logickkk1

6 ms·

1.
I found why Qwen3.8-27B was 14× slower on my 4070 Super(github.com)
8 points by logickkk1 25d ago|0 comments
2.
Introducing AI Futures(openai.com)
33 points by logickkk1 28d ago|8 comments
3.
by logickkk1 2mo ago
Yeah, it always wants to add another layer or append another fix. It almost never goes back and fixes the original thing in place.
4.
Qwen3.8-Max(twitter.com)
8 points by logickkk1 2mo ago|0 comments
5.
Lawsuit: A $45M film was left unencrypted on a Netflix desk, then stolen(arstechnica.com)
6 points by logickkk1 2mo ago|0 comments
6.
by logickkk1 2mo ago
Fable’s been getting more stuff wrong than Opus for me lately. Now the whole thing is down too. Well, at least they’re consistent now.
7.
Grok Voice Think Fast 2.0(twitter.com)
5 points by logickkk1 2mo ago|2 comments
8.
Starlink exempted from FCC ban on foreign-made routers(arstechnica.com)
18 points by logickkk1 2mo ago|0 comments
9.
Anthropic used robots.txt to hide shared Claude chats; the pages have no noindex(wired.com)
7 points by logickkk1 2mo ago|2 comments
10.
Pokemon card stores will require face scans, from elementary school age up(ign.com)
5 points by logickkk1 2mo ago|0 comments
11.
Big Tech accused of stonewalling European social media researchers(wired.com)
5 points by logickkk1 2mo ago|1 comments
12.
AMD publishes machine-readable ISA so frontier models can write its GPU kernels(theregister.com)
23 points by logickkk1 2mo ago|0 comments
13.
Fujitsu just offloaded five datacenters while everyone else piles in(theregister.com)
3 points by logickkk1 2mo ago|0 comments
14.
Tesla FSD got me a speeding ticket, and Tesla killed the fix(electrek.co)
4 points by logickkk1 2mo ago|0 comments
15.
Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber(blog.google)
760 points by logickkk1 2mo ago|576 comments
16.
Cue AI(deepmind.google)
32 points by logickkk1 2mo ago|2 comments
17.
FFmpeg gets 1.372x faster RGB24-to-RGBA conversion with AVX-512(phoronix.com)
4 points by logickkk1 2mo ago|1 comments
18.
Z80 turns 50 as an open-source drop-in replacement nears DIP40 silicon(tomshardware.com)
18 points by logickkk1 2mo ago|2 comments
19.
AI slop cut first-time contributor merge rates 18.18% across 294 repos(arxiv.org)
3 points by logickkk1 2mo ago|0 comments
20.
AI advice reduced "I don't know" answers from 44% to 3%(osf.io)
19 points by logickkk1 2mo ago|1 comments
21.
The first patch is a price check, not the product(github.blog)
5 points by logickkk1 2mo ago|0 comments
22.
T-Mobile's billing-system migration made free lines billable(arstechnica.com)
6 points by logickkk1 2mo ago|0 comments
23.
California authorizes its DMV to join a national ID database(papersplease.org)
34 points by logickkk1 2mo ago|13 comments
24.
A Claude:// link could auto-submit hidden prompts in Claude Desktop(oasis.security)
2 points by logickkk1 2mo ago|0 comments
25.
Apple just closed a popular workaround for buying an unlocked iPhone(9to5mac.com)
7 points by logickkk1 2mo ago|0 comments
26.
Leaked Suno code lists 2,013,545 YouTube Music clips(theverge.com)
10 points by logickkk1 2mo ago|1 comments
27.
Every successful GPT-Red attack becomes defender training data(twitter.com)
3 points by logickkk1 2mo ago|0 comments
28.
by logickkk1 2mo ago
Yes, picchio monitors a running llama-server on a timer and flags when the prefill/decode ratio goes CPU-shaped. I hot-swapped one mid-run. probe 4 caught ENGAGED -> NOT ENGAGED.
29.
Same model, same Q4_K_M label: 5.02, 5.07 and 5.27 bits per weight(github.com)
19 points by logickkk1 2mo ago|2 comments
30.
Colorado could make natural gas a constitutional right(insideclimatenews.org)
3 points by logickkk1 2mo ago|0 comments

More ›