Archive · month
NVIDIA has released Isaac ROS 5.0 – a GPU-accelerated set of packages built on the open-source framework ROS, designed for developing robotic applications with physical AI models capable of perception, reasoning, and action.
Meta released a hotfix for a zero-day vulnerability in the macOS Muse app, discovered by researcher Patrick Wardle. The flaw allowed any local process to redirect speech transcription to a foreign server and thereby obtain an authentication token for full account takeover.
New: Any local process could access the authentication token without macOS permissions; An attacker could change the transcription endpoint to their own server and take over the account; Patrick Wardle created working proof-of-concept attacks for demonstration purposes; The goal of the exploit was to obtain the authentication token, not direct control of the agent
A Wired reporter used Gemini to create AI clones of two of his editors from their articles, podcasts and Slack messages. The clones captured the basic communication style but slipped into typical AI clichés and, on one occasion, drew on an incorrect source about another person.
The Austrian Academy of Sciences, together with Mistral and Sail Reply, is releasing the Apollo model, trained on 600 million words of ancient Greek. It fills in missing passages in damaged papyri and manuscripts and is available free through a chatbot for academics, but the final choice of words remains with experts.
An interview on Živě.cz with AI education advocate Marek Dědič about how AI has become a routine part of schools, why bans are not enough, and how to teach students to think critically about chatbots.
OpenAI released the GPT-6 Sol and Luna models on 22 September 2026 at prices 50% lower than GPT-5.6 (Sol 2/10 USD, Luna 0.10/0.50 USD per million tokens). However, an independent analysis by Artificial Analysis shows that actual capabilities remain at the level of their predecessors and have even declined on some tasks.
New: Anthropic released Claude Opus 5.5 on the same day (2026-09-23), roughly an hour after OpenAI; Claude Opus 5.5 is 20% cheaper than the previous Opus 5.0 ($4/$20 vs $5/$25); Opus 5.5 improved communication style and token efficiency; The model releases launch a 'price war' between the two largest AI companies; GPT-6 Luna and Sol are part of a broader strategy of price competition, not an isolated release
Google has published details about the models Gemini 3.8 Flash TTS and Flash-Lite TTS (GA from 22. 9. 2026): over 100 languages, voice creation from a text description, voice cloning from a 30-second recording, 2000+ preset voices, SynthID watermark.
New: The models support more than 100 languages; Ability to generate voices from text descriptions; Voice cloning from 30-second audio samples; Over 2000 preset voices with regional variants; Integrated SynthID watermark for detecting AI speech
Anthropic has made it possible in the beta version of Claude API (header inline-tools-2026-09-15) to define and modify tools directly in the system message midway through a conversation, without having to edit the tools array or invalidate the prompt cache.
Hugging Face has added native support to the transformers library for the GGUF format for local inference on Apple Silicon, including quantization levels such as Q4_K_M that shrink models to a fraction of their original size.
Jun Kim, the creator of the oMLX project, has become a full-fledged contributor at Hugging Face and will lead its further development. The project remains under the Apache 2.0 license, and the goal is to speed up the conversion of transformers models to the MLX format for local execution on Apple Silicon.
Ron Johnson, the creator of Apple stores, is skeptical that AI agents (e.g. Universal Commerce Protocol from Google or shopping through ChatGPT from OpenAI) will replace in-person shopping for more expensive goods such as laptops.
California Governor Gavin Newsom signed a package of seven laws intended to prevent the energy and water costs of AI data centers from being passed on to residents – the package introduces a new rate, mandatory disclosure of water consumption, and conditions for expedited project approval.
The USA and China accuse each other of propaganda operations in the debate on AI regulation and data center construction. American figures (Trump, Sacks, Tan) speak of Chinese-backed psy-ops, while Chinese state media accuse the USA of seeking unilateral control.
According to estimates from Apptopia, in its first 12 days the Muse app from Meta had more downloads (1.8 million on iOS in the USA/Canada vs. 1.3 million for ChatGPT) and daily active users (642 000 vs. 231 000 in the USA) than ChatGPT in the comparable period.
Grok 4.6 from xAI has been available in Amazon Bedrock since 18 August 2026, with a context window of 500 thousand tokens and four reasoning effort levels, targeting long-running agents and work with code.
An analysis by SemiAnalysis describes how the Mixture of Experts architecture changes the requirements for inference infrastructure – which tensors remain active, how data moves between the prefill, midfill and decode phases, and how orchestration layers such as NVIDIA Dynamo or Mooncake coordinate this.
The company xAI has released Grok 4.7 at $2/million input tokens and $6/million output tokens. On the Artificial Analysis index (v4.3.2), it achieves a score of 46, while Claude Fable 5.1 and GPT-6 lead with a score of 53 each. In the agentic coding test Terminal-Bench 4.0, it scores only 26 %.
Tabby, a startup founded by former accountant Ahad Ali, offers small businesses AI-powered accounting automation with connections to bank data through Plaid. After 14 months, it has 5 500 businesses, roughly 100 000 dollars in ARR and is seeking 1 million dollars in a pre-seed round.
Posit has integrated the Positron IDE with Amazon SageMaker AI — data scientists can combine R, Python, XGBoost, Shiny and Quarto in one environment with an optional AI assistant connected through Amazon Bedrock. It requires an ml.t3.xlarge instance or larger.
Benchling deployed a security architecture for running code generated by AI agents across thousands of tenants on Amazon Bedrock AgentCore – with a separate AWS account, DNS Firewall and per-job credentials. According to the company, there have been no security incidents since launch.
AI Radar monitors Czech and international sources every day, looking for changes that truly deserve attention.
The result is a quick overview of what has actually changed in the AI world, rather than another stream of articles.
Use the CS/EN switch to read the same Radar in Czech or English. English content is published after its translation has been checked, so new and older items may appear later.
Each row represents one event — not one article. At a glance, you can see its significance, credibility and main point.
Illustrative example, not a current news item.
Agent mode is available to all paying users
Until now, the mode was available only on the highest plan; it is now available on all paid tiers without a waitlist.
The detail page contains a fuller summary, its significance and original sources. Practical impact appears in the detail and the For individuals and For businesses views. An AI Flash item reaches the main selection only after it has been expanded and meets the publication rules.
We first summarize each event in the same way for everyone. Based on those same facts, we then explain what the change means for your own use and what it could mean for how a company operates.
Today’s briefing is the same for everyone. Pages For individuals and For businesses can be found in the main navigation — they select only events relevant to the given use case.
Reset history · Outage history · Subscribe in a feed reader (RSS) · Methodology and label explanations
