Quis_Cust

[Quis_Cust @Ipsi_Cust]

@MiaAI_lab Seriously you think they could have hit the ground with something cooler then that. Most of us local AI Cats are cooking way more cooler and capable shit then this..

[Quis_Cust @Ipsi_Cust]

@coldniko Hey so do you think its possible to https://t.co/ab36MUeC5C from @Blackfrost_AI running on Strata on a 12g vram /96g rig at there lowest quant? Have you released any tools for bridging compatibility?

[Quis_Cust @Ipsi_Cust]

@bygregorr @support_huihui Ive been looking for some different Abliteration quants and its rather scarce for Next Flash. Have you found any that have been worth it?

[Quis_Cust @Ipsi_Cust]

@RoundtableSpace This thing stinks like malware or something sneaky

[Quis_Cust @Ipsi_Cust]

@Blackfrost_AI @OmarchyLinux

[Quis_Cust @Ipsi_Cust]

@coldniko @itsmissglitch Abilterated quants are welcome or at least conversion tech for them would be nice if you have any ideas

[Quis_Cust @Ipsi_Cust]

@coldniko @itsmissglitch I have 12G of VRAM and 96G of ram. Im happy to get a stack that can serve Next Flash at a decent quantity at speed. I tried to scope out Strata but couldn't get a good fit for my hardware on the compatible quants

[Quis_Cust @Ipsi_Cust]

@puterlad @sxhivs I have a VMOS magic Box server which has 12 virtualized androids slots on it. Its hardware and I have my own swarm stack that sits on top of it as an adaptive cross fleet controller

[Quis_Cust @Ipsi_Cust]

@boneGPT To signin to grok bot it hoped through cursor then to grok then to X signin then back to cursor to sign into grok bot. Surely there has to be a better way

[Quis_Cust @Ipsi_Cust]

@arc_4n0n @boneGPT Distributed Denial of Sneaky

[Quis_Cust @Ipsi_Cust]

@ayayalar @NousResearch It should ne able to also run agentic flows o the phone itself. If you look at what MacDroid cam do we could do something similar where it opems the door to all kinds of internal mobile phone workflows as well. I feel like a lot of people have been calling for this

[Quis_Cust @Ipsi_Cust]

@ayayalar @NousResearch I am serious. I already have all the primitives and a mobile app development, testing and deployment harness. I guess the task is making it modular so it can integrate into the home server harnesses. We could have it be a meta-harness to control downstream harnesses.>>>...

[Quis_Cust @Ipsi_Cust]

@FishAudio DRAMA

[Quis_Cust @Ipsi_Cust]

@ayayalar And the Chinese will drop another turd on their valuation. Ive never understood the infatuation for anthropic. Looking at their cost versus GLM 5.3 flash and performance its will to me how they can justify a 10 figure valuation. With the right harness GLM cooks

[Quis_Cust @Ipsi_Cust]

@theiaincameron In my m?);- 'm

[Quis_Cust @Ipsi_Cust]

@ItsmeAjayKV Ive noticed Grok isnt the best at UI/UX to be honest compared to deepseek or GLM 5.3

[Pirate Face @thepirateface]

Introducing the Pirate Face tracker - its distributed hash table (DHT) and BitTorrent tracker.

Every listed magnet points clients to:
udp://tracker.pirateface.co:6969/announce

The endpoint speaks BitTorrent to torrent clients, helping peers carrying the same model find one another faster and easier.

Read more ↓

[Pirate Face @thepirateface]

🧲 Community torrents · 22 Sept 2026

30 model torrents added:
• hexgrad/Kokoro-82M
• Qwen/Qwen3.6-27B
• zai-org/GLM-OCR
• lightx2v/Minimax-h3-Turbo
• unsloth/gemma-4-12B-it-qat-GGUF
• google/gemma-4-12B-it-qat-q4_0-gguf
• + 24 more

Browse new magnets ↓