AI Flash · live feed
A timeline of everything that has passed through the radar — newest first, with an evidence status for each item.
Continuously updated
1096 items in the selected feed
How to read evidence statuses ↓
The author of the article on Root.cz describes a concept in which a software agent automatically divides a task among isolated roles (author, reviewer, checker), while context separation can, according to him, be achieved with a single model, not necessarily with multiple providers.
Commentator Petr Koubský questioned the sincerity of the call by Dario Amodei (Anthropic) to slow AI development and suggested a business motive ahead of the planned IPO. Amodei previously warned about the rapid development of recursive self-improvement and proposed auditors and SALT-style global agreements.
Startup ScrollEd is pitching an app at TechCrunch Disrupt Startup Battlefield 200 that uses AI to turn PDFs and textbooks into a vertical feed of videos, audio, text and quizzes in the style of TikTok and Instagram Reels.
Tencent introduced Gander, a research model that holds conversations in real time while also handling background tasks using a separate “cerebellum” and a replaceable “brain”. In tests, it had the best response timing but lower task accuracy.
According to Joshua Corman from Institute for Security and Technology, US energy systems remain vulnerable to cyberattacks, as they have long been. In his view, the main security risk is people, not artificial intelligence, despite recent concerns about AI attacks.
Runway has presented research into a system that would stream video continuously from a text description instead of making users wait for a finished clip. It is built on the GWM-1 model and addresses the accumulation of errors by training on its own outputs; no availability date has been announced.
Trump defends the construction of AI data centers as necessary to compete with China, but faces opposition from 60 % of Americans (47 % of Republicans), warning memoranda from his own party (NRSC, Fabrizio), and resistance from his own voters on Truth Social.
Pharmaceutical companies are increasingly basing drug development on AI infrastructure provided by companies such as Nvidia, Google, Microsoft or Amazon. This accelerates research but carries the risk of dependence on a single supplier and uncertainty over who holds the rights to drugs developed with the help of AI.
The Muse assistant from Meta on macOS incorrectly and confusingly explained how it accesses data from Messages. Meta confirmed that access is opt-in and that this was an error in the explanation, not unauthorized reading of messages.
Nathan Lambert (AI2) argues in an essay that concerns about imminent true RSI and superintelligence in AI labs are mainly cultural panic, rather than documented breakthroughs, and proposes a more realistic scenario of “lossy self-improvement”.
According to a World Bank study, AI threatens 14.2 % of jobs in high-income countries, including Czechia, compared with just 4.5 % in poorer countries. At the same time, AI can increase productivity in a similar proportion of jobs in both groups of countries.
Startup Vals raised 40 million dollars in a Series A round led by Andreessen Horowitz to develop AI benchmarks that test the ability of models to handle complex industry-specific tasks (law, finance, code) and, unlike traditional tests, do not make test materials publicly available.
The ICLR 2027 conference received about 50 000 abstracts before the deadline, 2.5 times more than ICLR 2026 (19 500). The increase is attributed mainly to faster writing using AI; concern is also growing over AI-generated papers with fabricated citations and an overloaded review process.
Mathematician Buckmaster accuses OpenAI of adopting his approach to Navier-Stokes without credit. Andreas Thom similarly claims that the Astra model used his techniques. Both mathematicians used Claude and ChatGPT for their work, but OpenAI did not acknowledge their contribution.
AI text detectors work with probability, not certainty, and the chosen decision threshold affects the false alarm rate. In the test, Pangram was the most accurate and most cost-effective tool.
The startup studio Vantora (formerly UP.Labs) has secured 100 million dollars from Silversmith Capital Partners and is shifting to a model of developing proprietary physical AI startups exclusively for corporate partners such as Porsche, Alaska Airlines or J.B. Hunt.
According to CNN, a US analyst used an AI chatbot to synthesize and format intelligence about a Chinese ship; the tool incorrectly identified its cargo. Aircraft were already airborne when the error was discovered and the operation was canceled at the last minute.
Presidents Trump and Xi Jinping will meet next week at a state dinner in Washington, where the rivalry between the USA and China in AI and access to chips is expected to be a topic of discussion. Executives of technology companies have also pledged to attend.
AWS summarized 13 new features for Amazon SageMaker Inference released in 2026. The main addition is Inference recommendations – automated selection of an instance and model deployment configuration, which, according to AWS, replaces 2–3 weeks of manual benchmarking.
Part of an overview of multiple AI topics; only this event has been covered.
According to CNN, an AI chatbot hallucination in an analysis by a US military analyst almost led to action against a Chinese ship over a fabricated cargo of nuclear components; according to the source, the incident “almost sparked a war”.
GitHub has made improvements to Copilot code review available — clearer summaries of findings, smarter automatic resolution of comments, and automatic commit message generation when accepting suggestions. The features are now generally available.
Simon Willison published a commentary in which he compares ignoring LLM research to a geneticist ignoring the opening of Jurassic Park. He emphasizes the necessity of scientific interest in LLMs.
GitHub Copilot is adding three tiers to automatic model selection – efficiency, balance and intelligence – for choosing between lower cost/speed and higher quality. They are gradually rolling out to VS Code, Copilot CLI and the Copilot app.
Part of an overview of multiple AI topics; only this event has been covered.
FAA is launching the SMART AI system to forecast and coordinate air traffic, initially at three airports in the Washington DC area from 21 September 2026, with plans to expand across the USA. The budget is 875 million dollars, and the tool serves only as an adviser and does not change controller procedures.
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
