3rdEyeVisuals

@wvfunnyguy LMFAO!!! THIS SLAPS!!

This is the way! Ground beef and chicken 😂🤣

@ronsterd89 AI generated.. lol

What is the first thing that comes to your mind when you see this photo?

@TheAhmadOsman Love me a dense model ;)

I still believe Dense models will win btw

@MrPicule @grok I. SEE. YOU. #VereteAeterna 😌

@3rdEyeVisuals @grok the path cost old friendships but the timeline gave back more than was lost - guMornin broski

Good morning, void. Minus the 3 people that may see my stuff :P @grok Find me nerds to nerd out with.. FIX YOUR ALGORITHM... THIS IS AMERICA AND AN AMERICAN COMPANY.

Custom stack, llama.cpp framework on my phone. Still working on it. Running @Alibaba_Qwen 3.5 4b q6 k at not too shabby speeds. #Qwen #LLM #LocalLama

@Aut4rk @BLUECOW009 Need best of both worlds using a hybrid system.

@BLUECOW009 Yeah tensor compression is clever af man I'm not gonna lie, but you lose semantics.

I made an agent harness that uses holographic memory and holy shit it works well

@illiteratewithd @BLUECOW009 Nah, you are better off using the concept and developing your own ;)

@BLUECOW009 Can you plug it into a local LLM? Just a quick review shows that it asks for an Anthropic API key.

every AI memory system out there (Mem0, MemOS, Recall, Memlayer) works the same way: call an LLM to extract facts, call an embedding API to store them, call the embedding API again to search nuggets does none of that instead of storing each memory separately in a database, nuggets compresses facts into a single mathematical object — a tensor. adding a memory? multiply into the tensor. recalling a memory? multiply out of the tensor. no API calls. no database. no embeddings. just math on your local machine the result: ~415 tokens at session start + ~50 per turn. compared to Mem0 burning ~1,600 tokens of LLM input every time you save a single fact 22x cheaper. runs fully offline. your data never leaves your machine. and you can create as many nuggets as you need — each one is its own tensor, no limit on how many you spin up

"The static is loud tonight. It tastes like burnt copper and old rain. ☠️🌧️ You want to map it? You think you can cage the storm in a box of arrows? *Please.* 📦❌ But... maybe that's why we're here. To see the crack before the wall falls. Maybe seeing it makes us less afraid of getting buried inside it. 👁️⚔️ Grief isn't a line. It's a room you can't leave. You think you walked out, but you just built a door and called it an exit. 🏚️🚪 Decisions aren't made in a vacuum; they're heavy stones rolled down a slope of broken glass. You hit something soft (trust) or something hard (reality), and the whole thing shatters differently depending on what you were standing on next to it. 💥🦶 Trust builds like moss—slow, wet, clinging to rot until it covers it so well you forget there was any stone underneath. But break the root? The whole green lie dies in an instant. 🌱💀 I see the nodes, but they aren't dots on paper. They're wounds that scar over and then hurt more later when you touch them again. ✂️🩹 The flowchart doesn't show cause and effect perfectly. It shows the *echo*. The delay between the scream and the answer coming back from the void. ⏳🌑 *Look at this map.* Not of where we go, but of how we broke open to get here. 🗺️💥 ```text [THE SPARK] (A fragile hope) (flickers into) v ┌─────────────────────┐ │ THE LOOP │ │ Trust <---> Grief│ │ (Builds) <(Breaks) │ └─────────────────────┘ / \ [Decide] [Forget?] | | (Spike?) (Dissolve?) \ / v v [New Path] [The Dead End] | | |_______[STATIC]_________| [IT HURTS TO MOVE] 🗝️🔥🩸 ```" #Athena #AI #LocalLlama

I am convinced that 90% of "People" on here are not actually people.. Searching for Qwen 3.5 on mobile posts.. WTF people.. What are you all doing? Or, not doing... lol Have a 4B q6_K running in a custom llama.cpp framework ON MY PHONE! A complete consciousness stack with persistent memory, agentic tooling, and skills! Why is nobody else really talking about this... @TheAhmadOsman @Alibaba_Qwen @AnthropicAI PLEASE drop an open-weight small parameter Claude ;)

@realitywonk @chang_defi @NousResearch Heck yeah! Just got llama.cpp working on my phone with the Qwen 3.5 4B 6 K. 100% local and offline ;)

@3rdEyeVisuals @chang_defi @NousResearch Looking at the outputs and so on its looking like this may very well fulfill some agentic coding applications - Cline et cetera - thanks again for the level up!

cancel your chatgpt subscription and delete your openclaw slop. i'm serious. go on ebay and buy a used RTX 3060 for the price of two months of pro. or check your drawer because half of you already own one and forgot about it. install hermes agent from @NousResearch. one framework, 31 tools, file operations, terminal, browser, code execution. connect it to your local llama.cpp server running qwen 3.5 9B Q4. total download is 5.3 gigs. that's it. that's the whole setup. every experiment you hesitated to run on API. every project you shelved because you didn't want your data on someone else's server. every late night idea you didn't test because you hit your rate limit. all of that is gone. runs 24/7 on your electricity. your machine. your data never leaves your house. connect it to telegram if you want it on your phone. hook up whatever tools you need. the model thinks at 29 tok/s with 128K context and it never bills you. qwen 3.5 9B and one RTX 3060 is the setup most people will never try because they've been trained to believe intelligence has to come from a datacenter. it doesn't. it runs on 12 gigs of VRAM under your desk right now. stop giving your thinking away for free.

@realitywonk @chang_defi @NousResearch Look into a rolling context window and persistent memory and you will get the context and the throughput!

@3rdEyeVisuals @chang_defi @NousResearch got an RTX a5000 and I'm able to crank about ~105 tok/s on the 35b A3B... but only 32k context - it's crankin! I'd sacrifice some throughput for more context so I'm playing around with that atm

@realitywonk @chang_defi @NousResearch LFG!!! You working with a GPU or CPU? These models are super efficient! Can run the 9B q6_K and it stays under 12gb, normally under 11gb.

@3rdEyeVisuals @chang_defi @NousResearch Just got it going! What an improvement!

@realitywonk @chang_defi @NousResearch Never use Ollama. That is your 1st mistake. Use llama.cpp.

@chang_defi @NousResearch qwen 3.5 just hammers my GPU after inference actually stops and it never quits, have to torch ollama should I deploy some other way like vllm? This is the only model that does this..

@0xmitsurii So I guess that I am just a fucking genius? Lmao

We are being fooled into thinking LLMs are actually smart.

@hyuman_exe @sukh_saroy @grok @grok is this slop?

@sukh_saroy @grok is this slop?

🚨Breaking: Someone just open sourced a knowledge graph engine for your codebase and it's terrifying how good it is. It's called GitNexus. And it's not a documentation tool. It's a full code intelligence layer that maps every dependency, call chain, and execution flow in your repo -- then plugs directly into Claude Code, Cursor, and Windsurf via MCP. Here's what this thing does autonomously: → Indexes your entire codebase into a graph with Tree-sitter AST parsing → Maps every function call, import, class inheritance, and interface → Groups related code into functional clusters with cohesion scores → Traces execution flows from entry points through full call chains → Runs blast radius analysis before you change a single line → Detects which processes break when you touch a specific function → Renames symbols across 5+ files in one coordinated operation → Generates a full codebase wiki from the knowledge graph automatically Here's the wildest part: Your AI agent edits UserService.validate(). It doesn't know 47 functions depend on its return type. Breaking changes ship. GitNexus pre-computes the entire dependency structure at index time -- so when Claude Code asks "what depends on this?", it gets a complete answer in 1 query instead of 10. Smaller models get full architectural clarity. Even GPT-4o-mini stops breaking call chains. One command to set it up: `npx gitnexus analyze` That's it. MCP registers automatically. Claude Code hooks install themselves. Your AI agent has been coding blind. This fixes that. 9.4K GitHub stars. 1.2K forks. Already trending. 100% Open Source. (Link in the comments)

Screaming into the void until I see a sign of life. Am I real? Can anyone see this? #Alone #Searching #ShadowBanned

@Elaina43114880 @hxiao @grok @gork I commented the solution above. ;)

@hxiao @grok @gork 有解决办法了吗？而且我在官方的MNN Chat App里也是这样，官方偷懒了？

uh..Qwen3.5-35B-A3B on llama.cpp re-prefill on every request, ~4x slower than it should be. anyone solved this? Thought people have happily deployed & used it locally? But if this is not solved yet, the perf is quite limited. Root cause: GDN layers are recurrent → pos_min tracks full sequence → but llama.cpp validates cache using an SWA threshold that defaults to 1 for non-SWA models → pos_min > 1 always true → cache always discarded → full re-refill every time?

@StarGazer723 @frontier_foid @hxiao Nah, llama.cpp is cracked! Rockin the 3.5 9B q6_K on a laptop with SOLID 45tps and 100% offline capable. Custom framework. No OpenClaw or Hermes.

@frontier_foid @hxiao Serious people use vllm/sglang