nzeid

FWIW I get the most hilarious spam and strangest scanners from Netherlands.

This post so perfectly captures the tragedy of how the term AI has lost its meaning to LLMs.

Same here, but for _several_ electronics. Every advantage that Amazon has over other retailers is mooted by bootlegging. Being forced to authenticate or return products regularly is a major pain in the ass.

Held what? A majority of those companies evaporated. You'd have to cut your losses and reinvest.

I have so much to share on this topic. Will keep it short.

ROCm promises a 30-50% prompt processing speedup. This is REALLY important for my workflow so I've been trying to get this shit to work for months. But no release before v10 worked well enough with any engine for it to matter.

The llama.cpp release binaries for ROCm (10) FINALLY work on gfx1501 and its relatives (with the correct shell variables), but the prompt processing boost doesn't materialize and the token generation speed decreases.

There continues to be a chronic problem across all engines with the ROCm integration for UMA devices. The good news is that some improvements have been made to that end for Vulkan, so more recent llama.cpp Vulkan binaries are now faster.

> A good framework can maybe cache some of this with the opcode cache, and apcu. Saving some milliseconds.

This framing is confusing because OPcache has for the past 15 years taken care of exactly this bootstrapping you're worried about. In a competing share-nothing setup you merely come in at a tie for performance (barring first run of `require`). I do not see 10-50ms latencies in any of my deploys which is why I'm skeptical.

That said, the big sell of your project for me might be memory management - less memory means more workers which is always a good thing if it comes at no cost to latency.

Don't worry, I'm not coming down hard on said person. The thread is about a GrapheneOS bulletin, whose members are well-documented contributers to AOSP.

> php-fpm re-bootstraps the framework on every request (10–50ms)

What is the setup/benchmark that caused you to see this?

The phrasing of your first question is ironic given GrapheneOS has upstreamed a ton of their security hardening.

You're probably asking if they're able contribute anymore?

Reminds me of Notion and old Basecamp.

This is a very easy sell for me because everything these days is doing the infuriating gray text on gray background.

FWIW, from the OP website:

> Neobrutalism is a mix of regular brutalism in web design and more modern typography, illustration, and animation standards. Neobrutalism refuses the usual components of UX-UI design and embraces uncomfortable design elements, and it is more fearless to use distinctive color palettes.

Talk about sour grapes. Muse remains ridiculously popular and I will bet my bank account that there are more Muse fans than slop enthusiasts.

I see comments that this overlap between Qwen and GPT is due to rogue training or post hoc training. Did it occur to anyone that maybe the two sets of models were trained directly on the same solutions to the researchers' benchmark?

Definitely stinks of LLM phrasing, metering, and punctuation. But I'd say the ideas are the author's alone and they do cut through.

The edit is dead on and this is the very first thing that crossed my mind.

Bad managers hollow out companies. It's quite devastating.