The Way

@subquadratic We have do we check? 🤔

The transformer architecture used for ChatGPT, Gemini, and Claude has defined the last decade of AI. It also introduced a fundamental constraint: compute scales quadratically as context grows. Longer inputs, exponentially higher costs and accuracy that degrades well before the context window limit. SubQ changes that. It's the first LLM that breaks the quadratic scaling constraint delivering longer context, higher accuracy, and lower cost at the same time without tradeoffs. Read more here. subq.ai/introducing-su…

@mr_r0b0t what exists vs what's referenced: Files that exist in the repo: scripts/swarm.py scripts/worker.py templates/prompts.py Files missing: - scripts/seed.py — mentioned in README, SKILL.md - scripts/filter.py — listed in PLAN.md and in the architecture

For all the people who were curious about running 96+ deepseek-v4 powered @NousResearch Hermes agents in parallel, I packaged that skill up and have open sourced it on GitHub. Hope you enjoy DeepSwarm!

@mr_r0b0t @NousResearch Couldn’t find seed filter script

Today we’re releasing Qwen-Scope 🔭, an open suite of sparse autoencoders for the Qwen model family. It turns SAE features into practical tools： 🎯 Inference — Steer model outputs by directly manipulating internal features, no prompt engineering needed 📂 Data — Classify & synthesize targeted data with minimal seed examples, boosting long-tail capabilities 🏋️ Training — Trace code-switching & repetitive generation back to their source, fix them at the root 📊 Evaluation — Analyze feature activation patterns to select smarter benchmarks and cut redundancy We hope the community uses Qwen-Scope to uncover new mechanisms inside Qwen models and build applications beyond what we explored.Excited to see what you build! 🚀 🔗🔗 Blog: qwen.ai/blog?id=qwen-s… HuggingFace: huggingface.co/collections/Qw… ModelScope: modelscope.cn/collections/Qw… Technical Report: …anwen-res.oss-accelerate.aliyuncs.com/qwen-scope/Qwe…

@deepseek_ai 💋

The DeepSeek-V4-Pro discount has been extended until May 31, 2026, 15:59 UTC!

"how do you fit qwen 3.6 27b q4 on 24gb at 262k context" lands in my dms 5 times a week. here is the exact memory math. model bytes at idle = 16gb (q4_k_m of 27b dense) kv cache at 262k context with q4_0 for both k and v = 5gb total = 21gb on the card headroom = 3gb for prompts and tool call traces the magic is the kv cache type. most people leave it at default fp16 or push to q8 thinking quality wins. on qwen 3.6 27b dense at 262k: - fp16 kv cache = does not fit at all - q8 kv cache = fits at 23gb but runs 3x slower (double penalty: more vram, less speed) - q4_0 kv cache = fits at 21gb at full speed (40 tok/s flat curve, same speed at 4k or 262k) most builders never test the kv cache type because tutorials never mention it. it is the single biggest unlock on consumer 24gb hardware. flags i run: ./llama-server -m Qwen3.6-27B-Q4_K_M.gguf -ngl 99 -c 262144 -np 1 -fa on --cache-type-k q4_0 --cache-type-v q4_0 what they do: -ngl 99 = offload everything to gpu -c 262144 = 262k context window -np 1 = single user slot (do not enable multi-slot, eats headroom) -fa on = flash attention on (memory and speed both win) --cache-type-k q4_0 --cache-type-v q4_0 = the unlock if you are sitting on 24gb and not running this config, you are leaving 250k of context on the table. or worse, you are running q8 kv cache and burning 3x your speed for nothing. q4 is not a compromise on consumer hardware. it is the right call.

🔥 Parse your documents with SOTA OCR that beats 👀 Gemini 3 Pro, Qwen3-VL-235B, GLM-OCR using just 1.2B params! 🏆 🔥 1.2B params beats: • Gemini 3 Pro • Qwen3-VL-235B • GLM-OCR • PaddleOCR-VL-1.5 📈 95.69 on OmniDocBench v1.6 (new record) 🚀 65.5M training pages = pure data magic ✨ Tables • Formulas • Charts • Cross-page magic

BREAKING:🚨 NVIDIA just quantized Gemma 4 31B on Hugging Face 🔥 NVFP4 compression = 4x smaller weights with frontier-level accuracy. ✅99.7% of baseline on GPQA (75.46% vs 75.71%). 📈256K context window. 🧐Multimodal (text + images + video). vLLM-ready + Blackwell optimized. VRAM requirements: ⚡️Weights only: ~16–21 GB 🚀Everyday use: Runs on 24 GB GPUs 📈Full 256K context = 32 GB VRAM sweet spot (RTX 5090-class consumer GPUs) This is the 31B-class frontier model you can actually run locally on a high-end rig. Try it today👉 huggingface.co/nvidia/Gemma-4…

@sadkatwt Honey pot 😈

Antigravity is a legit scam use it for an hour and get rate limited for a week even on PRO plan. Also there was a bug they rolled out credit system and just after limits were about to reset there was another week added to the reset timer.

@pankajkumar_dev Unsubscribed

Antigravity, what are you guys doing? - I was waiting for today because the Claude model quota was supposed to reset so I could use it again. - But now its been extended by another 6 days… again. - I haven't been able to use Claude models for the past 2 weeks. At this point, it feels like we can't use them at all. - If this continues, its better to just remove Claude models entirely and now give Pro users an option for a refund on the remaining subscription. Right now, it just feels like a wasted subscription for those who paid specifically for Claude access.

@HarshithLucky3 Not yet… waiting for it 🤔

🚨Update for Google AI Pro users Antigravity just reverted back to the 5 hour reset for all Gemini models including Gemini 3.1 Pro The Claude models are showing a 1 week reset now and that is okay This is much better Google Looks like they listened to the feedback on X today Can anyone confirm?

@CloudflareDev WTF???

Introducing the new /crawl endpoint - one API call and an entire site crawled. No scripts. No browser management. Just the content in HTML, Markdown, or JSON.

@Google When in @antigravity ?

Today, we’re releasing Gemini 3.1 Pro. It’s the same core intelligence that powers Gemini 3 Deep Think, now scaled for your practical applications. It’s a smarter model for your most complex tasks. See 3.1 Pro in action 🧵↓

@antigravity Any repos you recommend?

Easily add skills from repos to your local project. 🛠️ Simply choose your config, pick your skills from the repo, and start using them immediately.

@VittoStack Legend

If you know what this is. We can be friends 🤝