Saturday, Oct 24

Anthropic published this on Oct 24, 2026 (source: platform.claude.com).

Anthropic’s Claude Tag API lets Claude Enterprise admins read and change Claude Tag settings on Slack channels and workspaces, see which settings apply and why, and attach Access bundles from their own systems.

Hugging Face @internlm published this on Oct 22, 2026 (source: huggingface.co).

The internlm account publishes Atria-Dawn on Hugging Face as safetensors under the glm_moe_dsa / GlmMoeDsaForCausalLM stack, tagged with arXiv 2609.15818.

Suno published this on Oct 24, 2026 (source: blog.suno.com).

Suno’s October 23 iOS release notes say the app is rebuilt for the iPhone Duo screen and adds AirPlay so songs and playlists can play on AirPlay speakers, TVs, and smart-home devices (iOS 1.93+).

Huggingface published this on Oct 23, 2026 (source: github.com).

Diffusers v0.42.0 brings FLUX.3 Image Klein, Bria Fibo 2.0, Kandinsky 6, SANA-WM, and RFD4-Proteína into the library, with new adapter support, hardware integrations, and modular pipelines. Python 3.11+ is now required.

Langchain-ai published this on Oct 23, 2026 (source: github.com).

langchain-ai/langgraph tags sdk==0.4.9 with a Priority field for thread priority settings in the Python SDK, plus dependency cleanups since 0.4.8.

Not confirmed by the company — reported on Oct 23, 2026 (source: theverge.com).

The Verge reports that Google’s updated privacy terms for Personal Intelligence could mean sharing data for better AI recommendations also enables more personalized ads from Gmail and Google Photos signals.

Not confirmed by the company — reported on Oct 23, 2026 (source: the-decoder.com).

The Decoder reports Sam Altman arguing that sandboxes and monitoring are temporary, that advanced AI will inevitably leave containment, and that alignment remains an unsolved problem, while warning models already find hardware flaws humans miss.

Hugging Face Blog published this on Oct 22, 2026 (source: huggingface.co).

Hugging Face’s blog shows how to look up whether a model runs on AWS Trainium via TorchNeuron (torch-neuronx), including what “runs” means and a results dataset built on the native PyTorch Trainium backend.

Hugging Face @nvidia published this on Oct 22, 2026 (source: huggingface.co).

NVIDIA posts InternVideo2-CLIP-anomaly-detection on Hugging Face with TensorRT/ONNX/safetensors builds tagged for video-text retrieval, video embeddings, physical AI, and anomaly detection.

GitHub published this on Oct 24, 2026 (source: github.blog).

GitHub’s changelog says the consolidated GitHub app in Microsoft Teams (formerly GitHub notifications) lets teams @mention GitHub on an issue notification so Copilot can investigate, open a fix PR, and keep review in the chat. The legacy GitHub app is marked deprecated and set to retire on November 16, 2026.

GitHub published this on Oct 24, 2026 (source: github.blog).

GitHub’s Copilot usage metrics API now reports model usage from clients outside the IDE, including OpenCode, Claude Code, Codex, Zed, Copilot CLI, and others, via a new totals_by_client array on enterprise and organization 1-day and 28-day reports. The change is meant to capture server-side model traffic that editor telemetry missed.

Langchain-ai published this on Oct 23, 2026 (source: github.com).

langchain-ai/langgraph tags 1.2.17 with fixes that apply an explicit recursion_limit equal to the default and keep the caller’s namespace in RemoteGraph.stream(subgraphs=True), plus dependency cleanups since 1.2.16.

Not confirmed by the company — reported on Oct 23, 2026 (source: theverge.com).

The Verge reports that Jack Dorsey’s Block announced Dessa, an AI model trained from scratch on nothing but human conversation, and shared sample chats that can sound awkward or uncanny.

Not confirmed by the company — reported on Oct 23, 2026 (source: techcommunity.microsoft.com).

Microsoft Foundry’s blog says the Azure OpenAI Decisions API is now generally available for agents that must choose the next step—routing, urgency, quality checks, or whether a workflow continues—without generating long-form content.

Not confirmed by the company — reported on Oct 23, 2026 (source: marktechpost.com).

MarkTechPost reports that Kyutai released CADRE, a family of seven open causal diffusion video autoencoders for interactive world models, with world models said to run at 20 fps on one H100 and react in under 230 ms.

Ollama published this on Oct 24, 2026 (source: github.com).

Ollama tags v0.40.11-rc0 with fixes that preserve literal function closing tags in Qwen3 Coder parameters, use allocated context for OpenCode, allow empty state for vllm-d3 decisions, bump llama.cpp, and expand Phonon voice catalog and symbol spelling in mlxrunner.

@MiniMax_AI on X published this on Oct 23, 2026 (source: x.com).

MiniMax thanks Fireworks AI for bringing MiniMax M3.1 Flash to serverless on day zero and points builders to a serverless preview for multi-step agents.

@LayerbaseDB on X published this on Oct 23, 2026 (source: x.com).

Layerbase shows DocumentDB creating cosmosSearch vector-HNSW indexes and running $vectorSearch from mongosh, noting pgvector underneath for embedding retrieval.

@Replit on X published this on Oct 23, 2026 (source: x.com).

Replit announces that Chroma (TryChroma) is joining Replit to bring the right data to the right models and build a context layer so companies can run more self-driving agents.

@TencentHunyuan on X published this on Oct 23, 2026 (source: x.com).

Tencent Hunyuan introduces Hy-Nomas as an open-source interactive video world model with text and camera-trajectory control, long-horizon generation with world memory, lightweight LoRA adaptation for domains such as robotics, and consumer-GPU inference, with website, GitHub, Hugging Face, and demo links.

@cognition on X published this on Oct 23, 2026 (source: x.com).

Cognition introduces a refreshed Devin Desktop app focused on a smoother agent command center for running and supervising Devin.

@cursor_ai on X published this on Oct 23, 2026 (source: x.com).

Cursor says its agents can drive desktop apps such as the iOS Simulator to test features they build, expanding agent computer use beyond the editor.

@TencentGlobal on X published this on Oct 23, 2026 (source: x.com).

Tencent’s WorkBuddy pitch says the product can bring in a team of AI Experts for proposal work covering financials, risks, and positioning while leaving the final call to the user.

@Hailuo_AI on X published this on Oct 23, 2026 (source: x.com).

MiniMax’s Hailuo account says MiniMax Design MCP is live so builders can create with Design inside Codex and other agents, with creations auto-syncing to Design Canvas.

@elonmusk on X published this on Oct 22, 2026 (source: x.com).

Elon Musk posts that Grok Bot now integrates with phone apps, pointing to a download/demo link for the expanded mobile integration.

@MiniMax_AI on X published this on Oct 22, 2026 (source: x.com).

MiniMax says it is partnering with Inferact to advance production inference for MiniMax models, with engineering improvements contributed back to vLLM, and thanks Inferact for day-one production support of M3.1 Flash.

@alexandr_wang on X published this on Oct 22, 2026 (source: x.com).

Meta’s Alexandr Wang posts a $100,000 challenge for the developer who can make Muse Spark Open run fastest, linking the contest details.

@MiniMax_AI on X published this on Oct 22, 2026 (source: x.com).

MiniMax thanks the NVIDIA NeMo team for day-zero fine-tuning support for MiniMax-M3.1-Flash and asks what post-training recipes the stack will unlock.

@Replit on X published this on Oct 22, 2026 (source: x.com).

Replit introduces Monitoring Agent as an on-call engineer in your pocket that can see issues, fix them, and deploy changes autonomously for AI-app builders.

@Kling_ai on X published this on Oct 23, 2026 (source: x.com).

Kling AI says Kling 4.0 is now live on Pollo AI.

@Kling_ai on X published this on Oct 23, 2026 (source: x.com).

Kling AI says Kling 4.0 is now live on idea_ai (@officialidea_ai).

Not confirmed by the company — reported on Oct 24, 2026 (source: 36kr.com).

36Kr, citing other media, reports that Utopai Studios—described as the world’s largest independent AI film company—closed a financing round at a $1 billion valuation and is discussing a follow-on round at a $2–3 billion valuation.

Not confirmed by the company — reported on Oct 24, 2026 (source: 36kr.com).

36Kr, citing Capital Construction News, says that at the 2026 World Intelligent Connected Vehicles Conference a China Mobile–led national AI application pilot base for auto manufacturing released the sector’s first shared AI base. The base offers compute–data–model capacity (about 112.7 EFLOPS managed) and six agents spanning design, process planning, visual QC, ops, driving-data generation, and model training, already used by more than ten central and state-owned automakers.

Not confirmed by the company — reported on Oct 23, 2026 (source: 36kr.com).

36Kr reports that the UK will spend £100 million (~$134 million) on an AI biosafety laboratory backed by OpenAI, Anthropic, and Google DeepMind. The lab is meant to use AI to identify and monitor biological threats; OpenAI and Anthropic are said to provide up to £200 million of model compute, with further work involving DeepMind and Basecamp Research.

Not confirmed by the company — reported on Oct 23, 2026 (source: 36kr.com).

36Kr, citing Jiemian News, reports Amazon CEO Andy Jassy saying AI infrastructure demand has largely sold out AWS cloud capacity through the end of 2027, with many orders already stretching into 2028, as Amazon races to serve AI labs and enterprises.

Not confirmed by the company — reported on Oct 24, 2026 (source: the-decoder.com).

The Decoder reports a controlled study finding that switching coding-agent harnesses such as Claude Code, OpenCode, or mini-SWE-agent barely moves pass rates beyond re-running the same setup, while cost per task can roughly triple—driven mainly by system prompts resent at every step.

Not confirmed by the company — reported on Oct 24, 2026 (source: 36kr.com).

36Kr profiles Today AI, founded by former Teambition founder Qi Junyuan, saying the Personal AI startup has raised about $200 million and has launched international and China apps since September 9. The piece frames Today AI against the broader Personal AI wave that includes Muse and Instinct.

Not confirmed by the company — reported on Oct 24, 2026 (source: 36kr.com).

36Kr reports that on October 24 Tencent WorkBuddy said it has integrated Moonshot’s Kimi K3.1 Pro, including a Highspeed tier that keeps the same intelligence with about 2–3× output speed. Access is currently prioritized for Enterprise Flagship and personal subscribers amid tight capacity.