ContentPosts from @kala..
Link
@kala shared a link, 4 months, 2 weeks ago
FAUN.dev()

Cato CTRL™ Threat Research: HashJack - Novel Indirect Prompt Injection Against AI Browser Assistants

A new attack method -HashJack- shows how AI browsers can be tricked with nothing more than a URL fragment. It works like this: drop malicious instructions after the#in a link, and AI copilots likeComet,Copilot for Edge, andGemini for Chromemight swallow them whole. No need to hack the site. The LLM .. read more  

Link
@kala shared a link, 4 months, 2 weeks ago
FAUN.dev()

AI and QE: Patterns and Anti-Patterns

The author shared insights on how AI can be leveraged as a QE and highlighted potential dangers to watch out for, drawing parallels with misuse of positive behaviors or characteristics taken out of context. The post outlined anti-patterns related to automating tasks, stimulating thinking, and tailor.. read more  

News FAUN.dev() Team
@kala shared an update, 4 months, 2 weeks ago
FAUN.dev()

Guido van Rossum: “AI Should Adapt to Python - Not the Other Way Around”

Python TypeScript

Guido van Rossum discussed Python's enduring relevance in AI and education at GitHub's Octoverse, emphasizing its clarity, accessibility, and community-driven growth despite TypeScript's rise.

Guido van Rossum: “AI Should Adapt to Python - Not the Other Way Around”
News FAUN.dev() Team
@kala shared an update, 4 months, 2 weeks ago
FAUN.dev()

DeepSeekMath-V2 Launches with 685B Parameters - Dominates Math Contests

DeepSeekMath-V2

DeepSeekMath-V2, an AI model with 685 billion parameters, excels in mathematical reasoning and achieves top scores in major competitions, now available open source for research and commercial use.

DeepSeekMath-V2 Launches with 685B Parameters - Dominates Math Contests
 Activity
@kala added a new tool DeepSeekMath-V2 , 4 months, 2 weeks ago.
News FAUN.dev() Team
@kala shared an update, 4 months, 2 weeks ago
FAUN.dev()

A New Challenger: INTELLECT-3's 100B Parameters Punch Above Their Weight

Ansible Lustre Slurm INTELLECT-3

INTELLECT-3, a 100B+ parameter model, sets new benchmarks in AI, with open-sourced training components to foster research in reinforcement learning.

A New Challenger: INTELLECT-3's 100B Parameters Punch Above Their Weight
 Activity
@kala added a new tool INTELLECT-3 , 4 months, 2 weeks ago.
Link
@kala shared a link, 4 months, 3 weeks ago
FAUN.dev()

20x Faster TRL Fine-tuning with RapidFire AI

RapidFire AI just dropped a scheduling engine built for chaos - and control. It shards datasets on the fly, reallocates as needed, and runs multipleTRL fine-tuning configs at once, even on a single GPU. No magic, just clever orchestration. It plugs into TRL withdrop-in wrappers, spreads training acr.. read more  

20x Faster TRL Fine-tuning with RapidFire AI
Link
@kala shared a link, 4 months, 3 weeks ago
FAUN.dev()

Code execution with MCP: building more efficient AI agents

Code is taking over MCP workflows - and fast. With theModel Context Protocol, agents don’t just call tools. They load them on demand. Filter data. Track state like any decent program would. That shift slashes context bloat - up to 98% fewer tokens. It also trims latency and scales cleaner across tho.. read more  

Code execution with MCP: building more efficient AI agents
Link
@kala shared a link, 4 months, 3 weeks ago
FAUN.dev()

Hacking Gemini: A Multi-Layered Approach

A researcher found a multi-layer sanitization gap inGoogle Gemini. It let attackers pull off indirect prompt injections to leak Workspace data - think Gmail, Drive, Calendar - using Markdown image renders across Gemini andColab export chains. The trick? Sneaking through cracks between HTML and Markd.. read more Â