Archive · month
Amazon published a guide and AWS CDK architecture for memory lifecycle management for agents in Bedrock AgentCore - a combination of TTL expiration, scoring and consolidation that addresses the problem of agents referring to outdated context.
NVIDIA released Cosmos 3, an open omnimodal world foundation model for robotics and autonomous vehicles, under the OpenMDW-1.1 license. AWS published a guide to building a pipeline on Amazon SageMaker HyperPod in which one model replaces three separate GPU queues.
Amazon described a procedure combining Amazon Textract and Amazon Bedrock for RAG processing of complex multi-page documents (PDF, DOCX, TXT, HTML, PNG, XLSX), intended to reduce errors and hallucinations compared with uploading raw files. The code is on GitHub.
Since 7 September, Project HydraFusion has been available in experimental mode (/experimental) in GitHub Copilot CLI. It is a system that automatically selects and combines AI models for a given task; according to GitHub, it achieved higher quality at lower costs than Claude Opus 5 in the TerminalBench 2.1 test.
New: Specific rollout date: 2026-09-07 (missing from the existing event); Available in experimental mode through the /experimental command; Part of GitHub Copilot CLI (scope clarification); Can be selected like any other model in the interface; Automatic routing without manual configuration by the user
Instagram incorrectly labels conventionally edited photos (Canva Background Remover, skin retouching) as “AI Content”, while actual AI content often escapes detection. A similar problem is recurring after the incident in February 2024.
Restaurants and food brands are increasingly replacing food photographs with AI-generated images that, according to The Verge, often look distorted and repulsive – with strange textures, holes and unnatural shapes.
Applicants pay tools such as Jobscan for AI optimization of their CVs for ATS filters, while recruiters deploy their own AI screening in ATS because of the flood of similar applications — according to the CEO of Greenhouse, this is a “doom loop” in which AI on both sides makes the problem worse.
According to Bloomberg, Crusoe has raised 3 billion dollars in a new funding round at a valuation of 30 billion dollars; the round is led by Atreides Management and Valor Equity Partners.
The recommended replacement for Claude Opus 4.7 in GitHub Copilot is Claude Opus 5.5; the original announcement listed Claude Opus 5. The retirement of four models takes effect on 2 October 2026 across all environments of the service.
New: The recommended replacement for Claude Opus 4.7 is Claude Opus 5.5, not Claude Opus 5.
Thinking Machines, the lab founded by Mira Murati, is in talks for a funding round of 1 billion dollars at a valuation of at least 40 billion dollars, led by Accel. This is a lower valuation than the 50 billion dollars the company originally sought.
The company Pangram hired a journalist who used its detection score to publicly ridicule users suspected of using AI. It ended the collaboration, but CEO Max Spero continues to publish scores – yet the tool measures only AI involvement, not the extent of human contribution.
AWS published a guide to connecting Microsoft Outlook with the generative AI assistant Amazon Quick, which summarizes email threads, suggests replies, and schedules meetings. The integration uses Microsoft Graph API and OAuth 2.0 authentication.
AWS published best practices for creating agentic automations in Amazon Quick Automate: an emphasis on understanding the process, clearly defined agent responsibilities, human oversight, and deterministic safeguards against fragile workflows.
NVIDIA and Microsoft announced new tools at IFA 2026 for faster inference and easier setup of AI agents running locally on NVIDIA hardware. Compact NVIDIA RTX Spark computers running Windows are expected to be available from October 2026.
Google is launching Gmail Live, Docs Live and Keep Live voice modes for the Gmail, Docs and Keep mobile apps, available to paying Google AI Plus/Pro/Ultra users; they are due to arrive later for Workspace customers.
OpenAI has resolved an incident involving elevated error rates in ChatGPT and Codex, and all monitored components are operational again. Users of Codex Remote Control may need to pair their mobile device again.
GitHub is gradually reopening Copilot Business and Enterprise sign-ups for payments by card or via PayPal and introducing stricter account verification and revised billing from 1 October 2026.
According to Vals AI, the model Claude Fable 5.1 independently solved the previously unsolved Cyphral Distich cipher by Sir Thomas Urquhart from 1653, doing so in 44 minutes without human help. The other models tested did not solve the cipher.
The Grok outage on Android on 3 September 2026 was part of simultaneous morning outages affecting Claude, ChatGPT/Codex and Grok. SpaceX cited an outage at a data center in Memphis as the cause of the Grok outage, but Anthropic and OpenAI did not confirm a common cause.
New: The Grok outage was part of a simultaneous all-day incident involving Claude (6:23–9:16 PT) and ChatGPT (7:43–8:17 PT); The cause was an identified outage at the xAI/SpaceX compute center in Memphis; xAI and Anthropic had compute partnerships with SpaceX from May 2026; Google Gemini also reported problems, but the vendor did not confirm them; No shared third-party provider (cloud, CDN) was identified as the cause
Hugging Face released open-source multimodal encoders NeoMME (260M and 800M parameters, Apache 2.0) for document retrieval from page images — according to the company, 2× faster than ColModernVBERT and with an index reduced in size by 255×.
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
