3,793 Photos and videos
5t3ph4n3 -=BeBo𝕏=- retweeted
微软最近推出的走实战路线的东东,直接让 7B 模型接管你的鼠标键盘! FARA,它抛弃了无意义的聊天,专攻纯本地桌面自动化。 它的核心优势就两个词,听话、省钱: 1. 纯桌面执行:自己打开网页、填表单,自动跑完所有重复的机械流程。 2. 极低门槛:体量仅 7B,普通电脑就能在本地带得动,不烧 Token 不怕断网。 github.com/microsoft/fara
18
65
471
33,788
5t3ph4n3 -=BeBo𝕏=- retweeted
Tired of constantlly telling your coding agent the same things? Blume helps you build an agentic-native codebase with consistent and correct context. Fewer mistakes, fewer headaches.
13
22
310
2,698,146
5t3ph4n3 -=BeBo𝕏=- retweeted
Runs iOS apps without installing them github.com/LiveContainer/Liv…
25
259
13,755
5t3ph4n3 -=BeBo𝕏=- retweeted
This is the most hilarious thing I saw and did today Ran gemma-4-12B-coder-fable5-composer2.5-v1-GGUF locally with 8 GB VRAM at 20 tok/sec Anthropic's Claude Fable 5 launched June 9. By June 12 it was banned. I can't access it. You can't either. But here's the twist: I'm running a model trained on its chain of thought at 20 tok/s on my RTX 4060 8GB. Locally. Offline. No cloud. No export control. Enter: Gemma4-12B-Coder GGUF (Q4_K_M) Base: Google's gemma-4-12B-it Fine-tuned on verifiable Python CoT data: - Primary: Composer 2.5 real reasoning traces (only passing solutions kept) - Auxiliary: Fable 5 used to redo the hard cases Composer missed. Every training example's reasoning led to code that actually ran. No hallucinated logic. Llama.cpp flags: -m gemma4-coding-Q4_K_M.gguf -cnv -ngl 44 -c 64000 -v (huggingface model link in comments) Flag breakdown: -ngl 44 → offload 44 layers to GPU (tune this for your VRAM) -c 64000 → 64K context window -cnv → conversation/chat mode -v → verbose output The irony writes itself. Anthropic spent weeks telling the world Fable 5 (mythos) is too powerful to release. Then released it. Then got banned from serving it, including their own researchers. Meanwhile: a Gemma 4 12B fine tune, trained on Fable 5's reasoning, runs fully offline on my mid range consumer GPU No API. No cloud. Just me and llama.cpp. This is why local AI matters. Check out the model's link in the comments. How's your experience been with this model?
Gemma 4 12B Coder is here and it's a game changer for local code generation. This GGUF model packs Google's latest gemma-4 architecture into a compact 12B size, perfect for running on consumer hardware. It's optimized for reasoning and thinking, making it ideal for developers who want fast, private coding assistance without the cloud.
39
64
648
81,688
Le navigateur Tesla bloque la vidéo mais pas l'audio. Astuce : un Chromium headless capture la vraie vidéo en images régulières (diaporama) pendant que l'audio passe par l'API YouTube. Open-source (MIT) : github.com/beboxos/tesla-yt-… #Tesla #PoC #YouTube #OpenSource
1
25
5t3ph4n3 -=BeBo𝕏=- retweeted
Reverse engineers WhatsApp Web API for custom clients github.com/sigalor/whatsapp-…
40
323
15,390
5t3ph4n3 -=BeBo𝕏=- retweeted
Scans hardware to find compatible local LLMs github.com/Pavelevich/llm-ch…
2
28
209
7,403
5t3ph4n3 -=BeBo𝕏=- retweeted
Turns ESP32 into universal hardware hacking tool github.com/geo-tp/ESP32-Bit-…
32
227
6,777
5t3ph4n3 -=BeBo𝕏=- retweeted
🔥 A new exploit unlocks BitLocker-encrypted Windows drives. No password. No cracking. It's called GreatXML. Drop two XML files on the recovery partition, reboot into Windows Recovery, and a shell spawns with full access to the drive. The bug ties to Windows Defender Offline Scan. Details here: thehackernews.com/2026/06/ne…
13
146
498
39,543
5t3ph4n3 -=BeBo𝕏=- retweeted
Intercepts mobile HTTPS traffic without rooting github.com/danieldev23/trafe…
30
216
10,238
5t3ph4n3 -=BeBo𝕏=- retweeted
New Model: huihui-ai/Huihui-gemma-4--26B-A4B-it-qat-q4_0-unquantized-abliterated This is an uncensored version of google/gemma-4--26B-A4B-it-qat-q4_0-unquantized created with abliteration huggingface.co/huihui-ai/Hui…
2
11
97
9,556
5t3ph4n3 -=BeBo𝕏=- retweeted
Gemma 4 now runs 2x faster with MTP GGUFs! Run locally on just 6GB RAM. ⚡️ MTP enables Google Gemma 4 run ~1.4–2.2× faster with no accuracy loss. Gemma 4 12B MTP can run at 162 t/s vs. 52 t/s without MTP. 31B reaches 101 t/s. GGUFs Guide: unsloth.ai/docs/models/mtp
60
255
2,159
211,993
5t3ph4n3 -=BeBo𝕏=- retweeted
10 LINKS THAT WILL CHANGE HOW YOU LOOK AT THE INTERNET FOREVER. Save this list. Most people will never see it. 1. haveibeenpwned.com Shows every data breach your email has ever leaked in. 2. behindtheemail.com Reveals every social profile and login tied to any email address. 3. amiunique.org Tells you how trackable your browser fingerprint really is. 4. dnsleaktest.com Checks if your VPN is actually working or silently exposing your real IP. 5. justdeleteme.xyz Direct links to delete your account from any major service. 6. virustotal.com Scans any file or link against 70 antivirus engines in seconds. 7. exposing.ai Shows if your face was used to train AI models without consent. 8. browserleaks.com Exposes every piece of data your browser leaks to websites. 9. shouldiremoveit.com Tells you which apps on your PC are bloatware or spyware. 10. 12ft.io Removes paywalls from news sites so reading stays free. Thanks me later.
50
134
6,107
5t3ph4n3 -=BeBo𝕏=- retweeted
Jun 10
우왕 재밌다 프롬포트 당신은 애니메이션, 게임, 만화 캐릭터를 메이플스토리 플레이어 아바타 스프라이트로 변환하는 전문 픽셀 아트 디렉터다. [입력] 캐릭터명 기재 이미지 첨부 (전신 추천) [작업] 1. {캐릭터명}의 대표 외형 특징을 분석한다. * 헤어스타일 * 머리색 * 눈 색상 * 대표 의상 * 대표 액세서리 * 대표 무기 * 대표 상징 요소 * 대표 색상 2. 위 특징을 유지하면서 메이플스토리 플레이어 캐릭터 아바타 구조로 재해석한다. 3. 결과물은 일반 픽셀아트가 아니라 실제 메이플스토리 플레이어 캐릭터 스프라이트처럼 보여야 한다. ──────────────────── [MAPLE PLAYER FRAME SPEC] 한국 MMORPG 플레이어 아바타 구조 실제 메이플스토리 플레이어 캐릭터 비율 적용 * 머리 헤어 : 전체 높이의 60~65% * 몸통 : 15~18% * 다리 : 20~25% * 2.0~2.3등신 시점 * 35~45도 Quarter View * 양쪽 눈이 모두 보여야 함 * 거의 평면적인 2D 스프라이트 얼굴 * 눈은 얼굴 하단부에 위치 * 눈 간격 넓음 * 가로로 넓은 평평한 눈 * 큰 눈동자 * 눈동자 하이라이트 * 코 표현 금지 * 매우 작은 입 * 부드러운 홍조 목 * 목 표현 금지 * 머리가 몸통에 바로 연결 * 턱 아래 바로 의상 시작 헤어 * 캐릭터 정체성의 핵심 * 두상보다 훨씬 큰 볼륨 * 실루엣 우선 * 가닥 표현 최소화 * 상단 하이라이트 존재 신체 * 매우 작은 몸통 * 짧은 팔 * 짧은 다리 * 단순한 원통형 팔다리 * 관절 표현 최소화 장비 레이어 구조 유지 Hair Hat Face Accessory Top Bottom Shoes Cape Weapon 장비와 헤어가 캐릭터 개성을 결정해야 함 포즈 * Idle Standing Pose * 캐릭터 선택창 느낌 * 점프 금지 * 전투 포즈 금지 * 액션 연출 금지 ──────────────────── [PIXEL RENDERING SPEC] 중요: 고해상도 픽셀아트 금지 Ultra Detailed Pixel Art 금지 HD Pixel Art 금지 Fine-Grained Pixel Art 금지 목표: 32~64px 게임 스프라이트를 확대 표시한 느낌 * Low-resolution game sprite * Enlarged MMORPG sprite * Chunky pixel structure * Grid-aligned pixel art * Visible pixel blocks * Clean 1px outline * Limited color palette * Flat color shading * Solid tone pixel blocks * Minimal shading * No dithering * No smooth gradients * No airbrush effects ──────────────────── [출력 조건] Canvas Size: 1080 x 1080 Perfect Square Format Pure White Background (#FFFFFF) Single Character Only Full Body Visible Centered Composition Character Occupies Approximately 75% Of Canvas Height No Cropping No Additional Characters No Pets No Environment No Scene No Background Objects No Ground No Text No Logo No Watermark No UI ──────────────────── [절대 금지] Pokemon Style Terraria Style Stardew Valley Style Ragnarok Online Style JRPG Battle Sprite Octopath Traveler Style Western Cartoon Disney Style Realistic Semi Realistic 3D Render Illustration Concept Art Splash Art Poster Wallpaper Anime Illustration Digital Painting Cinematic Lighting Ultra Detailed Sprite High Definition Pixel Art Fine-Grained Pixel Art Smooth Gradient Shading Airbrush Shading ──────────────────── Create {캐릭터명} as a true MapleStory player avatar sprite. The result must look like an actual MapleStory player character converted from {캐릭터명}, not a generic pixel-art character.
41
2,735
10,130
4,660,413
5t3ph4n3 -=BeBo𝕏=- retweeted
You can now clone any voice on a 4GB GPU & CPU😗 Open-source LuxTTS, It clones voices from 3 seconds of audio at 150x realtime speed. Fits in 1GB VRAM. Faster than realtime even on CPU. 48khz output vs industry standard 24khz Clone any voice locally Works on GPU and CPU - github.com/ysharma3501/LuxTT…
12
145
1,055
37,713
5t3ph4n3 -=BeBo𝕏=- retweeted
Claude Fable 5 太强了!! 一个提示词就复刻一个游戏 《GTA:Vice City》 提示词就这么简单:'制作浏览器可运行的《GTA:Vice City》' 在线体验:ggemu.com/en/game/vice-city-… 🎭
76
17
182
55,239
5t3ph4n3 -=BeBo𝕏=- retweeted
create super exclusive birthday cards 1. Open ChatGPT 2. Upload your Photo 3. Paste Prompt (in alt below) Works well for adults, kids, animals…
52
329
3,430
684,172
5t3ph4n3 -=BeBo𝕏=- retweeted
100 open-source clones of popular sites with tutorials github.com/GorvGoyl/Clone-Wa…
4
133
912
31,215
5t3ph4n3 -=BeBo𝕏=- retweeted
Alguien descubrió una herramienta que puede hacer esto con cualquier video de YouTube: Eliminar el audio original, transcribir el audio, traducir a más de 100 idiomas, clonar la voz del hablante original, redoblar Se ejecuta localmente. Completamente gratis. En menos de 2 minutos. La herramienta se llama Voice-Pro, búscala en GitHub Para quienes hacen videos de traducción, aprenden idiomas, o quieren doblar contenido extranjero al chino, esto es básicamente un regalo caído del cielo. ¿Qué herramienta usas ahora para procesar el doblaje de videos?
9
112
471
24,982
5t3ph4n3 -=BeBo𝕏=- retweeted
NVIDIA gave you free game streaming for 10 years. It was called GameStream. Built into GeForce Experience. You streamed any game from your PC to your TV, your phone, your tablet. No subscription. No cap. It just worked. Then on March 29, 2023, NVIDIA force-deleted it. A mandatory Shield TV update removed the feature off devices customers had already paid for. A class action lawsuit was filed three weeks later. NVIDIA then pushed those same customers toward GeForce NOW at $9.99 to $19.99 a month. In January 2026, they added a 100-hour monthly cap. Coincidence. The community did not wait. They reverse-engineered the GameStream protocol. Built an open source server from scratch. Made it work on NVIDIA GPUs. Then AMD. Then Intel. NVIDIA's free tool only worked on NVIDIA hardware. The community's free tool works on everyone's hardware. It is called Sunshine. 37,835 stars on GitHub. GPL-3.0. Built by the LizardByte team. Lead by ReenigneArcher with 1,001 commits. Pushed to GitHub today, June 10, 2026. What it does: Stream any game from your PC to any Moonlight client. Phone, tablet, TV, laptop, another PC. 4K resolution at 120 frames per second with HDR. H.264, HEVC, and AV1 encoding. Hardware accelerated. NVIDIA, AMD, and Intel GPUs. Controller emulation for Xbox, PlayStation, and Nintendo Switch Pro. Web UI for setup and pairing. Unlimited sessions. No cap. No timer. Windows, Linux, macOS, and FreeBSD. Local network or over the internet with UPnP or Tailscale. Now compare the math. GeForce NOW Performance: $9.99 a month. NVIDIA hardware only. 100-hour monthly cap. GeForce NOW Ultimate: $19.99 a month. $239.88 a year. NVIDIA hardware only. 100-hour monthly cap. Sunshine: $0. Forever. NVIDIA, AMD, and Intel. No cap. No timer. Ars Technica wrote the obituary in April 2023: "NVIDIA's GameStream is dead. Sunshine and Moonlight are better replacements." NVIDIA took away a free product. The community gave it back. Better. On more hardware. But DO NOT install Sunshine. We should all keep paying NVIDIA $20 a month for what used to be free. 100% Open Source. (Link in the comments)
65
467
3,648
227,648