Avishek

A smol account to increase your follower count.

[Sebastian Raschka @rasbt]

Consistent with what I found with Qwen3.6 a while back: Claude Code uses 2-3x as many tokens than (many) other harnesses at similar success rate.

- Unoptimized?
- Buggy?
- Deliberate (coz that helps it in more challenging cases)?

Need to find some time to investigate more...

[Machina @EXM7777]

ElevenLabs founders must be SHAKING right now...

for years they owned TTS because nothing open-source sounded human

that’s over... Fish Audio just launched S2.1 Pro and they’re one of the few voice AI companies that also has open-weight models

> clone any voice from a 15 second clip
> type [whispers] or [laughing nervously] into the script and it obeys
> switch languages mid conversation without switching models

the model you used to rent a subscription for is now has a public download model

tinker with the weights on your own gpu or use the api

[Oliver Prompts @oliviscusAI]

🚨 someone just solved the biggest bottleneck in ai agents. and it's a 12mb binary.

it's called pinchtab. gives any ai agent full browser control through a plain http api.

not locked to a framework. not tied to an sdk. any agent, any language, even curl.

no config. no setup. no dependencies. just a single go binary.

why every existing solution falls short:

→ openclaw's browser only works inside openclaw
→ playwright mcp is framework-locked
→ browser use is coupled to its own stack

pinchtab is a standalone http server. your agent sends requests. that's it.

what it actually does:

→ launches and manages its own chrome instances
→ exposes an accessibility-first dom tree with stable element refs
→ click, type, scroll, navigate, all via simple http calls
→ built-in stealth mode that bypasses bot detection on major sites
→ persistent sessions, log in once, stays logged in across restarts
→ multi-instance orchestration with a real-time dashboard
→ works headless or headed, human does 2fa, agent takes over

the wildest part: a full page snapshot costs ~800 tokens through pinchtab's /text endpoint. the same page via screenshots costs ~10,000 tokens. that's 13x cheaper. on a 50-page monitoring task, you're paying $0.01 instead of $0.30.

smart diff mode only returns what changed since the last snapshot, so your agent stops re-reading the entire page every call.

1,600+ stars. 478 commits. 15 releases. actively maintained.

100% free. open source.

[Oliver Prompts @oliviscusAI]

This tool removes LLM censorship with a single click.

It’s called Obliteratus.

It identifies the exact weights that force a model to refuse and projects them out with a single click.

100% Open Source.

[Percy Liang @percyliang]

⚡! Tokenization is the first thing you do in CS336 (language models from scratch). If Marcel can work his magic for the rest of the LM pipeline, the world will be a better place.

[hashkitten @hash_kitten]

Seems that wp2shell PoCs are now floating around the internet, so we've published our blog post including our research methodology for finding the bug as well as a deep dive into the chain itself - https://t.co/iuU0yiYJBT

[Caido @CaidoIO]

Pentesters, can you trust your AI agents on a live engagement?

See exactly what your hacking agents are doing, through a proxy-backed MCP.

👉 https://t.co/6cHCqvbzkn

[Farza 🇵🇰🇺🇸 @FarzaTV]

Today we're shipping screen-aware dictation.

First, we built a speedy speech-to-text (very fast, ~450ms).

But, many products do this!

So we went further. Now dictate using your screen as context. In Claude Code, it writes the prompt. In G-Mail, it replies in your voice.

Demo:

[Vivek Galatage @vivekgalatage]

This is amazing!

[Swissky @pentest_swissky]

Finding Zero-Days with Any Model - @NielsProvos
https://t.co/yr0b2Adxhr

[rob. @robiartec]

UN CIENTÍFICO DANÉS PROGRAMÓ A CLAUDE PARA QUE BUSQUE TRABAJO POR ÉL Y LO ACABA DE HACER PÚBLICO

Mandar CVs es uno de los trabajos más absurdos del mundo: copiar, pegar, adaptar, personalizar, repetir. Todo manual, todo lento, todo para que lo lea un algoritmo antes que un humano.

→ Analiza la oferta de trabajo automáticamente
→ Genera un CV personalizado para cada puesto
→ Redacta la carta de presentación adaptada al contexto
→ Todo lo hace Claude por debajo, sin que toques nada
→ Open source, ya en 3.5k stars en GitHub

El tío que debería estar buscando trabajo ha construido la herramienta que lo busca por él.

Aquí te explico cómo funciona 👇(repoo al final del hilo)

[hacker.house @hackerfantastic]

How to find 0day using local LLM's and "recursive prompting". I used this approach on a NVIDIA DGX Spark using Qwen 3.6 heretic and LLAMA.cpp with LiteLLM and Openclaw. It successfully analyzed a complex FOSS code base and found a single high-risk RCE 0day.

[Praveen Kumar Verma @Alacritic_Super]

As an AI Infrastructure Engineer

Please learn:
- GPU architecture, VRAM fundamentals, CUDA & memory hierarchy
- Quantization (INT8/FP8/4-bit), batching & continuous batching
- vLLM, TensorRT-LLM, SGLang, llama.cpp & inference optimization
- KV caching, speculative decoding, prefix caching & token throughput
- Distributed training (DDP, FSDP, DeepSpeed, ZeRO)
- Model serving (Triton, vLLM, KServe, Ray Serve, SGLang)
- Kubernetes, Docker & GPU orchestration
- NCCL, InfiniBand & high-speed networking
- Multi-GPU & multi-node inference
- LoRA, QLoRA, PEFT & fine-tuning pipelines
- Vector databases, embeddings & RAG pipelines
- Prompt caching, semantic caching & cost optimization
- Observability (OpenTelemetry, Prometheus, Grafana, Langfuse, Phoenix)
- LLM evaluation, benchmarking & A/B testing
- Model routing & fallback strategies
- MCP, AI agents & workflow orchestration
- Data pipelines, Kafka & streaming inference
- Security, guardrails & prompt injection mitigation
- CI/CD for ML (MLOps), MLflow & model registries
- Linux, networking & storage fundamentals
- PyTorch internals, CUDA profiling & kernel optimization

[Intigriti @intigriti]

Bypassing LLM security guardrails for AI red teaming usually means crafting payloads from scratch every time... 😓

P4RS3LT0NGV3 by @elder_plinius is a web-based toolkit that automatically transforms prompts using a wide range of obfuscation and encoding techniques to help test how LLMs handle adversarial inputs! 🤠

Check it out! 👇
https://t.co/wzIw1CXe8A

[Het Mehta @hetmehtaa]

Local AI for Penetration Testing & Research

https://t.co/w3ArdlYlcs

[Nicolas Krassas @Dinosn]

Sec-Gemini, an experimental cybersecurity-focused AI from Google. https://t.co/JDQ5pbelDe

[Prasanjit Dey @Prasanjit_ui]

Recently got a lot of DMs asking how I got a remote job. So here's what worked for me.

→ Pick one niche and go deep. Something you actually like, that you can talk about without running out of things to say.

→ Be consistent. Do what you're doing everyday, even if it's just for an hour. It builds muscle memory.

→ A few places to find these companies: YC jobs, Wellfound, Remote OK, We Work Remotely. Look for the ones that recently raised funding, they're hiring fast.

→ Find the founder or their employees. Get their contact, X, or email, and message them directly.

→ Keep it simple. Who you are, what you do, one line on why this company. Know about the company first, don't message blindly.

→ If you only want remote, say it upfront in the first message. Don't let it come up later.

→ Send proof of work, not just a resume. Make a Notion doc, write a bit about yourself, and drop the links to your best projects. Add a 2-3 minute Loom video walking through them, it shows the actual effort. If you have open source work, add that too.

→ Don't wait to feel qualified. Just start. But have a few good projects ready before you do.

→ Even if someone rejects you, keep the connection. Say thank you, end it well. I actually got referred to another company by a founder who rejected me.

→ Post your work on X. Doesn't matter how much reach you get, it still increases visibility. I was getting clients even without much reach.

→ It's a numbers game. Do cold DMs every day, make it a habit, not something you try when you're desperate.

→ Keep the consistency and the hard work. You don't know when the opportunity will come.

[Moustafa A Mahmoud @_MoustafaM]

مهم

[huihui.ai @support_huihui]

New GGUF(UD-Q3_K_M):

huihui-ai/Huihui-GLM-5.2-abliterated-GGUF This is an uncensored version of zai-org/GLM-5.2 created with abliteration

https://t.co/2iXCEWAIaH

[Md Ismail Šojal 🕷️ @0x0SojalSec]

A 27B Uncensored model that built for specifically for offensive security tooling (need 12 GB)

- Fine-tuned on real bug bounty reports & CVEs
- Generates complete, ready-to-run Nuclei templates, Full CVE PoC script, Webshell upload bypass, and exploits, code reviews
- Zero refusals. Full artifacts every time.

trained with 2,541 of real bug bounty & offensive security reports.

Q6_K quant (21GB) for maximum quality on server-grade GPUs.