Entryism Enjoyer🐵🧦🏗️ (@SecPerkinsStan) header
Entryism Enjoyer🐵🧦🏗️ (@SecPerkinsStan) profile picture

He/Him. Fallback: @secperkinsstan@mstdn.social

Joined July 2016
Photo by Entryism Enjoyer🐵🧦🏗️ (@SecPerkinsStan)Photo by Entryism Enjoyer🐵🧦🏗️ (@SecPerkinsStan)Photo by Entryism Enjoyer🐵🧦🏗️ (@SecPerkinsStan)Photo by Entryism Enjoyer🐵🧦🏗️ (@SecPerkinsStan)Photo by Entryism Enjoyer🐵🧦🏗️ (@SecPerkinsStan)Photo by Entryism Enjoyer🐵🧦🏗️ (@SecPerkinsStan)Photo by Entryism Enjoyer🐵🧦🏗️ (@SecPerkinsStan)Photo by Entryism Enjoyer🐵🧦🏗️ (@SecPerkinsStan)Photo by Entryism Enjoyer🐵🧦🏗️ (@SecPerkinsStan)Photo by Entryism Enjoyer🐵🧦🏗️ (@SecPerkinsStan)
Entryism Enjoyer🐵🧦🏗️ retweeted
Pretty extraordinary Nvidia data point: Israel’s economy grew at an 8% annualized pace in Q3. Roughly HALF of that growth came from overseas production, largely tied to revenue generated by Nvidia’s Israeli arm. Strip that out, and growth was 4%.
27
89
12
818
73,778
Entryism Enjoyer🐵🧦🏗️ retweeted
The “they leaked it because Dems won the midterms” thing is less plausible to me than “they waited until after the midterms to leak so it wouldn’t hurt the GOP at the ballot box.”
9
25
4
342
44,753
Entryism Enjoyer🐵🧦🏗️ retweeted
Hearing that SemiAnalysis cut its 2028 AI accelerator capex forecast by 28%, citing slower growth at frontier labs and tightening capital markets in a note to institutional clients
82
99
70
1,200
354,730
Entryism Enjoyer🐵🧦🏗️ retweeted
Who absorbs the cost of China's AI buildout? Three stories this week showed the bill splitting across fab partners, staff and public equity. On November 12, SMIC, China's largest contract chipmaker, reported third-quarter total profit doubled to $651m under international accounting standards, yet $223m, or 34.3%, went to minority shareholders in its subsidiaries. Even after its SMIC North buyout completed its equity transfer on June 12, SMIC South in Shanghai is taking in $7.78bn in fresh joint-venture capital. At Tencent's 28th anniversary broadcast, management said 50% of staff should act like the army in core businesses and 50% like the navy exploring AI Native, Chinese outlet The Paper reported on November 12. A person close to Tencent denied rumors of 50% layoffs, calling it resource reallocation for the AI era. Alibaba announced on November 10 that T-Head, its in-house chip unit, filed confidentially on September 11 to list H shares on Hong Kong's main board, with the offering size and stake reduction not yet finalised. As I read the week, Chinese tech firms are shifting the AI cycle's cost across state fab partners, internal headcount and outside shareholders. Every Sunday I pick the China tech stories that mattered that week, say what I make of each, and link back to where I first wrote about it. Today's letter is free to read. hellochinatech.com/p/chinas-…
1
3
1
4
1,622
Entryism Enjoyer🐵🧦🏗️ retweeted
The open model world faces two main challenges in a post-Jensen world: 1. Political exposure. Open models’ allies are retreating in DC. 2. Investment. Open model builders need to find profitable products far sooner w/o his backstop. Will feel it sooner than people expect.
10
48
17
355
65,057
Entryism Enjoyer🐵🧦🏗️ retweeted
Nobody has a hopeful vision, a whacked out fanfiction cult was allowed to frame the elite discourse, there isn't even a live movement to implement what socialism was supposed to be, the industrialists building it are the least trustworthy figures in the history of industry...
2
3
60
1,219
Entryism Enjoyer🐵🧦🏗️ retweeted
This is the month of the Big Update for a great many people, it seems, and now Dwarkesh Patel has joined us as well. Dwarkesh this morning:
73
181
29
2,422
155,687
Entryism Enjoyer🐵🧦🏗️ retweeted
''>A Turkish man invents the steam turbine 200 years before the industrial revolution >only uses it to spin döner kebab''
>A Turkish man invents the steam turbine 200 years before the industrial revolution >only uses it to spin döner kebab
111
7,178
45
105,929
1,070,334
Entryism Enjoyer🐵🧦🏗️ retweeted
Another Chinese Kintex7 clone manufactured in 28nm. It even has a DDR4 memory interface. Cost for the 160T variant. 60 USD.
31
39
4
691
42,910
Entryism Enjoyer🐵🧦🏗️ retweeted
Really interesting work.
Why is AI slop so frustrating to read? AI writing often obscures main ideas and invents awkward phrases, making sentences more complex than necessary. Many of these slop patterns violate writing principles established in English writing books. In our benchmark, AI writing breaks these principles far more often than human writing, with Claude Code replies by far the worst offenders. We show examples of the most common slop patterns in this thread
2
3
63
7,609
Entryism Enjoyer🐵🧦🏗️ retweeted
I got a 1.9B model to watch a restaurant camera and tell staff which tables need cleaning 🍽️ It runs on my laptop. No cloud, and it generates zero tokens. Built over a weekend with @liquidai d1-2B
309
402
71
7,131
406,513
Kimi and 1017 are closer in perf than they may appear and the frontier is further away from both
Replying to @teortaxesTex
Yeah, the drop by Kimi on ArXivMath is also a bit sus, ngl
1
3
39
3,731
Entryism Enjoyer🐵🧦🏗️ retweeted
GPT-4 level intelligence has gotten roughly 1800x smaller
50
127
17
2,713
125,010
Entryism Enjoyer🐵🧦🏗️ retweeted
I think it’s only downhill from now on.
Prediction: next gen non-cyber models will reject formal verification tasks because if you can formally verify something, you can technically enumerate all bugs that invalidate the property you verify; so a sufficiently good verifier is a potentially amazing hacking tool.
3
29
1,904
Entryism Enjoyer🐵🧦🏗️ retweeted
“We also invite you to subscribe to our email alert service”
The Houthis are running an email subscription service What a time to be alive
19
222
4
3,365
48,617
Entryism Enjoyer🐵🧦🏗️ retweeted
Zyphra's new paper proposes Zyphra Hierarchical Attention (ZHA), which organizes the KV-cache into a tree of compressed summaries. Queries search the tree from top to bottom, narrowing down which tokens deserve attention. With just 1B tokens of additional distillation training, ZHA preserves most of the performance of full attention on models up to 35B parameters. On RULER at 256K context, it achieves 93.3% vs 96% accuracy, while attending to just 1-6% of the KV-cache. alphaxiv.org/abs/2610.42112
9
28
2
211
9,963
Entryism Enjoyer🐵🧦🏗️ retweeted
This is pretty much how things have gone... ANTHROPIC: "AI is extremely perilous and should be carefully regulated" CALIFORNIA GOV: "Ok, we will regulate you." ANTHROPIC: "No! Good regulation is really hard to get right. We know so little about what to do. Instead we should let companies carefully iterate over time." CALIFORNIA GOV: "Oh ok. How about we do a thing where you get to write what the rules are but we will hold you to them? Then you can make your own commitments as you think is best but they will be true commitments!" ANTHROPIC: "Nah, we'd rather make fake commitments." CALIFORNIA GOV: ...
If Anthropic wants me to think these steps are real/credible, they should put it in their SB 53/RAISE Act Frontier Compliance Framework. State law gave Anthropic and OpenAI a mechanism for making credible statements about their behavior, and they mostly choose not to use it on the risks they are most worried about (neither company includes anything in their compliance frameworks about internal deployment RSI risk, though Anthropic says maybe at some point in the future they will. Now seems like as good a time as any???!!!???)
4
28
572
23,990
Entryism Enjoyer🐵🧦🏗️ retweeted
"We opted not to buy any posttraining data, instead devising our own synthetic data pipelines, quality filtering, and coverage metrics." has quite large implications when combined with the fact the whole thing cost 32M.
Today we release Ripple-65B Preview, our 65B dense LLM we pretrained from scratch. Ripple sets a new state-of-the-art for U.S. open-weight models in agentic coding tasks, surpassing significantly larger models including Muse Spark Open-816B, Beam-501B, Inkling-975B, and Nemotron 3 Ultra-550B. Learn more: ricursive.com/ripple
2
6
1
96
7,920
Entryism Enjoyer🐵🧦🏗️ retweeted
This is hilarious. No matter how many times MediaTek asked, nobody at Intel would get off their ass. MediaTek actually had to hire a former top Intel exec just to get them moving lmao. How is Intel’s culture still this much of a mess? LOL
62
34
9
683
98,497
Entryism Enjoyer🐵🧦🏗️ retweeted
*sshole is out if you run sshd on alpine, openwrt or any musl box with password auth on: update now official writeup: sshhole.com
86552c7e1e5d46882844cffabb44d454129ff22927f13797b151c9db4519d47d
188
890
446
5,450
1,572,712