🅰🆁🅻🅾

1.7K posts

🅰🆁🅻🅾 banner
🅰🆁🅻🅾

🅰🆁🅻🅾

@arlogilbert

Founder, Father, Author, Jew, Texan. 🇺🇸 🇮🇱 ✡️🤘

Austin, TX Katılım Mart 2008
1.2K Takip Edilen7.4K Takipçiler
🅰🆁🅻🅾
🅰🆁🅻🅾@arlogilbert·
@AlibabaGroup When do we get it in the US region? I see it in Signapore and China, but latency :(
English
0
0
0
4.3K
Alibaba Group
Alibaba Group@AlibabaGroup·
Qwen3.7-Max is live! 🚀 Introducing the latest proprietary model, built for advanced agentic coding, complex reasoning, and long-horizon execution. It’s here to transform how we approach complex tasks.
Alibaba Group tweet media
English
54
181
1.4K
942.5K
🅰🆁🅻🅾
🅰🆁🅻🅾@arlogilbert·
AMD ROCm is great, but hit-or-miss for AMD model self-hosters. Just boosted tokens/sec from 99 to 160 by switching from ROCm to Vulkan. Qwen 3.6 35B Q4 MTP on AMD Radeon W7900 Vulkan backends have received massive optimization lately. ROCm is great but Vulkan has gotten better faster.
English
1
0
0
58
🅰🆁🅻🅾
🅰🆁🅻🅾@arlogilbert·
@Alibaba_Qwen When will it be available in ModelStudio for US Virginia? I see it in Singapore and China but no love for us Americans?
English
0
0
1
373
Qwen
Qwen@Alibaba_Qwen·
📣Meet Qwen3.7-Max — our latest flagship, made for the Agent Era. A versatile foundation for agents that actually get things done: 🧑‍💻 Coding agent, end to end. Frontend prototypes, multi-file refactors, real debugging — nails it. 🗂️ A reliable office and productivity assistant. Get your work done through MCP integrations and multi-agent orchestration. ⏱️ Long-horizon autonomy. 35 hours straight on a kernel optimization task — 1,000+ tool calls, zero hand-holding. 🔌 Scaffold-agnostic. Claude Code, OpenClaw, Qwen Code, or your own stack. Consistent reliability everywhere. API's up on Alibaba Model Studio. You can also take it for a spin on Qwen Studio. Go build something wild!🏃🏃‍♂️ 📖 Blog: qwen.ai/blog?id=qwen3.7 ✅ Qwen Studio: chat.qwen.ai/?models=qwen3.… ⚡️ API:modelstudio.console.alibabacloud.com/ap-southeast-1…
Qwen tweet media
English
229
508
3.9K
612.5K
Sudo su
Sudo su@sudoingX·
anyone here actively running grok build, what are you finding? i have been testing it for a few days and have my own early read, but the user base is small enough that i would rather hear from a few of you before i write anything up. specifically curious about long running tasks, how the agent holds up over hours, multi file work, and how it compares to whatever you were using before. drop your honest read in the replies, i will amplify the good ones.
English
33
0
54
5.6K
🅰🆁🅻🅾
🅰🆁🅻🅾@arlogilbert·
@NetworkChuck I don't understand the almost cult-like love of OpenClaw. My Hermes instances stay up and get work done while OpenClaw users argue about religion.
English
0
0
0
50
🅰🆁🅻🅾
🅰🆁🅻🅾@arlogilbert·
Sure, so we agree he's imperfect. People aren't "Liars" or "Not Liars". You lie sometimes I assume, does that mean you are a liar or does it mean you sometimes lie? Either way, Platners comments had nothing to do with whether or not Kyle was honest. Platner hadn't ever met him, never worked with him, didn't get into the same Elite program that Kyle was in, and he's making up shit about how a highly decorated soldier was somehow lying and killing civilians. Whether or not you like Chris Kyle, that's not the behavior of a senator.
English
2
0
0
56
Charlie J. Johnson
Charlie J. Johnson@Charliemagne·
@arlogilbert @FreeBeacon Yes, it’s ok to point out liars are liars and not reliable narrators. Chris Kyle literally said he gunned down unarmed people in NOLA.
English
1
0
5
73
Washington Free Beacon
Washington Free Beacon@FreeBeacon·
Left-wing Maine Senate candidate Graham Platner smeared the late “American Sniper” Chris Kyle in a 2024 podcast interview, suggesting that Kyle shot innocent civilians in Iraq to inflate his kill numbers. Platner also bristled at the fact that Kyle and members of his platoon were described as “heroes,” calling it “a weird practical joke.”
English
765
1.1K
3.5K
2.8M
Charlie J. Johnson
Charlie J. Johnson@Charliemagne·
@arlogilbert @FreeBeacon All people are flawed, but most don’t commit their flaws to writing, then sell them for money — except for Chris Kyle, Tim Kennedy, and Robert O’Neil, I guess.
English
1
0
5
150
🅰🆁🅻🅾
🅰🆁🅻🅾@arlogilbert·
@Charliemagne @FreeBeacon The point isn't that Chris was perfect. The point is that Platner is a dumb drunk who has no problem shitting on dead soldiers. Just what we'd want in a fucking senator.
English
0
0
0
29
🅰🆁🅻🅾
🅰🆁🅻🅾@arlogilbert·
@Charliemagne @FreeBeacon Find me an unflawed soldier and then explain to me how you decide that some dead soldiers are okay to shit on while others aren't.
English
2
0
0
127
Deno
Deno@deno_land·
Introducing Claw Patrol! A security firewall for AI agents. It sits between your agents and production, parsing and gating HTTP, SQL, Kubernetes, and SSH on the wire. Built from our experience giving ops agents production access without sharing credentials. MIT licensed.
English
20
21
366
56K
🅰🆁🅻🅾
🅰🆁🅻🅾@arlogilbert·
@danielhanchen I'm using it on AMD Radeon wx7900 48GiB and getting 98 t/s with 2x concurrency. Performance has been awesome. MTP took it from 56 to 96 in my setup. Thanks!
English
0
0
0
28
Daniel Han
Daniel Han@danielhanchen·
Unsloth Studio now has auto speculative decoding & MTP support for GGUFs! Get up to 2x faster inference with no accuracy loss! We ran many experiments from small models to MoEs, and optimized the params for Mac, GPUs & CPUs. There's also a new toggle for MTP / ngram or auto!
Unsloth AI@UnslothAI

4-bit Qwen3.6 MTP GGUF managed to search 70+ sites from a single prompt. Try this locally on 20GB RAM via Unsloth Studio. Unsloth now supports auto MTP + speculative decoding & auto-selects the best MTP settings for your device (Mac, CPU, GPU). GitHub: github.com/unslothai/unsl…

English
6
16
185
14.7K
🅰🆁🅻🅾
🅰🆁🅻🅾@arlogilbert·
@LindseyGrahamSC Chris Kyle died helping another soldier learn to cope with PTSD. Chris Kyle was my friend. Platner is a loser now, was a loser then, and will be a loser when he chokes on whatever cheap liquor he swills before every podcast.
English
0
3
6
249
Lindsey Graham
Lindsey Graham@LindseyGrahamSC·
To my Democrat colleagues: Raise your voice in condemning the most vile assault and slander in recent memory by a candidate for the U.S. Senate against an American soldier risking his life for our country. The soldier in question received the Purple Heart fighting our enemies. As to Platner, at what point does it become obvious this guy is unworthy and unfit to be dogcatcher, much less a member of the U.S. Senate? The bottomless pit of left-wing wackos just got deeper. Hopefully it’ll hit a wall when the good people of Maine reject Platner’s outrageous conduct on multiple fronts.
Fox News@FoxNews

“Dumb motherf****** didn’t deserve to live.” Far-left Democrat Senate candidate Graham Platner mocking a U.S. soldier who was shot four times in Afghanistan in a resurfaced Reddit comment. The post made fun of Purple Heart recipient Pfc. Ted Daniels after he took incoming fire during a 2012 clash with the Taliban. Platner’s now deleted comment went on to say: "At least his stupidity and fat ass wheezing are available for all future infantrymen to witness and hold in contempt. Poor marksmanship on the Taliban's part is the only reason this mouthbreather made it home, he managed to make every possible s*** decision possible when it comes to small unit combat."

English
537
608
2.7K
594.5K
🅰🆁🅻🅾
🅰🆁🅻🅾@arlogilbert·
@rscottri @FreeBeacon Pussies like you wouldn't have said that to his face. He was shot by another soldier who he was trying to help get over PTSD.
English
20
0
0
3.1K
🅰🆁🅻🅾
🅰🆁🅻🅾@arlogilbert·
@DODOREACH It's nicely done. The only bug I've identified so far is that the first click on Edit for a saved host always opens with all blanks values, second click always works.
English
0
0
1
26
dodo-reach 🦤
dodo-reach 🦤@DODOREACH·
@arlogilbert please open issues, PRs, or just dm me when you hit bugs or rough edges 😊 still one guy trying to make Swift, weird configs and SSH behave. Feedbacks helps a lot!
English
1
0
0
29
🅰🆁🅻🅾
🅰🆁🅻🅾@arlogilbert·
Hermes > OpenClaw Running agents locally sounds cool, but if you want an always on agent you either need a VPS or a local machine and you need a channel/method by which you can interact with the agent. Until now your choices have mostly been chat platforms or wonky web interfaces. Although telegram is "fine" it's really not built for managing and interacting with agents beyond simple chats. Installed and tested this Hermes desktop app and it is quite awesome! Early, has little bugs, but it lets you work on a native desktop app and uses SSH to connect to your Hermes server. The end result, it feels like hermes is running locally.
dodo-reach 🦤@DODOREACH

Hermes agent's tui got REALLY good @NousResearch and @Teknium improved it so much that hiding it inside a weak web/chat wrapper feels wrong now Hermes Desktop v0.9.0 is out and it's not a fake chat UI. It’s a clean, safe door to your agent SSH to your host open Hermes work 🚀

English
2
1
5
593
Youssof Altoukhi
Youssof Altoukhi@Youssofal_·
Managed to get QWEN 3.6 27B running in Cursor with localhost (no Ngrok) at 120 - 140 TPS on 2x 3090s.
English
36
9
307
20.3K
🅰🆁🅻🅾
🅰🆁🅻🅾@arlogilbert·
@Youssofal_ Impressive, I'm on an AMD Radeon WX7900 48GiB VRAM and get 90t/s peak. The big question nobody has asked, how's the concurrent request performance? With Hermes & OpenClaw at least I've found that I need to plan for 3-5 concurrent requests.
English
0
0
0
79
🅰🆁🅻🅾
🅰🆁🅻🅾@arlogilbert·
Use a smaller model than you think you'll need. That model that your 128GiB "just barely fits" is useless for anything other than PoC, it won't function for real work. Layer on memory, title generation, sub-agents, recurring tasks, and you'll quickly find that you need to support 5-10 concurrent connections for one person to work.
English
1
0
0
82
Hermes Agent Tips
Hermes Agent Tips@HermesAgentTips·
When working with Local LLMs what was one thing you wished you knew when you first started your Local LLM journey?
English
16
1
11
1.7K