Archive
October 19 – October 25
New user profiles allow API requests to be linked to specific end customers. According to Anthropic, this means safety enforcement action can affect one customer instead of an entire organization.
Transformers and Diffusers now support TPU through TorchTPU, which adds another backend to PyTorch. According to Hugging Face, Optimum TPU is no longer needed to run them on TPU. However, some tests are not yet passing.
The release of Voice Agents has been announced. Through Voice Agents SDK, the service provides managed processing for real-time, two-way voice communication, with interfaces for telephony, browsers and custom audio input.
AnyLanguageModel 1.0 for Swift has been released. The same session code supports the system model, local models and hosted services. According to the authors, tool calling and structured generation work across all providers. The library also runs on Linux.
Organization administrators can disable the web fetch tool or configure allowed or blocked domains in Claude Console. Disabling it causes a 400 error for Messages API requests that include the tool; in Claude Managed Agents sessions, the agent receives an error result.
Claude Managed Agents sessions have respected organizational web search permissions since 6 October and domain restrictions since 14 October. Results must satisfy both organizational rules and the web_search tool settings.
Admin API from Anthropic now supports notifications about restrictions or bans on user profiles within an organization and the removal of those measures. The feature is in beta, and access requires a request to the customer team.
Amazon Bedrock has introduced outputs based on JSON Schema and strict validation of tool parameters on the bedrock-runtime and bedrock-mantle endpoints. Support covers all Claude models available on the respective endpoint.
New customers can get started with Amazon SageMaker Studio after signing up in 2–3 steps, usually without a credit card. According to AWS, the environment is set up in about 20 seconds. A 100 USD credit is included; the free period ends after six months at the latest.
Microsoft introduced the Guardian Agents feature in Project Perception to monitor corporate AI agents. It cites the unpredictability of models as the reason oversight is needed and recommends separating the provision of models from the authority to control their actions.
Microsoft introduced Copilot for Legal, which adds more than 20 legal AI skills and plugins to Microsoft 365. According to the company, it supports contract drafting and review as well as the preparation of legal analyses.
The new REST API allows model availability policies in GitHub Copilot to be read and updated at the enterprise, enterprise team, enterprise application, and organization levels. GET and PATCH operations are generally available.
GitHub ended support for six models in most GitHub Copilot environments as of 19 October 2026. It recommended replacements for the affected workflows and integrations; Copilot Enterprise administrators may need to enable access to the alternatives.
TorchNeuron enables applications using PyTorch to move to Trainium chips. According to Amazon Web Services, changing one line of configuration is sufficient, with no need to rewrite code.
The guide from GitHub shows how to open a session from the GitHub Copilot app in VS Code while preserving the conversation, code changes and working context. It then covers reviewing changes, making manual edits and verifying the project locally.
Kling AI has released Kling 4.0 with generation of up to 30 seconds of video and support for up to ten input keyframes. Access through MCP and CLI has been added. The Kling 4.0 Flash variant is limited to 720p and 20 seconds; the company is still developing API support.
AWS has made Pegasus 1.5 from TwelveLabs available in Amazon Bedrock. According to AWS, the model finds video segments based on specified criteria and returns structured metadata with timestamps. It is available in the us-west-2 region in on-demand mode.
Deepgram and AWS ported the Nova-3 model to AWS Trainium2 chips. According to their measurements, a configuration with eight replicas on a single instance processed 11–70 % more audio per dollar than the compute options previously used. The results apply to batch transcription.
The creator of cleanyourtv published a prompt on GitHub that, according to his description, he used to have Claude Code modify a TV over Wi-Fi and create a simple home screen with apps. The solution does not require root access, but it is still in development.
Part of an overview of multiple AI topics; only this event has been covered.
Bidirectional KV cache transfer in vLLM allows earlier computations to be reused in subsequent conversation turns. In its own measurements, AWS reports time-to-first-token speedups of up to 1.6-fold for Qwen3-32B and up to 3.3-fold for zai-org/GLM-5-FP8.
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
