brrrock

5.2K posts

brrrock banner
brrrock

brrrock

@bchap1n

powershell and llm inference struggles / shitposts currently tolerating this toxic platform to learn about AI SRE @Proofpoint (opinions are my own)

Sunnyvale, CA Присоединился Haziran 2009
1.9K Подписки466 Подписчики
Marc-André Moreau
Marc-André Moreau@awakecoding·
If you're wondering why the GitHub Copilot app on Windows mysteriously doesn't see many command-line tools which you know you've installed and are available in PATH, I think I figured out why: #issuecomment-4789727188" target="_blank" rel="nofollow noopener">github.com/github/app/iss…
English
3
2
9
1.7K
brrrock
brrrock@bchap1n·
@sudoingX 1964 Fender Jazzmaster. Probably worth 13k
English
0
0
0
23
Sudo su
Sudo su@sudoingX·
anon. what's the most expensive thing you own that exists solely because of a hobby
English
61
0
41
9.1K
brrrock
brrrock@bchap1n·
@sudoingX Fender Telecaster + Bassman amplifier
English
0
0
0
10
Sudo su
Sudo su@sudoingX·
what piece of technology ruined your standards forever
English
27
1
21
10K
brrrock
brrrock@bchap1n·
The powershell it wrote is actually ok, so may need to tweak my PSES/LSP benchmark thing.
English
0
0
0
15
brrrock
brrrock@bchap1n·
Ornith 35b gguf on single 3090 native windows latest llama.cpp +87 tok/s 🫠 benchmark + deepseek analysis is kinda funny though lol.
brrrock tweet mediabrrrock tweet media
English
1
0
0
87
brrrock
brrrock@bchap1n·
🦹‍♂️time to code some evil shit?
brrrock tweet media
English
0
0
0
11
brrrock ретвитнул
Jiwei Li
Jiwei Li@JiweiLi1·
Excited to share Ornith, our latest family of open-source models specialized for agentic coding. Ornith achieves SOTA performance among open-source models of comparable size on a variety of coding benchmarks (Terminal-Bench 2.1, SWE, NL2Repo, OpenClaw, SWE Atlas, etc) Feedback is deeply appreciated! 📖Tech Blog: deep-reinforce.com/ornith_1_0.html 🤗Huggingface: huggingface.co/collections/de…
Ornith@ornith_

Aloha! 🌺 Meet Ornith-1.0, a family of open-source LLMs specialized for agentic coding. Ornith-1.0 spans the full parameter sizes including 9B Dense, 31B Dense, 35B MoE, and 397B MoE. It achieves state-of-the-art performance among open-source models of comparable size on coding benchmarks including: ✅Terminal-Bench 2.1(77.5) ✅SWE-Bench(82.4 on verified, 62.2 on pro, 78.9 on Multilingual) ✅NL2Repo(48.2) ✅SWE Atlas(41.2 on QnA, 42.6 RF, 39.1 TW) ✅ClawEval(77.1) Post-trained on top of gemma4 and qwen3.5, Ornith-1.0 employs a novel self-improving training strategy in which reinforcement learning is used to generate not only solution rollouts, but also the task-specific scaffolds that drive those rollouts. By jointly optimizing the scaffold and the resulting solution, the model generate higher-quality solutions in agentic coding.😎 All models are released under the MIT license, enabling full commercial and research use. 📖Tech Blog: deep-reinforce.com/ornith_1_0.html 🤗Huggingface: huggingface.co/collections/de…

English
37
27
334
23K
Max Zanoga
Max Zanoga@zanoga·
@bchap1n Nah, way more boring - it just processes data all day. But at night… 😈
English
4
0
12
1.3K
Max Zanoga
Max Zanoga@zanoga·
Finally finished building my AI datacenter! 🚀 32x3090s across 4 servers (8 GPUs each), all connected over InfiniBand. The whole setup is solar-powered with a massive battery bank and generator backup. More technical details and benchmarks coming soon.
Max Zanoga tweet media
English
584
406
6.3K
786.2K
brrrock
brrrock@bchap1n·
@nixcraft Maybe so but... the Chinese are the ones releasing quality open weight models. We love Qwen. if you want support on this you gotta be giving back to open source.
English
3
0
37
925
nixCraft 🐧
nixCraft 🐧@nixcraft·
Oh, so now scraping data without permission is bad for AI training? 😂 how ironic 😉
nixCraft 🐧 tweet media
English
65
372
2.9K
41.4K
xuanming zhang
xuanming zhang@xuanmingzhangai·
@AiBattle_ Well, it’s still a personal view, but thanks for all the support!
English
6
0
45
30.7K
AiBattle
AiBattle@AiBattle_·
Looks like we may be getting the open-weight Qwen 3.7 models after all - Any plans for open-weight releases of the 3.7 family? - Qwen staff: Absolutely will do
AiBattle tweet media
English
8
3
105
8.3K
brrrock
brrrock@bchap1n·
@LottoLabs dual 3090 rig is the lowest I would go for. far more practical value for +1000 bucks.
English
0
0
1
8
Lotto
Lotto@LottoLabs·
I love hardwaremaxxing and getting high specs etc but at the end of the day If it works it works
~sarlev (e/acc)@sarlev_

@LottoLabs as someone with a build 10x this price, i would recommend going with this build.

English
3
0
20
4.1K
brrrock
brrrock@bchap1n·
@pupposandro @nb4ld @luceboxai i definitely had the order page open and watched the early discount expire. very compelling product
English
0
0
1
35
Sandro
Sandro@pupposandro·
Thanks for the purchase and support @nb4ld! Can’t wait to get it delivered and hear your feedback. We’ve received many applications for @luceboxai recently. I'm personally having a call with each one. We knew there was a problem to solve here, but we didn’t expect the demand to be this strong. Now working hard to get everyone their machine soon!
Nicolas@nb4ld

Just ordered my Lucebox. A machine I can put on a lawyer's desk and say "your AI runs here now, privately." Can't wait to push the frontier of domestic AI. Proud to support the amazing work @luceboxai.

English
8
3
25
5.3K
brrrock
brrrock@bchap1n·
@karpathy @Scobleizer at this moment in time it feels like bloat. Enterprise is carefully watching token spend and very worried about how to lock down agents. Last thing they want is to take it off the leash and put it in Slack.
English
0
0
0
202
Andrej Karpathy
Andrej Karpathy@karpathy·
This is a new paradigm for interacting with Claude that is significantly more "inline" with all the other human activity org-wide. Once you do all of the under the hood engineering work to make this "just work" (e.g. across tools, integrations, compute environments, memory, security, etc.), Claude basically joins the team in a seamless way - you can talk to it as you would talk to a person and it can help with a very large variety of workloads. Imo this is the 3rd major redesign of LLM UIUX. The first paradigm was that the LLM is a website you go to, the second was that it is an app you download to your computer. This third one is that it is a self-contained, persistent, asynchronous entity with org-wide tools and context, working alongside teams of humans. It really takes a while to wrap your head around it, but it works and it is awesome.
Claude@claudeai

Introducing Claude Tag, a new way for teams to work with Claude. In Slack, Claude joins as a team member with access to the channels and tools you choose. Tag Claude in and delegate tasks to it while you focus on other work.

English
1.1K
1.7K
21.3K
6.2M
brrrock
brrrock@bchap1n·
@0xSero true... but China does MUCH more intense versions of those awful things to pursue carefully planned long term objectives. The US does their awful things to protect the short-term interests of landlords, banks, and oil.
English
0
0
0
9
0xSero
0xSero@0xSero·
@LLMJunky I find it hard to care about the CCP when our government is cannibalising the soul of our people Every thing we claim of them we are guilty of - National CREDIT system - Government that doesn't care about freedom - Endless wars - Economic bullying etc..
English
3
0
7
221
brrrock
brrrock@bchap1n·
@tyminski_marek good call! maybe give you time to add gender fluid character attribute sliders. I am so glad you guys are Woke2
English
4
0
0
1.5K
Marek Tyminski
Marek Tyminski@tyminski_marek·
Lords of the Fallen II Development Update. Lords of the Fallen II is now set for a Q1 2027 release window. As development progresses, we have worked closely with the Gameplay Feedback Team, a dedicated group of seasoned Souls-like veterans within our Launch Creative Team. Their valuable ongoing input, combined with the team’s vision, has identified meaningful opportunities to further refine and strengthen the overall experience. These enhancements will benefit from additional integration, iteration, and polishing time, enabling the team to deliver the highest possible quality at launch. This updated release window also strategically positions Lords of the Fallen II outside of a highly competitive holiday period, ensuring the game receives the dedicated attention it deserves. The team remains fully energized by the strong response to our recent reveals and is deeply focused on delivering a standout dark fantasy action RPG and a worthy successor to the Lords of the Fallen franchise. We are grateful for the continued passion, feedback, and support from our community, and we look forward to sharing more updates in the coming months.
Marek Tyminski tweet media
English
555
405
4.1K
526.6K
brrrock
brrrock@bchap1n·
@BullTheoryio Korea will stay competitive as an industrial nation by refusing to subsidize a casino economy while Samsung and SK hynix carry the country. Foreign banks crying "policy uncertainty" when a nation stops optimizing exclusively for short-term = 🤡
English
0
0
0
541
Bull Theory
Bull Theory@BullTheoryio·
🚨 SOUTH KOREA JUST PROPOSED TAXING UNREALIZED GAINS. And this is one of the major reasons behind today's massive selloff in the Korean market, now being called BLACK TUESDAY in Korea. At a forum hosted by South Korea's ruling Democratic Party, lawmakers called for comprehensive taxation that would treat unrealized gains on stocks and real estate as taxable income, even before the asset is ever sold. The ruling party has pushed escalating wealth tax measures throughout 2026, including a February forum proposing to lower the real estate capital gains exemption threshold from 1.2 billion won to 800 million won, and an April push to abolish the long term holding tax deduction entirely. Today is the first time this campaign has explicitly extended to taxing unrealized stock gains. Under current law, investors are taxed only when they sell a stock and realize a profit. Under this proposal, investors could owe tax on paper gains they have not sold or collected, simply for holding a stock that went up in value. The Netherlands tried almost this exact policy four months ago. On February 12, 2026, Dutch lawmakers passed a law taxing unrealized gains on stocks, bonds, and crypto at a flat 36% every year, whether or not anything was sold. The backlash was immediate. A petition against it gathered more than 61,000 signatures, and Shopify CEO Tobi Lutke called it "the dumbest thing any government on planet earth is pursuing right now." Just 13 days after the bill passed, the Dutch finance minister announced the government would scrap the unrealized gains portion entirely, admitting the law "cannot pass as is." This lands directly on a South Korean market that just ran up nearly 95% over the past year, built largely on heavy retail buying with borrowed money. A tax on gains that exist only on paper is a direct threat to the exact rally that created that exposure in the first place.
Bull Theory tweet mediaBull Theory tweet mediaBull Theory tweet media
Bull Theory@BullTheoryio

🚨 THIS IS WHY EVERYTHING IS CRASHING AT THE SAME TIME TODAY. Gold, silver, and tech stocks are all down together right now, which usually signals forced selling across markets. AI SEMICONDUCTOR DELEVERAGING South Korea's Kospi crashed 10% today, triggering a circuit breaker for the second time this month. Samsung and SK Hynix both fell more than 12%. A local media report said SK Hynix is slowing down expansion of its newest chip and shifting focus to a lower priced, commodity grade chip to cover a shortfall. For a company at the center of the AI memory boom, that reads as a signal that demand assumptions are being walked back. On top of that, Korean investors had been buying chip stocks with record amounts of borrowed money, after regulators had already warned that the sector's rally had run too hot. Once the selling started, that leverage forced even more selling, which is what turned this into a circuit breaker instead of a normal pullback. QUARTER-END REBALANCING JPMorgan warned that quarter-end rebalancing could force up to $165 billion in equity selling worldwide. Big pension funds and sovereign wealth funds have to rebalance back to fixed stock-to-bond targets after a strong run in stocks. The window runs through June 30, so we are still in the middle of it. A HAWKISH FED 9 of the Fed's 19 policymakers are projecting at least one rate hike this year, and markets are pricing a 70% chance of a hike by September. That alone raises the cost of holding risk right now. USD/JPY AND A POSSIBLE YEN INTERVENTION USD/JPY had violent wicks yesterday, the kind of move that shows up when Japan steps in to defend the yen. The pair was already sitting near levels where Japan has intervened before. If that is what happened, it disrupts the yen carry trade, where investors borrow cheap yen to fund positions in stocks and other assets globally. Unwinding that trade hits unrelated markets at the same time, which lines up with gold, silver, and stocks all dropping together today. The move was on the smaller side, so a confirmed intervention isn't certain. But it's the one thread connecting everything else here. TECH SPECIFIC WEAKNESS The Nasdaq closed down 2.33% yesterday, and futures point to another 2.50% drop today. The Dow closed up +0.29% the same day, almost entirely from one stock, Caterpillar. SpaceX fell 16% yesterday, its third straight losing day, down from $176 to $154. Alphabet fell 5% on reports of AI talent leaving. Amazon, Meta, and Microsoft all fell with them.

English
279
801
3.1K
636.9K
brrrock ретвитнул
Steve Hall
Steve Hall@ProfHall1955·
Must read. In one interview Michael Hudson identifies the root of today's grossly unfair, unstable economy and sums up almost everything I've been clumsily trying to get across on this site over the past couple of years. Please share. nakedcapitalism.com/2026/06/michae…
English
27
311
1.1K
271K