Replit Cloud Agents Opens Enterprise Codebase Contributions, Non-Technical Staff Can Submit Tested PRs Directly
Replit launches Cloud Agents, available to Pro and Enterprise customers, letting non-engineering staff such as designers and support teams submit tested pull re…
Gamma launches Gamma 5 plus multiple updates to credits, export, preview, and more
Gamma has released Gamma 5, which it says rebuilds the way you think, design, and edit. You can feed in a sentence, document, PDF, image, or an old presentation…
OpenAI GPT-6.1 Terra goes live on OpenRouter at $1/M input, $5/M output
OpenRouter announced that OpenAI's GPT-6.1 Terra is now available. The model scores on par with GPT-6 Astra on Agents' Last Exam while costing roughly one-tenth…
Apple Creator Studio update: Final Cut Pro integrates MotionVFX, Keynote, Pages and Numbers gain smart assistant and MCP support
Apple has released an update for Creator Studio: Final Cut Pro for Mac now ships with the MotionVFX visual effects library and connects to a Content Hub stocked…
Apple launches 14-inch MacBook Pro with M6 chip, starting at $1,999
Apple has launched a 14-inch MacBook Pro powered by the all-new M6 chip, featuring a 12-core CPU and 12-core GPU. Graphics performance is up to 30% faster than …
Krea launches plugin for Adobe Premiere
Krea has released Krea for Adobe Premiere, letting users build storyboards, generate video and add sound effects without leaving the timeline. A trial is now av…
Unsloth releases quantized GGUF for MiniMax-M3.1-Flash, 2-bit version runs locally on 150GB of RAM
Unsloth has released quantized GGUF files for MiniMax-M3.1-Flash. The 2-bit version cuts size by 45% while retaining 91.5% top-1 accuracy and can run locally on…
Kling 4.0 Goes Live on Runware API with 30-Second 1080p Video and Native Audio
Runware announced that Kling 4.0 is available on its API from day one of launch, generating videos of up to 30 seconds in 1080p with native audio support. Featu…
Moonshot AI's Kimi K3.1 Pro Launches on OpenRouter as a 1M-Context Multimodal Open-Weight Model
OpenRouter announced that Moonshot AI's Kimi K3.1 Pro is now live, positioned as a 1M-context multimodal open-weight model post-trained from Kimi K3 and aimed a…
Kling 4.0 launches on fal with API access opened to all customers
fal announced that Kling 4.0 is now publicly available, with all customers able to access it via API. The model supports native video generation of up to 30 sec…
Replit launches Monitoring Agent, an on-call agent for apps that can automatically fix and deploy
Replit has launched Monitoring Agent, a monitoring agent that acts as an autonomous on-call engineer for applications. It can read production logs and metrics t…
Claude web, desktop, and mobile to enable auto mode by default
Claude officially announced that starting next week, sessions on the Pro, Max, and Team plans across the Claude web, desktop, and mobile apps will run in auto m…
OpenAI Officially Launches Decisions API for Low-Latency, Low-Cost Small Decisions
OpenAI announced the general availability of the Decisions API built on GPT-6 Luna, designed for bounded-answer problems such as classification, routing, and sc…
Kling 4.0 Goes Live on OpenArt with First, Middle and Last Frame Guidance and Up to 10 Reference Images
OpenArt has announced that Kling 4.0 is now live, letting users guide videos with first, middle and last frames, use up to 10 reference images, and generate fro…
Kling 4.0 Is Now Live on OnSolo
Kling announces that Kling 4.0 is now available on OnSolo. The model generates up to 30 seconds of native video in a single pass and stays stable through fast a…
OpenAI Releases Codex Skill to Help Websites Self-Check Exposure to Critical PHP Vulnerability
OpenAI has provided a Codex skill for the critical PHP vulnerability that guides users to run the github.com/openai/php2shell-checker repository to check whethe…
MiniMax Code Desktop Launches Mini App Long-Running App Feature
MiniMax has introduced Mini App in local sessions of the MiniMax Code desktop app (macOS / Windows) — a long-running application with both frontend and backend …
NVIDIA Showcases Full Local AI Hardware Lineup with RTX Spark, DGX Spark and DGX Station at GTC Berlin
At GTC Berlin, NVIDIA showcased its full local AI hardware lineup: RTX Spark targets personal agents on Windows PCs, DGX Spark targets Linux developers with sup…
NVIDIA at GTC Berlin announces Bolt, Volvo Cars and other European automakers building L3/L4 vehicles and robotaxi fleets on Hyperion
At GTC Berlin, NVIDIA announced that Bolt (in partnership with Lucid), Mercedes-Benz, Volkswagen Commercial Vehicles, Einride and other European automakers and …
OpenAI API now supports 12-hour prompt caching, with caches kept permanently for Codex and Work users
The OpenAI developer account announced that the API now offers optional 12-hour prompt caching, letting long-running agents reuse instructions, tools, and conte…
Ant Group and the vLLM Community Open-Source the Quantized Inference Kernel Library Humming
Ant Group and the vLLM community have open-sourced Humming, a general-purpose high-performance quantized inference kernel library, available at github.com/vllm-…
Anthropic opens Claude Duet research preview, with Claude taking over daily work as a proactive agent
Anthropic has opened the research preview of Claude Duet, a standalone app that lets Claude work as a proactive agent. It is now available on iOS, Android and c…
Perplexity Computer now runs fully locally on Mac, no cloud inference required
Perplexity CEO Aravind Srinivas announced that Perplexity Computer can now run entirely locally on a Mac, with no cloud inference and no WiFi needed. Both the m…
Google and Meta Release TorchTPU: A Native PyTorch Backend for TPU
Google and Meta have released TorchTPU, an open-source project that provides a native PyTorch backend for Google TPU. Migrating a model only requires changing .…
llm-d 1.0 released: Kubernetes-native distributed inference serving becomes production-ready
llm-d, the open-source inference project led by IBM, Red Hat, Google and others, released llm-d 1.0 at the PyTorch Conference North America — a Kubernetes-nativ…
DeepSeek Open-Sources Large-Scale Inference Framework DeepInfer Under Apache-2.0
DeepSeek has open-sourced DeepInfer, a high-performance large-scale inference framework, under the Apache-2.0 license at https://github.com/deepseek-ai/DeepInfe…
DeepSeek Fully Open-Sources Its Internal Inference Framework DeepInfer
DeepSeek announced that it is fully open-sourcing its internal inference framework DeepInfer, and is simultaneously open-sourcing the production-grade token cod…
Databricks Releases MemCurrent, a Production-Grade Persistent Memory Architecture for Agents
Databricks introduces MemCurrent, a persistent memory architecture for production agents that combines episodic memory with derived memory. Memories are read, w…
Inception's Diffusion Model Mercury 3 Goes Live on OpenRouter
Inception's diffusion model Mercury 3 is now live on OpenRouter, generating over 1,000 tok/s with 1M context support — 4x faster than Mercury 2.5. Inception rep…
Google releases TorchTPU-native inference engines for vLLM and SGLang
In a PyTorch-related announcement, Google introduced native PyTorch inference engines for vLLM and SGLang built on TorchTPU, developed in partnership with Infer…
Meta's open-weight model Muse Spark Open goes live on Fireworks, targeting long-context coding and agentic work
Meta's open-weight model Muse Spark Open is available for serverless calls on Fireworks from Day-0. It is an 816B-parameter MoE model with 42B parameters active…
Modal launches Sticky Sessions: requests in the same session always route to the same container
Modal has released Sticky Sessions. Built on Modal Servers, it pins every request in a given session to the same container; you can create one using the @modal.…
AWS releases TorchNeuron: a native PyTorch backend for Trainium
AWS announced TorchNeuron, an open-source native PyTorch backend for Trainium chips, available in public Beta under the Apache 2.0 license starting October 20, …
WorkBuddy Intl Adds Support for DeepSeek-V4.1-Flash-1017, Free Until October 31
WorkBuddy Intl has been updated to the latest DeepSeek-V4.1-Flash-1017 model, available free of charge until October 31. Quoted official DeepSeek content says t…
DeepSeek Opens Its Self-Developed Search API
DeepSeek has opened its self-developed search API, powered by the AI-native DeepSeek search engine, which reliably handles hundreds of millions of search reques…
NVIDIA RTX Spark laptops go on sale today, headlined by 128GB unified memory and local AI
NVIDIA has announced that RTX Spark laptops are now officially on sale from ASUS, Dell, HP, Lenovo, Microsoft and MSI, starting at $2,599.
Kling 4.0 Goes Live Late at Night: Author Tests 30-Second Single-Take and Omni Reference
KlingAI officially launched its video model Kling 4.0 in the early hours. After burning through tens of thousands of credits in hands-on testing, the author fin…
vLLM Delivers Day 0 Support for Reflection Beam 501B MoE Model
vLLM announced Day 0 support for Reflection's Beam model, a 501B-parameter MoE model with 23B active parameters per token, optimized on NVIDIA Blackwell.
SGLang natively supports Reflection's Beam and powers its production API
The SGLang team announced native support for Reflection's Beam 1.0 (a 501B-parameter MoE with 23B activated per token), and it now powers Reflection's productio…
Reflection's Open-Weight Model Beam Lands on Ollama, Free for a Limited Time
Reflection has released Beam, its first open-weight model. As a day-zero partner, Ollama is offering it free for a limited time on Ollama Cloud, with zero data …
Krea 3 image model launches on ComfyUI via Partner Nodes
The Krea 3 image model is now available on ComfyUI through Partner Nodes, focused on reference-based generation and editing. The model emphasizes faster iterati…
Claude Opus 5.5 fast mode leaves research preview and drops the waitlist
Anthropic announced that Opus 5.5 fast mode on the Claude platform is no longer in research preview, and the waitlist has been removed. The mode delivers up to …
ChatGPT Adds Tab Reading and Control for Chrome Users
ChatGPT has officially announced an upgrade for Chrome users, bringing its computer use capabilities to Chrome. In chat, ChatGPT can view the tabs a user has op…
Black Forest Labs FLUX 3 Series Lands on OpenRouter: Image Klein Goes Open-Weights, Video Upscale Gets Cheaper and Adds 4K Generation
OpenRouter announced three new offerings from Black Forest Labs (@bfl_ai). FLUX 3 Image Klein is a new open-weights image model that runs on a single consumer-g…
Meta's Open-Weight Model Muse Spark Open Lands on OpenRouter
Meta has released the open-weight model Muse Spark Open, now available on OpenRouter and positioned as the strongest US open-weight model for coding and agentic…
Perplexity Open-Sources SPACE, the Sandbox Platform Powering Perplexity Computer
Perplexity CEO Aravind Srinivas announced that SPACE, the sandbox platform underpinning all of Perplexity Computer's sandboxing capabilities, is now open source…
Kling 4.0 Opens to All Users, Adds MCP and CLI Access
Kling (Kling AI) has announced that Kling 4.0 is now fully available to all users, accessible through the new creation interface at kling.ai, with added MCP and…
ModelBest and Partners Jointly Open-Source PilotDeck 2.0, Focused on End-to-End Office Workflow Delivery and Stable Long-Task Operation
On October 15, the AI Agent workbench PilotDeck 2.0, jointly developed by Tsinghua University's THUNLP, ModelBest, OpenBMB, and AI9Stars, was officially open-so…
Payload integrates with Figma Make, enabling data-driven site building, hosting, and management inside Figma
Figma announced that open-source content framework Payload is now integrated into Figma Make, automatically provisioning a database, file storage, user login, a…
Black Forest Labs FLUX.3 Image Klein Lands on Diffusers with Bounding-Box Layout Control
The Hugging Face team introduced that the FLUX.3 Image Klein model is available in Diffusers from day one, with a single model supporting text-to-image, multi-r…
Google introduces the three-layer structure of the google/adk-recipes repository and its Agent Platform integration
The Google Developers Blog introduces the google/adk-recipes repository, a refactored successor to google/adk-samples that uses the three layers core/, plugins/…
Anthropic Code Review is now generally available for Team and Enterprise organizations
Anthropic announced that Claude's Code Review has exited research preview and is now generally available to Claude Team and Enterprise organizations.
Meta open-sources Muse Spark Open; Modal launches shared endpoints with token-based billing
Meta has released the open weights of Muse Spark Open, now available on Modal Auto Endpoints and billed per token through a Shared Endpoint.
Meta Muse Spark Open Launches on Together AI
Meta's open-weight Muse Spark Open model is available on Together AI starting today, with Together AI as a Day 0 launch partner. The model focuses on long-horiz…
vLLM Delivers Day 0 Support for Muse Spark Open
vLLM has partnered with Meta Superintelligence Labs to support the open-source MoE model Muse Spark Open on the very day of its release. The model has 816B tota…
SGLang Delivers Day-0 Support for 816B Open-Source Multimodal MoE Model Muse Spark Open
The SGLang team has announced Day-0 inference support for Meta's open-source multimodal MoE model Muse Spark Open (816B parameters, roughly 42B activated, 1M to…
Tessl Verifiers Goes Live, Checking Every Agent Code Change with Plain-Text Rules in the Repo
Tessl has released Verifiers: rules are stored as JSON files in the repository, and on every code change an LLM judge returns pass or fail for each one. It can …
Lovable launches growth tools that handle customer acquisition, ad campaigns and SEO right from the same chat window
Lovable has launched growth tools, bringing growth planning, ad delivery and SEO capabilities into its existing app-building chat window. Once users set a goal,…
Google releases agentic review system Deep Review, supporting technical document review across multiple domains
Google has publicly released Deep Review, an agentic review system that provides feedback and verification for complex technical documents. Initially focused on…
vLLM Introduces an Online-Fitted Acceptance Estimator Without a Confidence Head, Enabling Adaptive Verification for Draft Models Such as EAGLE and MTP
vLLM extends confidence scoring support to all draft-model-based speculative types: an online-fitted logistic regression derives a logit(q_max) feature from the…
xAI Releases Grok Voice 3.0 Speech Model
xAI has released Grok Voice 3.0, a speech-to-speech model that ranks first on Artificial Analysis's Speech-to-Speech Index with a score of 86.9 and sub-second l…
ComfyUI launches NVIDIA VFX node, supporting video super-resolution, frame filling and SDR conversion to HDR
NVIDIA's RTX video effects model is available in ComfyUI as a custom node. The model is distributed with the node and runs on the local RTX GPU.
Google DeepMind’s Gemini 4 Argon launched on OpenRouter
Gemini 4 Argon is now available on OpenRouter and can be called through Google AI Studio and Google Cloud. The model scored 51.3% in AutomationBench and 91.7% i…
Hesai Launches 3D World Data Engine Kosmo and 8K AI Spatial Camera Kosmo One
On October 14, Hesai Technology officially unveiled Kosmo, a 3D world data engine, together with Kosmo One, an AI spatial camera with an integrated 8K imaging s…
Apple launches 7-inch smart speaker HomeView, priced at $299 and powered by Siri AI
Apple released HomeView equipped with a 7-inch Liquid Retina display, integrating entertainment, FaceTime, home monitoring and voice automation. Both the standa…
LangSmith launches AI Search to search agent trajectories by behavior using natural language
LangChain launches AI Search public beta in LangSmith, allowing agent trajectories to be queried using natural language descriptions of behavior, such as PII le…
NVIDIA officially announced that the RTX Spark software ecosystem is ready, and notebooks will be launched on October 16
NVIDIA introduces the ecological readiness of the RTX Spark platform. The RTX Spark notebook will be available on October 16 (Friday), equipped with up to 128GB…
Qwen releases AI glasses N1/N1 Pro and the first AI ear clip-on headphones, now available for sale
Qwen released three new hardware products: AI glasses N1, N1 Pro and AI ear-clip headphones, which are now available for sale on Taobao, Tmall, JD.com, Douyin a…
Microsoft releases self-developed chip Maia 300 and plans to massively expand Maia inference deployment
Microsoft announced that starting next year, it will expand the Maia chip family to multi-gigawatt deployments, using it as the preferred inference system for i…
xAI launches The X Subscription integrating Grok, Grok Bot, Cursor, and X Premium
xAI launches The X Subscription, combining Grok, Grok Bot, Cursor, and X Premium into one plan, which will be available to users worldwide in the coming days.
GPT-6.1 Sol has increased the default context length in Codex and Work to 500k
Tibo announced that the default context length of GPT-6.1 Sol in Codex and Work has been increased from 272k to 500k, an increase of about 80%. This move can re…
Sierra releases draft of Personal Agent Protocol (Poppy), adding 35 new design partners
Sierra released a draft of the Personal Agent Protocol (Poppy) protocol and announced the addition of 35 new design partners, including OpenAI, Meta, Bank of Am…
Tibo announces the release of ChatGPT
Tibo announced the release of ChatGPT and gave the access address https://chatgpt.com/. The tweet itself only contains a release statement and link, with no fur…
Prime Intellect rewrites Prime Agent in Rust, and more than 2,000 agents autonomously complete end-to-end migration
Prime Intellect released Prime Agent rewritten from scratch in Rust. Since its launch, it has been downloaded more than 300,000 times and processed more than 8 …
ts-rust released: TypeScript compiler, checker, and LSP ported to Rust by LLM
ts-rust (tsc-rs) transplants the compiler, type checker and language server of Microsoft/TypeScript (Go implementation) to Rust. The author said that all the co…
Claude launches Dashboards real-time dashboard and Motion animation explanation function
Anthropic releases two beta features, Claude Dashboards and Claude Motion.
Google open source ML Drift end-side GPU inference engine, replacing TFLite GPU delegate
The Google AI Edge team has open sourced ML Drift, a cross-platform end-side GPU inference computing engine, under the Apache 2.0 license, which serves as the c…
Google's open source AQuA environment quality agent automatically diagnoses Agent faults in production environments
Google has released and open sourced AQuA (Ambient Quality Agent), which regularly extracts production sessions from Cloud Trace, Cloud Logging or BigQuery in G…
GPT-6 Luna Decisions Launched on OpenRouter
OpenRouter announces the launch of GPT-6 Luna Decisions. OpenAI's Decisions API allows applications to choose the appropriate model, tool, or action, supports s…
LangChain refactors Deep Agents' Skills support, adding tool binding, fixed skills, and thread-local overrides
LangChain has restructured Deep Agents' Skills support and introduced three updates for scenarios where enterprise skill libraries increase to thousands of skil…
NVIDIA and Microsoft release RTX Spark and DGX Station for Windows, promoting AI Agent implementation on Windows PCs
NVIDIA and Microsoft announced at the Windows AI and Surface event in San Francisco the introduction of AI Agent hardware and software for Windows PCs.
Cursor announces pricing for Claude Haiku 5.5 and lowers the cache read price for Claude Sonnet 5.5
Cursor announced that the pricing for Claude Haiku 5.5 is $0.10 per M input tokens and $0.50 per M output tokens, with a rate of $0.50/M and $2.50/M for input e…
Claude Code v2.1.293 Released: Added Claude Haiku 5.5 and Fixed Numerous Issues
Claude Code releases v2.1.293, adding Claude Haiku 5.5 (claude-haiku-5-5) as the default Haiku model for the Anthropic API, supporting 1M context, priced at $0.…
Anthropic launches monthly Platform API quotas for Claude Max and Team plans
Anthropic is launching monthly Claude Platform API quotas for Claude Max and Team plans: Max 5x for $100, Max 20x for $200, and Team up to $500 with sharing ava…
Anthropic has halved the cached read price for Claude Sonnet 5.5 to $0.10 per million tokens
Anthropic announced that it has halved the cache read price of Claude Sonnet 5.5, reducing it to $0.10 per million tokens. The company stated that this lowers t…
OpenAI launches GPT-6 and Intelligent UI to all ChatGPT users
OpenAI launched GPT-6 for a wider range of users, and with GPT-6, introduced Intelligent UI in ChatGPT, which can generate graphics, buttons, forms, charts, and…
Anthropic launches open source vulnerability scanning service OSS Scanner, free for open source projects
Anthropic releases OSS Scanner, an optional open source vulnerability scanning service that uses its strongest models (including Claude Mythos) to regularly sca…
Google launches Developer Knowledge API ecosystem, providing official document retrieval for AI agents
Google has launched the Developer Knowledge API ecosystem as the official programmatic source for developer documentation of Google Cloud, Firebase, Android, an…
Claude for Google Workspace has entered beta, allowing direct editing in Docs, Sheets, and Slides
Claude announces the launch of Claude for Google Workspace (beta), which can cover Google Docs, Sheets, and Slides with a single installation.
LlamaIndex Releases OpenDocRouter: An API for Unified Calling of Multiple Document Parsing Models
LlamaIndex launched the OpenDocRouter platform, which parses documents (PDF, PNG, JPEG, or URL) into markdown through a unified API, connecting to five cutting-…
Mistral Large 4 Launches on OpenRouter Public Beta: 1T Parameters, 512K Context
Mistral Large 4 is now available for public beta testing on OpenRouter, with specifications of 1T parameters (49B activated), native multimodal capability, 512K…
Google DeepMind releases multimodal embedding model EmbeddingGemma 2
Google DeepMind released the open-weight multimodal embedding model EmbeddingGemma 2 on October 6, which maps text, images, video frames, and audio into a unifi…
The Cursor iOS app supports remotely controlling local agents.
Cursor has launched local agent remote control in iOS apps, allowing users to view and reply to agents running on their computers. By default, it is enabled for…
Anthropic Cowork switches to running model inference and VM on the cloud
Anthropic engineer Felix Rieseberg explains the difference between the old and new architecture of Cowork: The old version infers in the cloud and runs local VM…
OpenAI launches new visual ad format and expands ad measurement tools in ChatGPT
OpenAI announced the launch of a new visual ad format in ChatGPT, which will be tested in image generation scenarios in the United States starting this month. T…
Together AI launches Together Link to connect open source models to existing coding agents with one click and reduce fees by more than 50%
Together AI released Together Link, which connects the coding agent tools already used by the team to the open source models on Together AI, claiming to save mo…
Prime Intellect releases inference platform Prime Inference, GLM-5.3 endpoint has been launched
Prime Intellect releases the inference platform Prime Inference, which provides serverless endpoints and reserved capacity, serves cutting-edge open source mode…
ChatGPT launches Finances financial management function
ChatGPT officially announced the launch of Finances financial management function, the entrance is http://chatgpt.com/finances. Functions include finding forgot…
Google releases next-generation federated learning system based on TEE, Gboard has deployed it
Google announced a next-generation federated learning system that leverages a Trusted Execution Environment (TEE) to provide fully verifiable and auditable data…
Google releases next-generation federated learning system based on TEE, Gboard has deployed it
Google announced a next-generation federated learning system that leverages a Trusted Execution Environment (TEE) to provide fully verifiable and auditable data…
NVIDIA DGX Spark launches 64GB version, available starting October 23 for $4,999
NVIDIA announced the launch of a new configuration of 64GB unified memory for DGX Spark, which will be available from Acer, ASUS, Dell, Gigabyte, HP and MSI on …
FLUX 3 Image is now available on Krea, supporting multiple rounds of editing and 4K generation
Krea has announced that FLUX 3 Image is now available on its platform. New capabilities include multiple rounds of editing without changing other pixels, layout…
FLUX 3 Image launches on OpenRouter, supporting native 4K generation and multi-reference editing
Black Forest Labs' FLUX 3 Image is now available on OpenRouter. It is the flagship image model that supports Vincentian graphics and multi-reference editing, an…
Suno launches Speech beta: integrated speech and background music generation
Suno launches Speech beta, calling it the first audio model that can generate speech and original background music as a complete track. Users can create by inpu…
Claude Code launches mods function to customize behavior and UI with TypeScript
Claude Code introduces mods function, which supports modifying model behavior, customizing UI and replacing own functions. Can be written with a few lines of Ty…
Modal Clusters officially released, providing multi-node GPU clusters via @modal.clustered
Modal announced that Modal Clusters are officially available. Multi-node clusters can be obtained through a decorator @modal.clustered. Communication between no…
ChatGPT can now build and deploy MCP servers directly
ChatGPT Sites can now host MCP servers. Users can build and deploy MCP servers directly in ChatGPT, and can also turn them into plug-ins and install them on web…
Modal launches Sidecars to provide low-latency trust boundaries for Sandboxes
Modal launches Sidecars (Beta), a trusted container that runs on the same host as the main Sandbox but is isolated to establish a security boundary between trus…
Modal releases multiple runtime product updates such as VM Sandboxes, Modal Clusters and more
Modal announced multiple product updates at its first Runtime conference. VM Sandboxes provide a complete Linux environment for Agent, already available in Line…
Modal officially releases VM Sandboxes: giving Agent a complete Linux virtual machine
Modal announced that VM Sandboxes is officially GA. Just set runtime="vm" to get a complete Linux VM that supports Docker, FUSE and kernel functions. It continu…
Perplexity opens Computer email delegation portal and runs tasks for free for a limited time
Perplexity opens Computer's email delegation function to everyone. No Perplexity account is required. Tasks forwarded or copied to computer@perplexity.com will …
Claude Code launches mods that can use TypeScript functions to rewrite prompt words and replace built-in functions
Anthropic launches mods for Claude Code, small TypeScript functions that hook into Claude Code’s event stream, rewrite prompt words, intercept or retry tool cal…
Manus 2.0 launches Game Dev function, allowing people without programming experience to make games
Manus introduced the Game Dev function after releasing Manus 2.0 on September 28, with the goal of allowing people without programming experience to use AI to c…
Google DeepMind releases SynthID Bio, embedding verifiable watermarks for AI-generated proteins
Google DeepMind released SynthID Bio on September 30, which introduces watermarking technology into synthetic biology, embedding invisible signatures into biolo…
Arena opens limited-time test of Claude Sonnet 5.5, Direct Mode available for 48 hours
Arena has announced the limited-time availability of Anthropic’s Claude Sonnet 5.5 (High) in Direct Mode until October 2 at 8am PT, and will remain available in…
Factory Automations is officially open: Droid can automatically execute engineering workflows on a scheduled or event basis
Factory announced that Automations is officially open to all users. After describing the workflow in natural language, Droid can be run according to schedule or…
OpenAI launches Codex cloud environment to reuse configurations and track tasks across devices
OpenAI launches Codex cloud environments, which can run Codex tasks in a reusable cloud environment. Warehouses, dependencies, scripts, and settings are already…
OpenAI DevDay 2026 releases a package of updates including Dots, GPT-6.1 Sol, and $500 subscriptions
The author summarizes the release of OpenAI DevDay 2026: the personal agent product Dots is launched to ChatGPT Pro, Business Premium and Enterprise users, supp…
OpenAI launches Codex cloud environment so agents can continue working after closing their laptops
OpenAI announced the launch of Codex cloud environments, allowing the agent to continue working even after users turn off their laptops. Reusable cloud environm…
OpenAI opens the ChatGPT platform and supports the use of plug-in extensions to build native applications
OpenAI announced an open platform that allows developers to build complete native applications with plug-in extensions and publish them directly within ChatGPT.…
ChatGPT subscription credits now available across 60+ partner products
ChatGPT subscriptions are now available directly for use in over 60 partner products, including usage for Devin, OpenCode, Lovable, and more. Users can use it b…
OpenAI releases resident agent dots, powered by GPT-6 Astra
OpenAI releases resident agents called dots, powered by GPT-6 Astra, with its own cloud computer that can continuously work for users 24/7 and connect to more t…
OpenAI launches new version of Codex Cloud, Agents API is open for preview and supports computer use
OpenAI has launched a significantly upgraded new version of Codex Cloud, which focuses on a configurable cloud environment. The author said that it is difficult…
ChatGPT launches Space team collaboration function and interactive document pages
ChatGPT officially announced the launch of Space as a new space for teams to collaborate with AI. New interactive document pages are available in Space, includi…
OpenAI launches a package of updates including ChatGPT plug-in extension, Space shared workspace and enterprise market on DevDay
OpenAI announced a series of ChatGPT updates on DevDay, including an open Plugin Extensions plug-in system, team shared workspace Space, document collaboration …
Microsoft Research releases Quine, the AI biological research system, and opens applications for the Quine Fellows program
Microsoft Research launches Quine, an AI research system that includes a model of the biological world and an interactive harness that jointly learns and reason…
OpenAI DevDay 2026 releases more than 20 updates including GPT-6.1 Sol and resident agent Dots
OpenAI made more than 20 announcements at DevDay 2026, including the release of GPT-6.1 Sol, which is strong for agentic coding, computer use, and professional …
Every actual test OpenAI DevDay 2026: 20 multiple releases and hands-on experience
Every reviews more than 20 products and features released at OpenAI DevDay 2026. After actual testing, the author believes that the Dots persistent agent has ch…
Artificial Analysis open source AA-AgentPerf-Local to test the inference performance of local AI agents on laptops and workstations
Artificial Analysis released the open source tool AA-AgentPerf-Local to test local inference performance by replaying 8 real agent tasks (168 rounds, the contex…
NVIDIA releases AI agent security platform, including Sentry system for real-time isolation of abnormal agents
NVIDIA releases an open AI agent security platform, including OpenShell security software and NVIDIA Sentry watchdog system. OpenShell can set boundaries for AI…
Meta launches Hologram immersive virtual image, Ray-Ban glasses and Quest headset will be launched in the fall
Meta has announced that it will launch the “Hologram” function on Ray-Ban Display smart glasses, Quest headsets and new thin and light headsets this fall. The f…
xAI releases Team Bots: Grok agents that can learn and work with teams
xAI launches Team Bots, allowing teams to share Grok Bots built around roles or workflows, integrating the four capabilities of Context, Plugins, Credentials, a…
NVIDIA releases open source runtime OpenShell to provide permission control and security sandbox for AI Agents
NVIDIA launches open source project NVIDIA OpenShell 0.1.0, a runtime for defining and enforcing AI Agent access permissions. It enforces permission controls ou…
NVIDIA releases open source intelligent agent security platform to provide chip-level continuous monitoring and isolation
NVIDIA launches NVIDIA Open Agent Safety Platform, which includes the open source runtime OpenShell, hardware layer Sentry and DOCA technology. The platform is …
Fireworks AI launches FireRouter with Opus, reducing coding task costs by 57%
Fireworks AI releases FireRouter with Opus, which can perform cache-aware routing between Claude Opus 5.5, GLM 5.3 and GLM 5.3 Flash, and is open as a serverles…
Claude opens the plug-in directory submission portal, and Plugins becomes the main way for third-party extensions
Anthropic announced that Plugins are the primary way to build third-party extensions for Claude. Plugins can package MCP connectors, Agent Skills, or both, and …
Satya Nadella announces Copilot’s biggest update yet, positioning it as a new OS for work
Satya Nadella announces Copilot’s biggest update yet, positioning it as a working new OS covering every model, device and task.
NVIDIA joins forces with Google DeepMind and other institutions to open a data set of predicted structures of protein complexes for more than 2,800 viruses
NVIDIA collaborates with global research institutions such as Google DeepMind and EMBL-EBI to openly release the predicted 3D structures of protein complexes of…
Claude Code clarifies that Cloud sessions are billed by subscription, with a one-time credit of $100 for Pro and $250 for Max.
The Claude Code team clarified that Cloud sessions, like other features of Claude Code, run within the Pro or Max subscription plan. This promotion is an option…
vLLM adds a distortion-free text watermark function based on Gumbel-max
vLLM announced support for distortion-free watermarking based on the Gumbel-max algorithm, integrating it into the sampling pipeline of Model Runner v2, and int…
Claude Code cloud session is officially launched, Pro and Max subscribers can receive a one-time credit
Claude Code cloud sessions are generally available, out of research preview, and can continue to run on Anthropic-hosted infrastructure after the laptop is clos…
Claude Marketplace is online to discover tools, agents and service partners
Claude has officially announced the launch of the Claude Marketplace, where users can discover tools, agents, and expert partners. Main uses include adding conn…
Anthropic launches Claude Marketplace, bringing together plug-ins, agents and service partners
Anthropic launches Claude Marketplace, which brings together plug-ins and connectors, agents and products, and service partners into one portal.
Greg Brockman announced a major upgrade of GPT Voice, supporting the use of tools and logging into ChatGPT Work
GPT Voice gets a major upgrade and now works with tools like Mailbox, Calendar, Slack, and more, and is powered by GPT-6 Astra, Sol, and Luna. The voice functio…
ChatGPT Voice upgrade: can call email, calendar, Slack plug-in and driven by GPT-6 Astra, Sol, Luna
OpenAI announced an upgrade to ChatGPT Voice, which can use plug-ins such as email, calendar, and Slack, and is powered by GPT-6 Astra, Sol, and Luna. Voice now…
OpenAI updates ChatGPT Voice: supports email, calendar, Slack plug-ins and is driven by GPT-6 Astra, Sol, Luna
OpenAI announced an update to ChatGPT Voice, which will be rolled out globally in the latest version of the application starting today. Voice functionality is n…
Antigravity SDK supports local models and can run agents completely offline
Google announced that Antigravity SDK supports local model workflow. It will first support Gemma 4 26B A4B through LiteRT of Google AI Edge. It can run the agen…
OpenAI opens Daybreak cyber defense program to Ukrainian government
OpenAI has announced the opening of its Daybreak program to the Ukrainian government to support cyber defense of civilian infrastructure, working with Ukraine’s…
Cursor releases Rollouts and Security Reviewer development bots
Cursor released two software development robots, Rollouts and Security Reviewer, to help teams send safe and reliable code to the production environment faster.
Qwen-Image-2.1 reaches the top of the Arena image editing and Wensheng diagram open source list
Qwen announced that Qwen-Image-2.1 is the first open source model in both Arena's Image Edit Arena and Text-to-Image Arena. The quoted content shows that its Im…
OpenAI launches improved prompt word caching system and diagnostic tools for GPT-6
OpenAI has launched an improved prompt word caching system for the GPT-6 series, which increases the cache hit rate by default and provides up to 90% cache inpu…
Claude Code v2.1.280 is released, adding Claude Opus 5.5 as the default Opus model
Claude Code releases v2.1.280, adding Claude Opus 5.5 (claude-opus-5-5) as the default Opus model, supporting 1M context, the price is $4/$20 per Mtok, and the …
Runway releases DaVinci Resolve plug-in to generate and redraw footage within editing software
Runway releases DaVinci Resolve plug-in that generates images and videos directly within Resolve Studio, redraws clips under the playhead, and imports the resul…
LiteParse September update: PDFium speed up by 20-25%, new visual positioning and is-complex routing API
LlamaIndex released the LiteParse 2.14.6 update. By optimizing the memory allocation of the self-maintained PDFium fork (built-in mimalloc) and other means, the…
Apple’s new Mac mini (M6/M5 Pro) and Mac Studio (M5 Max/M5 Ultra) go on sale today
Apple has announced that the new Mac mini and Mac Studio will go on sale on September 22. Mac mini equipped with M6 and M5 Pro, AI performance is improved by up…
Kimi releases browser extension, renamed from Kimi WebBridge
Kimi releases new Kimi Browser Extension, formerly known as Kimi WebBridge. Users can talk to Kimi in the sidebar of the browser and let it navigate web pages, …
Seedance 2.5 API launches Draft mode, 480P after trial and error, 1080P can be produced in one go
The Volcano Engine has launched a Draft (sample) mode for the Seedance 2.5 API. It first uses 480P to quickly generate samples to verify ideas. After selection,…
OpenRouter launches Batch API, offering half-price batch inference
OpenRouter publishes the Batch API, where asynchronous batch requests are completed by the provider within a 24-hour window, typically charged at 50% or less of…
transformers supports direct loading of GGUF quantization models, and local inference performance is close to llama.cpp
Hugging Face announced that transformers support directly running the GGUF quantization model. Passing in gguf_file through from_pretrained can load the GGUF ch…
vLLM releases vllm-metal v0.28.0: Support for concurrent inference services on Apple Silicon
vLLM officially released vllm-metal v0.28.0, which brings vLLM’s V1 scheduler, paged KV cache and OpenAI compatible server to Apple Silicon. The model is execut…
Kimi releases Kimi Code Desktop 1.0, macOS and Windows versions are launched simultaneously
Kimi (Moonshot AI) releases Kimi Code Desktop 1.0. As the official desktop client of Kimi Code, the macOS (Apple and Intel chips) and Windows versions are launc…
Anthropic merges Claude Cowork and chat into unified Claude, and launches Docs, Slides and more
Anthropic announced that Claude Cowork and chat have been merged into one Claude. Tasks no longer need to select entrances. Cowork and Design abilities can be u…
Claude Docs, Claude Slides and Claude Design are directly embedded in the conversation and can be exported to PowerPoint or PDF
Anthropic's Boris Cherny announces that Claude Docs, Claude Slides, and Claude Design have entered every conversation, generating demos, documents, and designs …
OpenAI launches new ChatGPT Ads feature: Sponsored Agents test and integrate HubSpot and Shopify
OpenAI launches several new AI-driven experiences for ChatGPT Ads. Sponsored Agents, which allow users to talk to clearly identified commercially sponsored agen…
Google releases multilingual AI achievements such as TranslateGemma, with language technology covering more than 300 languages
Google announced that its language technology has supported more than 300 languages, covering 86% of the world's population, and released TranslateGemma, a ligh…
Apple releases new generation of Apple Intelligence, Siri AI officially launched in beta version
Apple has released a new generation of Apple Intelligence, launching a completely reconstructed Siri AI that supports personal context understanding, screen awa…
OpenAI releases GPT-6 Astra and demonstrates community building cases
The OpenAI developer account announced the release of GPT-6 Astra and summarized the construction cases made by community developers based on it, including a 3D…
Google releases image tool Pics, based on Nano Banana to support precise editing and collaboration
Google releases image generation tool Google Pics, built on Nano Banana, now available at pics.new. Supports multiple options for local object editing, in-pictu…
OpenAI launches Data agent in ChatGPT Work
OpenAI launches a new Data agent in ChatGPT Work, which allows users to connect company data, analyze changes, and generate shareable interactive dashboards usi…
Cursor launches Projects: a coordinator agent that manages thousands of sub-agents to handle large-scale development tasks
Cursor releases Projects (beta), allowing users to handle large-scale work such as function development, migration, and continuous maintenance through the coord…
Hugging Face uses Gradio Workflow to rebuild Workflow1111 and reproduce the main functions of AUTOMATIC1111
Hugging Face released Workflow1111, using gr.Workflow to rebuild most of the functions of AUTOMATIC1111 with 73 nodes and 11 media pipelines, covering text gene…
OpenAI releases public beta version of Agents API
OpenAI launches a public beta version of Agents API, which will open the harness and infrastructure that drives Codex to developers through a single API call an…
OpenAI launches full-duplex speech model GPT-Live-1 in API
OpenAI releases the speech model GPT-Live-1 in the API, which can listen and speak at the same time. It supports delegation of inference and tool calls to back-…
Apple launches first foldable iPhone Duo, starting at $1,999
Apple released its first foldable screen iPhone Duo, which unfolds to a 7.6-inch inner screen and closes to a 5.4-inch outer screen, providing the largest displ…
Apple releases Health Sensing System and reconstructed Health app, Apple Watch Series 12 and Ultra 4 focus on readiness and Health Age
Apple has released new health and fitness features. Apple Watch Series 12 and Apple Watch Ultra 4 introduce the Health Sensing System, which supports heart rate…
Apple releases iPhone 18 Pro and iPhone 18 Pro Max
Apple releases iPhone 18 Pro and iPhone 18 Pro Max, equipped with a 48MP Fusion main camera with variable aperture, A20 Pro chip and new generation vapor chambe…
OpenAI launches GPT Image 2.5. The author actually measured the difference between the new version and image2.
OpenAI launches GPT Image 2.5, which features faster generation, higher definition, multi-round editing consistency, and regional annotation refinement. It also…
Meta intelligent product Muse is open for experience, and the official responds to users’ positive comments
Meta’s Muse intelligent product is open to more users for trial, and the entrance is https://muse.ai/join. Meta’s Chief AI Officer Alexandr Wang said that the t…
Runway releases Adobe plug-in for generating and editing directly within Premiere Pro and After Effects
Runway launches Runway Plugins, new panels that embed directly into Premiere Pro and After Effects to generate images and video, redraw clips, and place results…
OpenAI rolls out Astra to Plus, Pro, Business and Enterprise users
OpenAI announced that Astra is now generally available to Plus, Pro, Business, and Enterprise users in Codex and ChatGPT Work. Users can use it directly and wat…
Hugging Face releases open source tool funes to provide a locally held memory layer for coding agents
Hugging Face released the open source tool funes, which provides a local memory layer for coding agents such as Claude Code, Codex, pi, and Hermes, and indexes …
Claude supports background computer operation in Cowork and Claude Code
Claude officially announced that Claude Cowork and Claude Code have added the ability to use computers in the background. After users give tasks to Claude, it c…
Google Sheets launches Sheets canvas: Use Gemini to turn tabular data into interactive mini applications
Google Sheets has released a new feature, Sheets canvas, built on Gemini. Users can transform tabular data into interactive dashboards, learning trackers, seati…
NotebookLM short video overview is officially launched
The short video overview function has been officially launched on mobile and web pages for all English users.
Claude Cowork is open to mobile and web pages
Keep conversations and files synced across devices, with a beta version rolling out to Max users in the coming weeks. Cowork can complete complex tasks across f…
Rowboat: Open source, local-first desktop AI assistant
Index emails/meetings/Slack, etc. into an Obsidian style knowledge graph, providing persistent contextual memory; built-in email client, browser, meeting record…
