS + fmp4, as announced during WWDC2016.

HLS.js works directly on top of a standard HTML element.

detailed benchmarks and info wrt apple silicon cpus with llama.

LlamaBarn is a macOS menu bar app for running local LLMs.

While Vulkan can be a good fallback, for LLM inference at least, the performance difference is not as insignificant as you believe. I just ran a test on the latest pull just to make sure this is still the case on llama.cpp HEAD, but text generation is +44% faster and prompt processing is +202% (~3X) faster with ROCm vs Vulkan.

Note: if you're building llama.cpp, all you have to do is swap GGML_HIPBLAS=1 and GGML_VULKAN=1 so the extra effort is just installing ROCm? (vs the Vulkan devtools)

ROCm:

CUDA_VISIBLE_DEVICES=1 ./llama-bench -m /models/gguf/llama-2-7b.Q4_0.gguf ggml_cuda_init: GGML_CUDA_FORCE_MMQ: no ggml_cuda_init: GGML_CUDA_FORCE_CUBLAS: no ggml_cuda_init: found 1 ROCm devices: Device 0: Radeon RX 7900 XTX, compute capability 11.0, VMM: no | model | size | params | backend | ngl | test | t/s | | ------------------------------ | ---------: | ---------: | ---------- | --: | ------------: | -------------------: | | llama 7B Q4_0 | 3.56 GiB | 6.74 B | ROCm | 99 | pp512 | 3258.67 ± 29.23 | | llama 7B Q4_0 | 3.56 GiB | 6.74 B | ROCm | 99 | tg128 | 103.31 ± 0.03 |

build: 31ac5834 (3818)

Vulkan:

GGML_VK_VISIBLE_DEVICES=1 ./llama-bench -m /models/gguf/llama-2-7b.Q4_0.gguf | model | size | params | backend | ngl | test | t/s | | ------------------------------ | ---------: | ---------: | ---------- | --: | ------------: | -------------------: | ggml_vulkan: Found 1 Vulkan devices: Vulkan0: Radeon RX 7900 XTX (RADV NAVI31) (radv) | uma: 0 | fp16: 1 | warp size: 64 | llama 7B Q4_0 | 3.56 GiB | 6.74 B | Vulkan | 99 | pp512 | 1077.49 ± 2.00 | | llama 7B Q4_0 | 3.56 GiB | 6.74 B | Vulkan | 99 | tg128 | 71.83 ± 0.06 |

build: 31ac583

This guide highlights the key features of the new SvelteKit-based WebUI of llama.cpp.

The new WebUI in combination with the advanced backend capabilities of the llama-server delivers the ultimate local AI chat experience. A few characteristics that set this project ahead of the alternatives:

Free, open-source and community-driven
Excellent performance on all hardware
Advanced context and prefix caching
Parallel and remote user support
Extremely lightweight and memory efficient
Vibrant and creative community
100% privacy

If you back up your files to Backblaze B2 using software that uploads files when they change, Backblaze B2 retains old versions of the file. This is helpful because the old versions can be used if the original file is accidentally deleted. On the other hand, keeping old versions forever can clutter things. For an application like this, you may want a Lifecycle Rule that keeps old versions in a backup/ folder for 30 days and then deletes them, as in the following example:

Keep only the last version of the file This rule keeps only the most current version of a file. The previous version of the file is "hidden" for one day and then deleted.

Cockpit is a web-based graphical interface for servers, intended for everyone, especially those who are:

Thanks to Cockpit intentionally using system APIs and commands, a whole team of admins can manage a system in the way they prefer, including the command line and utilities right alongside Cockpit.

Cockpit makes Linux discoverable. You don’t have to remember commands at a command-line.

See your server in a web browser and perform system tasks with a mouse. It’s easy to start containers, administer storage, configure networks, and inspect logs. Basically, you can think of Cockpit like a graphical “desktop interface”, but for individual servers.

Cockpit uses APIs that already exist on the system. It doesn’t reinvent subsystems or add a layer of its own tooling.

By default, Cockpit uses your system’s normal user logins and privileges. Network-wide logins are also supported through single-sign-on and other authentication techniques.

Cockpit itself doesn’t eat resources or even run in the background when you’re not using it. It runs on demand, thanks to systemd socket activation.

The US wants a clear and enforceable commitment that Iran will not develop nuclear weapons – or even the capability to do so quickly.

Washington and Tehran signed a nuclear deal in 2015 under US President Barack Obama. The agreement put a limit on Iran’s uranium enrichment of 3.67 percent in return for sanctions relief. But Trump, who succeeded Obama, withdrew Washington from the deal three years later and slapped sanctions back on Iran. Since then, Iran has accelerated its uranium enrichment to 60 percent. To make an atomic bomb, 90 percent enrichment is required.

During Israel’s 12-day war on Iran in June, the US carried out air strikes on Iran’s three main nuclear sites, after which Trump claimed that Iran’s nuclear programme had been obliterated. But eight months later, he started a war against Iran by saying one of his main goals was to keep Iran from obtaining a nuclear weapon.

The war was launched while talks mediated by Oman were under way between Iran and the US. Oman had said a short time before the attacks began that a deal was “within reach”.

Iran is pushing for a broader regional ceasefire, including an end to fighting involving its allies, such as Hezbollah in Lebanon.

While Israeli Prime Minister Benjamin Netanyahu has expressed support for Washington’s decision to suspend the strikes on Iran, he said the ceasefire will not extend to Israel’s ongoing military operations in Lebanon.

Hours into the ceasefire, which began on Wednesday, Israel carried out dozens of attacks across Lebanon, killing more than 300 people in one day.

However, Tehran insisted the ceasefire included Lebanon, citing Pakistani Prime Minister Shehbaz Sharif’s ceasefire announcement on X, which unequivocally stated this was the case.

patch intel macs to get them working with newer OS versions.

Ghostty is a fast, feature-rich, and cross-platform terminal emulator that uses platform-native UI and GPU acceleration.