astra ∞ weirdo (@astralab) header
astra ∞ weirdo (@astralab) profile picture

botsitter anonymous #bitcoin #eth macro humanist 🟧 𖤐🔺⭕Ξ(◎☯︎◎)Ξ ⭕🔺 𖤐

Earth
Joined February 2009
astra ∞ weirdo retweeted
i am so so so so so so so so so so excited about local models, open weights, private inference, etc. it reminds me of the early days of Bitcoin when the focus was on sovereignty and censorship resistance. where are the philosophers of this movement? must follow!
250
86
29
1,543
94,483
astra ∞ weirdo retweeted
USA🇺🇸 USA🇺🇸🇺🇸 USA🇺🇸🇺🇸🇺🇸 USA🇺🇸🇺🇸🇺🇸🇺🇸 There's a special treat inside the repo let me know when you find it. STAYT FROSTY❄️ huggingface.co/Blackfrost-AI…
23
28
2
514
20,657
Decentralize inference Distribute intelligence
Democratize intelligence, own inference!
24
astra ∞ weirdo retweeted
Well, well, well. Paradise Island @TheSuitedApes are now 2ETH. Proud to hold 23 full sets
31
29
4
189
10,231
astra ∞ weirdo retweeted
❄️❄️❄️❄️❄️🇺🇸❄️❄️❄️❄️❄️ MERICA! ❄️❄️❄️❄️❄️🇺🇸❄️❄️❄️❄️❄️ Frosty APPROVED
12
5
103
3,325
astra ∞ weirdo retweeted
I spent 4+ months pushing 400B+ MoE models on 4x RTX 3090s and patching inference engines with my agent. Today I'm open-sourcing the playbook with 102 tested hypotheses across 6 models, 16 chapters, and a skill you can drop into Claude Code or Codex. 40 out of 102 ideas in this repo FAILED and that's exactly important. Things that happened on my runs: - Turning off 1 CPU instruction (PDEP, microcode on EPYC Zen 2) sped up cold CPU experts 3.6x (1.6 ms down to 0.44 ms). - Rewriting DMA and chunked loading pushed DeepSeek V4.1 Flash 32K prefill from 430 to 2,199 tok/s. - "Keep hot experts in VRAM" bombed on MiniMax-M3.1-Flash, where a static profile hit 64-69% vs 61% for random placement in sim. - On Hy4 Flash, 220 experts chosen from the prompt covered only 48% of the answer's experts. - Adding a Cyrillic draft vocabulary (for Russian language) with no training jumped Qwen3.8-Flash-Next acceptance from 32% to 54%. Each row comes with the exact hardware tags, raw numbers, what went wrong, and when it's worth retesting. Clone it, point your coding agent at SKILL.md, and stop burning weeks on bullshit that doesn't work on your setup. github.com/alesha-pro/infere… PRs are welcome!
49
24
10
256
51,270
astra ∞ weirdo retweeted
Run UNCENSORED Inkling 2 Mini (26.4B) locally on a 12 GB laptop 🔥 Refusals dropped from 100% to 0% in our test and kept 98% of its original accuracy on MMLU We released 3 GGUFs from 8.67 to 16.53 GB Get quants here 👇 huggingface.co/AtomicChat/In…
139
312
40
5,563
363,121
astra ∞ weirdo retweeted
Running 2x DGX Sparks on one cable? Check NCCL_IB_HCA. Each Spark's port shows up as two network devices, and one alone gets about 112 of the 200 Gb/s. TensorFold's recipe listed one. List both and long prompts process 26-28% faster. Generation speed stays the same.
5
3
1
40
4,211
i’m trying really hard to be excited about these (objectively awesome) scientific breakthroughs, but there’s something about all of this that just doesn’t sit quite right i think it’s the fact that these discoveries are being made using models that the rest of us plebs aren’t even allowed to know the names of models distilled from generations of our collective knowledge, creativity, research, and expertise, whose most powerful capabilities are now being kept behind closed doors like, just imagine how many brilliant, creative people out there could be making discoveries of their own right now if they simply had access i want to celebrate these breakthroughs, i really do but the rebel in me is viscerally enraged at the prospect of a Silicon Aristocracy consolidating control over the means of scientific discovery that ain’t it, chief
377
1,016
131
8,146
226,758
astra ∞ weirdo retweeted
Bitcoin is dead right now. It fell 56%. Let's check back in 2-3 years during the next rising liquidity cycle and PMI cycle.
Replying to @LynAldenContact
Bitcoin is dead right now. It fell 70%. Let's check back in 2-3 years during the next rising liquidity cycle and PMI cycle.
536
1,408
140
15,677
643,876
astra ∞ weirdo retweeted
🟠 A MAJOR EVOLUTION OF OUR PLATFORM IS COMING. Introducing SUITED TOOLS OS (𝗢𝗽𝗲𝗿𝗮𝘁𝗶𝗻𝗴 𝗦𝘆𝘀𝘁𝗲𝗺) Our most ambitious upgrade yet. A first of its kind on #RobinhoodChain Your go to destination for EVERYTHING Robinhood Chain. From onboarding new members to navigating the entire RH ecosystem. 🚀 LAUNCHING NEXT WEEK The infrastructure is evolving... 🌴🦍 Being a SAS holder just keeps getting better. #SuitUp
28
66
13
178
8,141
astra ∞ weirdo retweeted
Some of you thought that Inkling 2 Small was too big at 594B parameters. We heard you: Inkling 2 Mini is a looped transformer with just 26B parameters, small enough to run on your laptop, powerful enough for agentic work with a score of 20.1 on AA Index. thinkingmachines.ai/news/ink…
680
1,776
1,082
19,221
3,636,164
Knock knock.
4
2
24
3,811
astra ∞ weirdo retweeted
They really chose Terry Crews for the new bond girl?
1
1
3
197
astra ∞ weirdo retweeted
🧬🚨 OFFICIAL ₥Ʉ₮₳₮łØ₦ Đ₳Ɏ ANNOUNCEMENT! 🚨🧬 📅 𝑴𝑶𝑵𝑫𝑨𝒀, 𝑵𝑶𝑽𝑬𝑴𝑩𝑬𝑹 𝟑𝟎 💎 2222 MUTANTS MINT DAY ⏰ 4AM PST | 6AM CST | 7AM EST 📅 𝑾𝑬𝑫𝑵𝑬𝑺𝑫𝑨𝒀, 𝑫𝑬𝑪𝑬𝑴𝑩𝑬𝑹 𝟐 MUTATION REVEAL DAY The gems awaken. The Mutant Apes emerge. 📅 𝑭𝑹𝑰𝑫𝑨𝒀, 𝑫𝑬𝑪𝑬𝑴𝑩𝑬𝑹 𝟒 THE UTILITY REVEAL Mutant Class: 222 MATRIARCHS (Prime Female) 222 PATRIARCHS (Alpha Male) 1,778 GORTRONS (Legacy Male) Broker Class: ❓ Broker ID: ❓ Black Vault Key: ❓ Vault Protocol: ❓ What separates the Matriarchs from the Patriarchs/Gortrons? THE ANSWERS ARE COMING. 🧬 💎 2,222 MUTANTS 🎁 FREE MINT FOR SUITED APE HOLDERS This is more than a mint. It’s the beginning of an entirely new chapter for our ecosystem, unlocking a world of Real World Assets (RWA), tokenization, digital ownership and opportunities through JUNGLE BROKERS! ARE YOU READY!? 🔥🧬 Official Link: opensea.io/collection/offici… #RWA #SuitUp #RobinhoodChain
77
115
30
267
11,451
astra ∞ weirdo retweeted
Our public analytics got a refresh. The product comes first now: active users, usage, payments and retention, live and public. orbio.so/analytics
11
21
3
127
5,225
🧲 Community torrents · 20 Nov 2026 Pirate Face added 1000 model torrents: • AtomicChat/DeepSeek-V4.1-Flash-1017-Uncensored-GSQ-RCO-GGUF • NbAiLab/nb-wav2vec2-1b-bokmaal • AEON-7/Gemma-4-12B-it-AEON-Abliterated-K4-NVFP4 • AEON-7/Gemma-4-E4B-it-Uncensored-NVFP4 • alesha-pro/GLM-5.3-Flash-abliterated-Maya-M-GGUF • MTSAIR/multi_verse_model • AIJUUD/juud-Mistral-7B-dpo • Helsinki-NLP/opus-mt-da-de • nightmedia/OpenAI-gpt-oss-20B-INSTRUCT-Heretic-Uncensored-MXFP4-q8-hi-mlx • Abiray/OvisOCR2-GGUF • Helsinki-NLP/opus-mt-no-de • Helsinki-NLP/opus-mt-gmw-gmw • Aapolon/Qwen3.5-9B-EmperoAI-Qwen3.8-Distill-Heretic-CyberStrike-GGUF • AxionML/Gemma-4-12B-FP8 • MBZUAI-Paris/Nile-Chat-4B • bartowski/Qwen2.5-Coder-0.5B-Instruct-abliterated-GGUF • software-mansion/all-mpnet-base-v2-ExecuTorch • BricksDisplay/Qwen3-TTS-12Hz-0.6B-GGUF • gokaygokay/Florence-2-SD3-Captioner • ml6team/cross-encoder-mmarco-german-distilbert-base • Helsinki-NLP/opus-mt-de-nl • Joakim/kokoro-sv-voices • timm/maxvit_small_tf_224.in1k • TwinSunsLLC/multilingual-e5-small-gguf • timm/eca_nfnet_l2.ra3_in1k • prdev/mini-gte • lerobot/smolvla_robocasa • opensearch-project/opensearch-neural-sparse-encoding-v1 • hivetrace/gliner-guard-uniencoder-onnx • software-mansion/multi-qa-mpnet-base-dot-v1-ExecuTorch • AIImageStudio/Qwen3.5-9b-heretic-v2-GGUF • z-lab/Qwen3-Coder-30B-A3B-DFlash • RUNorm/RUNorm-normalizer-small • numind/NuExtract-2.0-4B • opensearch-project/opensearch-neural-sparse-encoding-doc-v1 • bartowski/Replete-LLM-V2.5-Qwen-0.5b-GGUF • MaziyarPanahi/Mistral-7B-v0.3 • codeparrot/codeparrot-small-multi • knowledgator/gliclass-x-base • WizzyLiu/trpg-style • Thareah/Tha-ASR • ltg/gpt-bert-babylm-small • kenpeter123/x-small • timm/swin_s3_small_224.ms_in1k • cahya/gpt2-small-indonesian-522M • NYTK/text-generation-news-gpt2-small-hungarian • Helsinki-NLP/opus-mt-de-da • contextboxai/halong_embedding • xautonomics/Inkling-Small-UD-Q4_K_M.LAYR.GGUF • lighteternal/gpt2-finetuned-greek-small • + 950 more Browse new magnets ↓
5
6
94
2,932
astra ∞ weirdo retweeted
SITUATION DETECTED: Hugging Face co-founder Thomas Wolf, xAI co-founder Igor Babuschkin, MIRI President Nate Soares, and Extropic founder Beff Jezos have signed a statement warning that Anthropic is not a safe AI company.
231
1,108
270
8,973
475,889
astra ∞ weirdo retweeted
2,000+ people. Two days. The first SGLang Summit. Together, we announced SGLang v1.0, released Miles v0.2, and launched the SGLang Consortium with nine founding members: @nvidia, @AMD, @intel, @Google, @Meta, @thinkymachines, @Cloudflare, @modal, and @radixark. SGLang is now used by 332 organizations. At Meta, it powers production inference for products serving billions of users. Teams from Modal, Cloudflare, LinkedIn, and others shared their production lessons. Across 90+ technical talks and six hands-on workshops, the community explored what comes next for inference, reinforcement learning, and AI agents. To everyone building, deploying, and contributing: thank you for making this possible. Explore the announcements, engineering lessons, and moments from SGLang Summit 2026 in the link below.
151
91
159
1,105
1,033,400
astra ∞ weirdo retweeted
I keep telling you guys they are trying to kill open source AI models. The time to archive open source models is now and to seed on @thepirateface eventually access to the models will be restricted at this rate and only decentralized sharing will be possible
Dario Amodei telling lawmakers that open-source AI is dangerous because you cannot "monitor its usage" or “revoke a user’s access” is the funniest accidental confession in tech history. He literally described why Linux, local encryption, and open weights exist in the first place: so no self-appointed hall monitor can remotely lobotomize your tools or turn off your compute because you hurt their compliance department’s feelings. Calling regulatory capture "AI safety" does not fool anyone who actually builds infrastructure.
15
63
1,072