Christian Schnell
@christianschnell.bsky.social
Supports your Digital Transformation Journey, CIO Advisory, Program and Project Management, Zurich 🇨🇭
Reposted by Christian Schnell
Furthermore, for each model size, they released the base, instruct, and reasoning variants to the community, each with image understanding capabilities, all under the Apache 2.0 license.
mistral.ai/news/mistral-3
mistral.ai/news/mistral-3
Introducing Mistral 3 | Mistral AI
A family of frontier open-source multimodal models
mistral.ai
December 2, 2025 at 7:45 PM
Furthermore, for each model size, they released the base, instruct, and reasoning variants to the community, each with image understanding capabilities, all under the Apache 2.0 license.
mistral.ai/news/mistral-3
mistral.ai/news/mistral-3
Reposted by Christian Schnell
ArtificialAnalysis.ai puts Kimi K2 nearly tied with GPT-5, < 1/3 the cost.
November 8, 2025 at 1:30 PM
ArtificialAnalysis.ai puts Kimi K2 nearly tied with GPT-5, < 1/3 the cost.
Reposted by Christian Schnell
Reposted by Christian Schnell
"Trust no AI."
OpenAI's New AI Browser Is Already Falling Victim to Prompt Injection Attacks
Experts confirmed almost immediately that OpenAI's latest AI browser, dubbed Atlas, is "definitely vulnerable to prompt injection."
trib.al
October 26, 2025 at 4:46 AM
"Trust no AI."
Reposted by Christian Schnell
BBC study finds AI chatbots still get news wrong 45% of the time #Technology #Cybersecurity Other
BBC study finds AI chatbots still get news wrong 45% of the time
Analysis from the BBC and other European news outlets has found that around 45 percent of AI chatbot responses based on news articles contain errors. The findings have potentially severe implications...
puretech.news
October 26, 2025 at 5:30 AM
BBC study finds AI chatbots still get news wrong 45% of the time #Technology #Cybersecurity Other
Reposted by Christian Schnell
First impressions of ChatGPT Atlas, as browser agents remain confusing, with insurmountable security and privacy risks including prompt injection attacks (Simon Willison/Simon Willison's Weblog)
Main Link | Techmeme Permalink
Main Link | Techmeme Permalink
October 21, 2025 at 7:11 PM
First impressions of ChatGPT Atlas, as browser agents remain confusing, with insurmountable security and privacy risks including prompt injection attacks (Simon Willison/Simon Willison's Weblog)
Main Link | Techmeme Permalink
Main Link | Techmeme Permalink
Reposted by Christian Schnell
Discord support breach may have exposed photo IDs for millions of users
Discord support breach may have exposed photo IDs for millions of users
Late last week, Discord reported on a breach that happened recently with its third-party customer support partner, Zendesk. But new information indicates the stolen data could be worse than originally thought, including millions of government photo IDs.
more…
9to5mac.com
October 8, 2025 at 7:10 PM
Discord support breach may have exposed photo IDs for millions of users
Reposted by Christian Schnell
Doesn’t sound ideal.
"We cannot control who sees what, we cannot see what users are doing, and we cannot verify that the software itself is secure," the memo says.
www.reuters.com/business/aer...
"We cannot control who sees what, we cannot see what users are doing, and we cannot verify that the software itself is secure," the memo says.
www.reuters.com/business/aer...
Anduril and Palantir battlefield communication system has deep flaws, Army memo says
The much-needed modernization of the U.S. Army's battlefield communications network being undertaken by Anduril, Palantir and others is rife with "fundamental security" problems and vulnerabilities, and should be treated as a "very high risk," according to a recent internal Army memo.
www.reuters.com
October 4, 2025 at 7:05 PM
Doesn’t sound ideal.
"We cannot control who sees what, we cannot see what users are doing, and we cannot verify that the software itself is secure," the memo says.
www.reuters.com/business/aer...
"We cannot control who sees what, we cannot see what users are doing, and we cannot verify that the software itself is secure," the memo says.
www.reuters.com/business/aer...
Reposted by Christian Schnell
"Believe me when I say: the things that ChatGPT has been telling users are probably worse than you think."
Former OpenAI Employee Horrified by How ChatGPT Is Driving Users Into Psychosis
The former OpenAI safety researcher had grim findings about how the company is handling these delusional spirals.
trib.al
October 5, 2025 at 6:57 AM
"Believe me when I say: the things that ChatGPT has been telling users are probably worse than you think."
Reposted by Christian Schnell
The 7 Cyber Security Trends Of 2026 That Everyone Must Be Ready For www.forbes.com/sites/bernar...
The 7 Cyber Security Trends Of 2026 That Everyone Must Be Ready For
From ransomware-as-a-service tools to state-sponsored cyber warfare, businesses face unprecedented threats that require immediate preparation and quantum-safe security.
www.forbes.com
September 27, 2025 at 9:57 AM
The 7 Cyber Security Trends Of 2026 That Everyone Must Be Ready For www.forbes.com/sites/bernar...
Reposted by Christian Schnell
How Cloudflare uses the world’s greatest collection of performance data to make the world’s fastest global network even faster
Cloudflare is using its vast traffic to send responses faster than ever before, by learning the characteristics of each individual network and tuning our conges…
#ai #news
Cloudflare is using its vast traffic to send responses faster than ever before, by learning the characteristics of each individual network and tuning our conges…
#ai #news
How Cloudflare uses the world’s greatest collection of performance data to make the world’s fastest global network even faster
Cloudflare is using its vast traffic to send responses faster than ever before, by learning the characteristics of each individual network and tuning our congestion control system.
blog.cloudflare.com
September 27, 2025 at 11:33 AM
Reposted by Christian Schnell
Google PM Ryan Salva is responsible for tools like Gemini CLI, giving him a front-row seat to the ways AI tools are changing software development.
How Google’s dev tools manager makes AI coding work | TechCrunch
Google PM Ryan Salva is responsible for tools like Gemini CLI, giving him a front-row seat to the ways AI tools are changing software development.
techcrunch.com
September 23, 2025 at 6:34 PM
Google PM Ryan Salva is responsible for tools like Gemini CLI, giving him a front-row seat to the ways AI tools are changing software development.
Reposted by Christian Schnell
Switzerland has launched Apertus, a national open-source LLM developed by EPFL, ETH Zurich, and CSCS. Built with transparency and compliance with European data laws, Apertus is positioned as a public alternative to commercial AI models.
#AI #OpenSource #LLM #Switzerland #Apertus
#AI #OpenSource #LLM #Switzerland #Apertus
Switzerland launches its own open-source AI model
Switzerland has just released Apertus, its open-source national Large Language Model (LLM) that it hopes would be an alternative to models offered by companies like OpenAI.
www.engadget.com
September 5, 2025 at 5:54 AM
Switzerland has launched Apertus, a national open-source LLM developed by EPFL, ETH Zurich, and CSCS. Built with transparency and compliance with European data laws, Apertus is positioned as a public alternative to commercial AI models.
#AI #OpenSource #LLM #Switzerland #Apertus
#AI #OpenSource #LLM #Switzerland #Apertus
Reposted by Christian Schnell
How Sakana AI’s new evolutionary algorithm builds powerful AI models without expensive retraining
M2N2 is a model merging technique that creates powerful multi-skilled agents without the high cost and data needs of retraining.
#ai #news
M2N2 is a model merging technique that creates powerful multi-skilled agents without the high cost and data needs of retraining.
#ai #news
How Sakana AI’s new evolutionary algorithm builds powerful AI models without expensive retraining
M2N2 is a model merging technique that creates powerful multi-skilled agents without the high cost and data needs of retraining.
venturebeat.com
August 31, 2025 at 9:58 AM
Reposted by Christian Schnell
There are solid use cases for AI. But they’re not the flashy gen AI apps that get investors excited.
“MIT found the biggest ROI in back-office automation—eliminating business process outsourcing, cutting external agency costs, and streamlining operations.” Not sexy, but this is where AI can really make people’s lives easier—by building interfaces with crappy software
fortune.com/2025/08/18/m...
fortune.com/2025/08/18/m...
MIT report: 95% of generative AI pilots at companies are failing
There’s a stark difference in success rates between companies that purchase AI tools from vendors and those that build them internally.
fortune.com
August 19, 2025 at 6:14 PM
There are solid use cases for AI. But they’re not the flashy gen AI apps that get investors excited.
Reposted by Christian Schnell
Red Teams Jailbreak GPT-5 With Ease, Warn It’s ‘Nearly Unusable’ for Enterprise
Red Teams Jailbreak GPT-5 With Ease, Warn It's ‘Nearly Unusable’ for Enterprise
Independent red teams have jailbroken GPT-5 within 24 hours of release, exposing severe vulnerabilities in context handling and guardrail enforcement, warning model is not enterprise-ready.
www.securityweek.com
August 10, 2025 at 8:36 AM
Red Teams Jailbreak GPT-5 With Ease, Warn It’s ‘Nearly Unusable’ for Enterprise
Reposted by Christian Schnell
Mistral Releases Devstral, an Open-Source LLM for Software Engineering Agents
Mistral AI announced the release of Devstral, a new open-source large language model designed to improve the automation of software engineering workflows, particularly in complex coding environments th…
#ai #llm #mistral
Mistral AI announced the release of Devstral, a new open-source large language model designed to improve the automation of software engineering workflows, particularly in complex coding environments th…
#ai #llm #mistral
Mistral Releases Devstral, an Open-Source LLM for Software Engineering Agents
Mistral AI announced the release of Devstral, a new open-source large language model designed to improve the automation of software engineering workflows, particularly in complex coding environments that require reasoning across multiple files and components. By Daniel Dominguez
www.infoq.com
May 27, 2025 at 6:26 PM
Reposted by Christian Schnell
Ransomware scum have put a target on the no man's land between IT and operations
Ransomware scum have put a target on the no man's land between IT and operations
Defenses are weaker, and victims are more likely to pay, SANS warns
Criminals who attempt to damage critical infrastructure are increasingly targeting the systems that sit between IT and operational tech.…
dlvr.it
May 14, 2025 at 6:39 AM
Ransomware scum have put a target on the no man's land between IT and operations
Reposted by Christian Schnell
Reposted by Christian Schnell
Reposted by Christian Schnell
In a total of 16 disciplines, the university ranks among the 10 best in the world. These are the latest results of the QS World #University #Rankings according to subject.
Article: ethz.ch/en/news-and-...
Article: ethz.ch/en/news-and-...
March 12, 2025 at 10:54 AM
In a total of 16 disciplines, the university ranks among the 10 best in the world. These are the latest results of the QS World #University #Rankings according to subject.
Article: ethz.ch/en/news-and-...
Article: ethz.ch/en/news-and-...
Reposted by Christian Schnell
1/13 New Paper!! We try to understand why some LMs self-improve their reasoning while others hit a wall. The key? Cognitive behaviors! Read our paper on how the right cognitive behaviors can make all the difference in a model's ability to improve with RL! 🧵
March 4, 2025 at 6:15 PM
1/13 New Paper!! We try to understand why some LMs self-improve their reasoning while others hit a wall. The key? Cognitive behaviors! Read our paper on how the right cognitive behaviors can make all the difference in a model's ability to improve with RL! 🧵
Scientists Just Linked Quantum Processors in a Historic Step Toward Scalable Supercomputers
scitechdaily.com/scientists-j...
scitechdaily.com/scientists-j...
Scientists Just Linked Quantum Processors in a Historic Step Toward Scalable Supercomputers
Scientists at Oxford have taken a massive leap toward large-scale quantum computing by successfully linking separate quantum processors into a fully connected system. This breakthrough overcomes a maj...
scitechdaily.com
February 10, 2025 at 8:35 AM
Scientists Just Linked Quantum Processors in a Historic Step Toward Scalable Supercomputers
scitechdaily.com/scientists-j...
scitechdaily.com/scientists-j...
Reposted by Christian Schnell
AutoGen 0.4 is out, Microsoft's multi-agent framework
AutoGen v0.4: Advancing the development of agentic AI systems
Announcing AutoGen 0.4, fully reimagined library for building advanced agentic AI systems, developed to improve code quality and robustness. Its asynchronous, event-driven architecture is designed to ...
www.microsoft.com
January 15, 2025 at 8:03 PM
AutoGen 0.4 is out, Microsoft's multi-agent framework