Sakana released Fugu Ultra v1.1, an update to the model router. It claims improvements of up to 7.9 points and that the system outperforms Fable 5, even though Fable 5 is not included in the router. Price: $5/$30 for millions of tokens. All figures come from Sakana; independent verification is lacking. An endpoint compatible with Claude Code has been added.
An upcoming meeting in Europe brings together policymakers, industry and academia to discuss European leadership in AI-powered robotics. euROBIN (a Horizon Europe network) is preparing an event with demonstrations of 20–30 advanced robots and a discussion of a joint European initiative for physical AI (robots that perceive, reason and act).…
European Commission — Shaping Europe's digital future (AI Act, EU regulation)
Original source ↗
An incident involving increased error rates for the Opus 4.8 model affected Claude.ai, Claude Console, Claude API, Claude Code and Claude Cowork. The incident has been resolved and is being monitored.
Duvo.ai, a startup founded by Tomáš Čupr and former colleagues from Rohlík, has secured Alza as a customer. The AI agents are intended to first map business processes and then redesign them using automation across systems such as ERP, email or supplier portals.
Although young people use chatbots, 37 % of teenagers mock AI content and more than half fear disinformation and deepfakes. More than half of the generation is concerned about the negative impacts of AI on society, including the loss of jobs and critical thinking. Young people perceive AI marketing by adults as pressure.
NVIDIA and South Korean university KAIST have established a joint AI research lab in Seoul focused on agentic AI. It is the first partnership of its kind between a Korean university and a global technology company. At the same time, collaboration between SK and NVIDIA on developing memory for AI platforms is expanding.
MKB (Monkey King Bang), a new model from the Shanghai Academy of AI for Science, unifies scientific research across six fields: DNA, RNA, proteins, small molecules, geophysics and medical images. It supports native outputs for biological sequences, molecular strings and meteorological fields. The model is built on…
An arXiv preprint describes an autonomous agent based on GPT-5.5 Pro that generated mathematical proofs disproving the sum-product conjecture over the real numbers in 7 of 8 attempts, each using an average of 132.4 thousand reasoning tokens. The authors released the code and proofs for reproduction.
A research paper presents AINTMA, a system of six AI agents for automated testing. It integrates LLMs for quality reports, reinforcement learning for test prioritization, and zero-trust security. It claims 88.4% accuracy (vs. 51.2% chance), a 43% reduction in testing time, and a reduction in defects from 8.3% to 2.1%.
According to TechCrunch, restrictions (guardrails) on models from Anthropic and OpenAI intended to prevent misuse for cyberattacks also hold back legitimate security researchers. The companies offer vetting programs with less restrictive limits, but researchers complain about inconsistent model behavior and bypass them with open-source…
An incident involving elevated error rates and API latency for gpt-image-2 lasted from the evening of 23 July to the morning of 24 July. All affected services were fully restored.
A guide to configuring guardrails (safety measures) in Amazon Bedrock for AI assistants that generate code. It addresses performance and cost issues when deploying to multiple developers, with a case study involving 15 developers and throttling errors.
In a decision dated 23 July, Judge Paul Felix noted that a court transcript contained errors resembling generative AI output. The warning addresses court reporters about the risks of using AI tools in legal documents without verification.
GitHub is adding the ability to ask the Copilot AI agent in its mobile app to investigate and fix failing GitHub Actions checks. The agent analyzes the error, creates a pull request with a proposed fix and assigns a developer to review it. The feature is available on iOS and Android.
OpenAI is launching Health in ChatGPT in the USA. According to the company, free users will get worse advice (GPT-5.5 Instant), while paid users will get GPT-5.6 Sol. Apple Health can be connected. Reportedly, 300 million people ask it healthcare questions each week.
Schneider Electric and AMD released a reference design for high-density AI data centers built on the AMD Helios platform – it supports racks up to 246 kW and clusters with up to 10.4 MW of IT capacity, and is intended for both new and retrofitted facilities.
Anthropic has expanded voice mode in Claude, previously based only on the Haiku model, to include the more powerful Opus and Sonnet. It now connects to Gmail, Google Calendar, Slack, Canva and Notion and can send an email or modify a meeting on behalf of the user through voice commands.
Meta's ads for AI services use David Bowie's song Five Years, which deals with the end of the world, but select only passages about family and nature. Critics point to manipulation or deliberate disregard for the original context.
Startup AegisAI, founded by former Google security managers, has raised 36 million dollars in a Series A round led by Battery Ventures to develop AI agents to combat targeted phishing attacks (spear phishing).
Patreon is laying off 93 employees (20 % of the team). CEO Jack Conte explains the decision by saying that AI is fundamentally transforming the technology industry and that companies that do not fully adopt AI tools will not be competitive in three years.
Jefferies built an AI agent-based trading assistant on AWS that provides traders with real-time analysis of data from multiple sources using natural language, without coding. The solution uses Anthropic's Claude, Amazon Bedrock, Knowledge Bases and MCP to connect to market data. The assistant accelerated data exploration from days…
Amazon Bedrock AgentCore is an optimization feature that detects silent AI agent failures — errors without error signals, such as incorrectly performed actions or omitted checks. The tool analyzes traces using a taxonomy of 11 failure categories and groups them into patterns. According to Amazon, it helps distinguish…
AWS is expanding Amazon Bedrock Managed Knowledge Bases with agentic retrieval mode — the new AgenticRetrieveStream API automates multipart queries instead of performing a single search. The model breaks complex questions into subquestions and searches iteratively, addressing the shortcomings of traditional top-k…
AI Radar monitors Czech and international sources every day, looking for changes that truly deserve attention.
MonitorsOfficial AI company blogs, specialist media, and research sources.
Selects and combinesFilters out information noise and combines articles about the same change into a single event.
Summarizes and explainsExplains significant events in English: what happened, why it matters and where the information comes from.
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.
Everything you need to navigate the AI world
Today’s briefingThe “What is worth attention” selection sits beside Live · AI Flash, followed by research and links to other Radar sections. On mobile, these blocks appear one below another.
AI FlashAn ongoing feed of brief updates with an evidence status. Links lead to a Radar detail page when one is ready, otherwise to the original source. You can also find reset and outage histories here.
Practical applicationsWhat new tools and features can do, what you can try and what their actual impact could be.
Model selectionModel comparison by type of work, capabilities, price and speed.
Research and archiveA separate research overview, topic search and older events by date.
One event, everything that matters
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.
Importance: ▮▮▮ majorOpenAIModels✓ 6
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.
▮▮▮ major · ▮▮ important · ▮ we're tracking = how significant the change is✓ 6 = six independent publishers, not the number of articles or feeds✓ official = a clear release, law or incident is substantiated by the relevant authority1 source = no independent confirmation yetbold = who is behind the changegray text = a brief summary of what happened
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.
The same news, two practical uses
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.
For individualsWhat you can use or try, how the change can help you at work and what to watch out for.
For businessesWhat impact the change could have on processes, costs, risks and other business decisions.
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.
