Introducing TurboQuant: Our new compression algorithm that reduces LLM key-value cache memory by at least 6x and delivers up to 8x speedup, all with zero accuracy loss, redefining AI efficiency. Read the blog to learn how it achieves these results: https://t.co/CDSQ8HpZoc
Rikugan - A reverse-engineering agent for IDA Pro and Binary Ninja - https://t.co/7AMhX1evmN
A reverse-engineering agent for IDA Pro and Binary Ninja that integrates a multi-provider LLM directly into your analysis UI. This project was vibecoded together with my friend, Claude Code.
You can now run 70B LLMs on a 4GB GPU.
AirLLM just made massive models usable on low-memory hardware.
𝗪𝗵𝗮𝘁 𝗷𝘂𝘀𝘁 𝗵𝗮𝗽𝗽𝗲𝗻𝗲𝗱
AirLLM released memory-optimized inference for large language models.
It runs 70B models on 4GB VRAM.
It can even run 405B Llama 3.1 on 8GB VRAM.
𝗛𝗼𝘄 𝗶𝘁 𝘄𝗼𝗿𝗸𝘀
AirLLM loads models one layer at a time.
Instead of loading everything:
→ Load a layer
→ Run computation
→ Free memory
→ Load the next layer
This keeps GPU memory usage extremely low.
𝗞𝗲𝘆 𝗱𝗲𝘁𝗮𝗶𝗹𝘀
• No quantization required by default
• Optional 4-bit or 8-bit weight compression
• Same API as Hugging Face Transformers
• Supports CPU and GPU inference
• Works on Linux and macOS Apple Silicon
𝗪𝗵𝗮𝘁 𝘆𝗼𝘂 𝗰𝗮𝗻 𝗱𝗼
• Run Llama, Qwen, Mistral, Mixtral locally
• Test large models without cloud GPUs
• Prototype agents on cheap hardware
Open-source alternative to Little Snitch for Windows and Linux that monitors all network activity with per app firewall controls, tracker blocking, secure DNS, and optional SPN https://t.co/SbHWzEvVBN by @SafingIO
We built a full-system iOS fuzzer using QEMU+AFL, dup2() I/O channels, hypercalls, syscall enumeration & __syscall tricks on undocumented architecture. Bridged gap between fuzzing theory & closed-source systems.
Instructions and code
https://t.co/elx6EA825s
Course and book
https://t.co/mPwIymEvmz
https://t.co/xwY1vsWwwo
The course has an exam and a certification.
#iOSFuzzing
#SecurityResearch #Fuzzing
Master fuzzing with AFL++/QEMU to discover vulnerabilities in large codebases. Learn full-system fuzzing, exploit edge cases in parsers & drivers, and hunt CVEs systematically. AFL + QEMU = CVEs 🔍
Below an example of @HexHiveEPFL Native Android System Service (NASS) Fuzzing
fuzzing installd paths