Max R

libertarian

[Het Mehta @hetmehtaa]

Found a very useful dataset for anyone building security agents.

• 2M+ sanitized SOC events
• 51K+ incident provenance graphs
• natural-language attack reports
• ATT&CK context

Much better benchmark than asking your agent to explain Mimikatz.

https://t.co/24v3kPTjtr

[Md Ismail Šojal 🕷️ @0x0SojalSec]

Uncensored Cyber-Agent Qwen3.6-35B-A3B, Opus reasoning distill, specialized for Pentest, can run locally.

- bRATH-format reports,
- tool-call style recon, and ATT&CK mapping
- pentest output: CVSS, CWE, MITRE ATT&CK

- https://t.co/7g7esSS7y8

[Nix0n @UnTalNixon_exe]

El baneo en Hugging Face que encendió la polémica por fin tiene un plan de continuidad a prueba de censura.

El modelo `audnai/penc1aw-GLM-5.3-abliterated-for-offensive-cyber` (diseñado específicamente para operaciones ofensivas y red teaming) ha sido listado oficialmente en Pirate Face porque ningún modelo abierto debería perderse por caprichos de moderación.

Lo que hay detrás de este movimiento:

• Resistencia de infraestructura: un plan de continuidad y refugio descentralizado para modelos de IA desinhibidos que son retirados de las plataformas mainstream.

• Potencial ofensivo puro: modelos "abliterated" optimizados para razonamiento técnico profundo, pentesting y flujos de ciberseguridad sin restricciones corporativas.

• La descentralización del código: la comunidad exige alternativas donde la censura no frene la investigación de vanguardia.

El ecosistema de modelos abiertos se mueve a plataformas alternativas para garantizar que la innovación no se detenga.

Te dejo el perfil completo del modelo y el repositorio directo por acá abajo para que lo analices a fondo 🛡️👇

[𝕏 Bug Bounty Writeups 𝕏 @bountywriteups]

👾 PenHunter - Modular Web Vulnerability Scanner

PenHunter is a powerful, modular web vulnerability scanner written in Go. It provides comprehensive testing for common web vulnerabilities including XSS, SQL Injection, LFI, SSRF, RCE, Open Redirect, and CSRF — with concurrent scanning, advanced detection methods, WAF evasion, and integrations with industry tools like dalfox, sqlmap, and nuclei.

https://t.co/LeTejfDOKU

By @fccdll

[Лентач @oldLentach]

Ровно 35 лет назад на поле аэродрома Тушино в Москве состоялся рок-фестиваль «Монстры Рока».

Вскоре после путча ГКЧП на сцену вышли AC/DC, Metallica, Pantera, Black Crowes и другие группы.

На фестивале побывало от 700 тысяч до 1,6 миллиона человек. Посмотрите, как это было.

[Max R @max_fwr]

Вы помните, когда зарегистрировались в X? Я — да. #МояГодовщинаВX

[Blackfrost_AI @Blackfrost_AI]

IT'S FINALLY FROZEN
🥶🥶CYBER-FROST-3.8❄️❄️

Built on Qwen 3. Flash model and fine-tuned with real-world engagement data and teachers from frontier models who are also derisked.

OFFICIAL EXL3 BY @ViC305 (Monday Sep 28, 2006)
https://t.co/VeJAFU4IZY

[Dhanian 🗯️ @e_opore]

10 Network Security Concepts to Master for a Cybersecurity Interview

1. Firewalls & Network Filtering (Packet Filtering, Stateful Inspection, NGFW)
2. IDS & IPS (Intrusion Detection, Intrusion Prevention, Signature & Anomaly Detection)
3. Network Segmentation (VLANs, Subnets, Microsegmentation)
4. VPN & Secure Remote Access (IPsec, SSL/TLS VPNs, Tunneling)
5. Network Protocol Security (TCP/IP, DNS, HTTP/HTTPS, SSH, TLS)
6. Zero Trust Network Security (Identity-Based Access, Least Privilege, Continuous Verification)
7. Network Access Control (NAC) (Device Authentication, Posture Checks, Access Policies)
8. DDoS Protection (Traffic Filtering, Rate Limiting, Load Balancing, Mitigation)
9. Network Monitoring & Traffic Analysis (Packet Capture, NetFlow, SIEM, Anomaly Detection)
10. Secure Network Architecture (DMZ, Proxies, Bastion Hosts, Defense in Depth)

📘 Grab the Cybersecurity Interview Handbook:
👉 https://t.co/Bgk5Ywp1UC

[nisten @nisten]

Opus 5.5 generated dataset covering synthetic clinical conversations and RAG-cards for every single human disease.
Apache 2.0
https://t.co/mH5M7M4CSm

[Md Ismail Šojal 🕷️ @0x0SojalSec]

New Uncensored 27B cyber model that actually fits on for Red teamer locally.

- OrcaSAQ-2 Cyber 27B Uncensored GGUF
- Zero refusal for payloads generation
- Made for defensive red teaming, vulnerability research, security coding,
- and authorized testing without sending source, logs, or findings to someone else’s API.
- terminal workflows
- 94.4% Top-1 agreement and 262K context.
- DFlash2 speculative decoding
this is the smallest serious cyber model in this class.

[Fred-New @Fred834567]

兄弟们！我发现了一个可以搞黄色的视频模型！

⛓️：https://t.co/trSynbTJK4

不是擦边图转视频，也不是穿个比基尼让你自己脑补，它就是冲着成人向内容微调的。

这个可以根据文字提示生成成人视频内容，该模型基于一个包含大约 1,250 个不同的 NSFW 子版块中排名前 1,000 帖子的数据集训练。

说白了，别人拿数据喂模型学猫学狗，这哥们直接把 Reddit 老色批们多年积累的“学术成果”端上桌了。

不过真准备开车，先别急着下 e10。

仓库作者后面已经把 e4–e20 归到旧版本，现在更推荐从 wan_1.3B_exp_e14.safetensors 开始，权重大概 2.8GB。

真跑的时候也别一上来就满脑子“狠狠干大制作”。先测三样：

✅️脸别换人
✅️手脚别长歪
✅️动作别像触电

这三��过不了，尺度越大越灾难。

本来想看点成年人该看的东西，最后给你生成一个三只手、两条腿打结的赛博妖怪，瞬间进入贤者时间。

这种模型我其实更建议本地跑，私人素材别往来路不明的云端扔，真想玩就玩虚构成年人。

先用普通场景测 人物一致性、动作连贯、镜头稳定，这三个不过关，后面提示词写再骚都白搭。

[OrcaRouter 🐳 @OrcaRouter]

Still cooking on Saturday.

We compressed our most popular local cyber model down to 15.7 GB.

Meet OrcaSAQ-2 Cyber 27B Uncensored GGUF — built for defensive red teaming, vulnerability research, security coding, terminal workflows, and authorized security testing.

54.7 → 15.7 GB
262K context
94.4% Top-1 agreement

Cyber capability should not require sending your source code, logs, or vulnerabilities to someone else’s cloud. Run it locally. Keep the data local.

Our smallest cyber model yet and one of the most capable we’ve evaluated in this size class. Have fun!
https://t.co/QhItgfVVCy

[Max R @max_fwr]

@odzhk В пятерке так же, еще появились пачки по 500г сахара по цене чуть ниже 3кг

[Hugging Models @HuggingModels]

Meet Qwen3.6-35B-A3B Uncensored Genesis Hermes Final GGUF: a multimodal beast that sees images, reads text, and talks back without filters. 361k downloads and 643 likes say it all. This is the uncensored AI you've been waiting for.

[小互 @xiaohu]

兄弟们 好东西

Qwen-Image-2.1无审查版本

移除了安全过滤
可以生成成人、NSFW等图像内容

而且优化了生成速度：

主模型放进显卡里跑保证速度，把巨大的“文本编码器”扔给普通电脑内存（CPU）去跑。 这样能帮你省出 9 到 17 GB 的显存，而且几乎不会拖慢出图速度！

https://t.co/xB8A5M6L8U

Qwen-Image-2.1同时把“文生图”和“图像编辑”融合成了一个模型，还原生支持透明图处理。

最多支持 10 张参考图：

可以把 6 个人的单人照直接合成为一张自然的多人合照

画面质感提升（字雅人美）：

文字排版不再“画��”：解决了以往生图模型文字容易扭曲、拼错的毛病，字体、排版更协调美观。

光影细节更逼真：皮肤质感、毛发细节与自然光影大幅改善，摆脱“塑料 AI 感”

[lucena @delucenaoficial]

Luxo na vida é não ter ninguém danificando o seu sistema nervoso.

[؜ @xxByte]

Active Directory Attack Architecture Map

https://t.co/CEolRHsi1z

[Hugging Models @HuggingModels]

Meet Qwen3.8-Flash-Next-Uncensored, a GGUF model tuned for ROCm FP4 on AMD Strix Halo. Built with imatrix quantization and llama.cpp, it brings fast, unfiltered text generation to Ryzen AI Max hardware. A niche gem for local AI fans.

[Max R @max_fwr]

@odzhk Фулл будет ?