Why it matters: A major translation player entering lifelike voice translation raises the bar for speech-to-speech products.
DeepL unveiled Aure, which it calls its most lifelike voice AI translation model yet, aimed at real-time spoken translation. Details on availability and languages were not specified in the headline.
Alconost launched Localize.io, a localization service that uses AI agents for translation and localization workflows while keeping human review in the loop. It targets teams wanting automation without losing quality control.
Why it matters: A new leader atop a public vocal-music benchmark signals intensifying competition in AI music generation.
OrcaRouter says its ACE Music v1 model leads the AA-Music-Vocal leaderboard with 1,160 Elo, ahead of Suno v6. It is a music generation model competing on vocal quality.
Why it matters: Board and governance turmoil at the dominant AI chipmaker could affect the whole AI supply chain.
Governance experts say they have never seen a situation like the unfolding crisis at Nvidia, as a US envoy urges diligence on China and Fortune reports the board may rein in Jensen Huang rather than fire him.
Devia, a Zurich-based Visium spin-out, emerged from stealth with AI agents that investigate pharmaceutical quality incidents, establishing causes and documenting responses with precedent checks.
BiteSpeed released AI Creative Studio, a tool that generates marketing creatives for e-commerce brands, positioning itself as a Higgsfield-style creative engine for online sellers.
Why it matters: Pushes AI agents from finding flaws into autonomously fixing enterprise vulnerabilities.
Cogent pairs OpenAI models with enterprise context so agents can automatically remediate vulnerabilities at scale.
More than 20,000 US veterans and transitioning service members have claimed a free year of ChatGPT Plus for career preparation.
Why it matters: Full-duplex, sub-200ms latency is a key benchmark for natural real-time voice agents.
Fish Audio released Live 3, a full-duplex voice model for voice agents. It listens while it speaks, distinguishes backchannels like "mm-hmm" from real interruptions, and reacts in as little as 186 ms. English only at launch.
Why it matters: Signals how the world's largest messaging app plans to weave AI and payments into chats.
WhatsApp CEO Kunal Shah laid out three priorities for the app: simplifying it, adding more AI features, and expanding payments and richer conversations.
Softfreak developed F-Avata Shell, which uses AI to automatically generate vulnerability assessment scripts.
LloydK launched DO-GATE, a security gateway designed to govern and protect enterprise use of generative AI, controlling access and data flows to AI services.
Block released Buzz 1.0, a team chat platform with plugins for running Claude, Codex, Goose and Pi agents locally, labeled beta in its repository.
Why it matters: Anthropic moves Claude from assistant to full workspace, directly taking on Microsoft and Google productivity suites.
Anthropic expanded Claude with spreadsheet, digital whiteboard and data dashboard tools, plus the general availability of Claude Data. Claude can also open apps like Asana, Canva and Vercel inside the chat.
Korean art-tech startup yourNEST launched Rheina 2.0, an emotion-aware AI recommendation engine, plus the MeM emotion-recording service and an art-commerce store, guiding users from mood input to artwork and art-tour recommendations.
Why it matters: Alibaba's flagship multimodal model pushes million-token context at low subscription pricing.
Alibaba opened a preview of Qwen4-Max, its next flagship model, supporting image and video inputs and up to a million tokens of context. Developers access it via the Token Plan subscription from $3 a month; Qwen Code 1.0 preview was also open-sourced.
OpenAI will discontinue ChatGPT's Advanced and Standard voice options from December 11, moving users to the newer Live voice mode after GPT-Live usage nearly doubled.
Amazon Bedrock now offers Moonshot AI's Kimi K3.1 Pro Highspeed with a 1-million-token context window via global, US and India inference profiles; Moonshot says output is 2-3x faster than K3.1 Pro.
Why it matters: Signals US policy stance on AI competition and cooperation with China.
Treasury Secretary Scott Bessent said the US and China should cooperate and compete on AI without illusions, as the two countries navigate AI rivalry.
Australian Finance Group announced a new AI tool for its broker network intended to reduce administrative work in loan processing; no pricing or availability details were given.
Zum launched Zoe, a conversational AI assistant embedded in its CMX platform that lets district operators trace late routes, coverage gaps and pay anomalies in plain language.
Lexroom put its Lex legal AI assistant on WhatsApp, giving the messaging app's 2B users access to legal help through chat.
Why it matters: Adobe is embedding agentic AI into mainstream creative and marketing software.
Adobe announced AI agents and tools for creatives and deeper integration with Google and Microsoft, and will add a CX Enterprise Coworker to GenStudio content workflows.
The University of Michigan launched an AI app store for its community, Spaces4Learning reported. The store gives students and staff a curated place to find AI tools.
Groupon relaunched LivingSocial as an AI agent-first marketplace as its billings rose 6%. The relaunch puts shopping agents at the center of the local-deals service.
Fiserv's Clover added AI tools to help small businesses reach e-commerce shoppers, per PYMNTS AI.
A 14-year-old created Autark, a tool that lets AI agents band together like a team and act without constant prompting, though users must enforce their own security boundaries.
Why it matters: First state-level AI safety regime imposes registration and testing on frontier labs.
New York's sweeping AI safety law takes effect, requiring major developers like OpenAI and Anthropic to register, disclose safety practices and report critical incidents or face fines; the state will independently test models.
After a four-week pilot, Zillow Rentals is deploying Actively AI's proactive agents to more sales teams; the agents prioritize accounts and prepare daily actions while Zillow uses Actively's APIs.
A Show HN project offers tooling to get coding agents off ~2019 JavaScript and onto modern web primitives, helping agents write up-to-date browser code.
VictoriaStack is an open-source observability stack combining VictoriaMetrics' metrics, logs and distributed-tracing products with AI-assisted troubleshooting, letting teams expand telemetry without replacing their monitoring setup.
Why it matters: Adds pre-execution guardrails for agent actions in enterprise apps.
Obsidian Security's Runtime Policy Decision Engine controls what AI agents do inside business apps before actions execute, alongside an MCP Security Proxy and Policy Decision API for its agent-security platform.
A community mod for Claude Code visualizes what the coding agent is doing as a live map: the agent sits in the center with stations for files, shell, network, browser, subagents, skills and hooks, shown in a pane beside the chat.
Why it matters: Brings offline voice input and output to on-device developer tooling on Android and cross-platform.
Google's AI Edge Gallery now supports offline automatic speech recognition and text-to-speech, letting developers build on-device voice apps via ML Kit on Android or MediaPipe Tasks and LiteRT cross-platform.
An Arduino C++ library for the Texas Instruments TCAN4550 and TCAN4551 CAN-FD controllers, wrapping TI's low-level C driver in an Arduino interface for car and robot projects. Runs on real hardware.
Why it matters: A shared naming and identity layer could become foundational infrastructure for agent-to-agent interactions.
The Linux Foundation launched an Agent Name Service to give AI agents verifiable identities, part of broader efforts to anchor the agentic web on verifiable identity.
Why it matters: Reflects shifting investor sentiment between AI software and AI chip leaders.
Microsoft's market capitalization passed Nvidia's for the first time since June 2025, with Microsoft back above $4 trillion and roughly $266 billion behind Nvidia's peak standing.
Cloudbrink is pitching endpoint visibility as the way to spot unsanctioned Shadow AI use on corporate networks.
WellSaid Producer plugs AI voiceover production into Claude, ChatGPT, Copilot and other Model Context Protocol workflows, and generates audio natively at 48kHz for film and video.
ILION Digital launched an AI platform in the US that it says cuts ad technical audit time by 94%.
Why it matters: Intel is pitching a full CPU-plus-GPU stack at the fast-growing agentic inference market.
Intel is expanding its CPU and GPU portfolio with the GPU D-series (Crescent Island) and next-gen Xeon Diamond Rapids, targeting large-model inference and enterprise orchestration.
Why it matters: Open self-driving models lower the barrier for automakers building on Autoware.
TIER IV released five open-source end-to-end autonomous-driving AI models for Autoware, ranging from camera-only to LiDAR, radar and HD-map configurations, and adapted Autoware for a Munich bus.
Why it matters: Extends Intel's newest 18A node into mainstream AI PC silicon.
Intel's Core Series 3 processor, codenamed Wildcat Lake, targets mainstream consumer and commercial notebooks for agentic AI, combining Intel 18A and TSMC 6nm process technology.
STMicroelectronics launched the STM32H7P industrial microcontroller, pairing a 520MHz Arm Cortex-M7 with hardware accelerators for DSP and AI workloads.
Why it matters: CPUs are becoming central to running AI agents, not just training models.
AMD detailed a three-generation EPYC roadmap and released Zen 6 EPYC chips for denser AI agent deployments, as data-center workloads shift toward inference and agent orchestration.
Why it matters: A multibillion-dollar 5GW build shows Southeast Asia becoming a major AI infrastructure hub.
SoftBank Group and Grab signed joint development agreements with Sarawak state oil company PETROS for AI data centers in Malaysia, putting US$1.85B toward 5GW of capacity.
Why it matters: An open world model for physical AI gives robotics researchers a runnable simulation backbone.
PAN 2 is an open-weights 29B latent diffusion world model for physical AI, generating up to 20 seconds of 720p, 24fps video under an Apache 2.0 license.
Tianji Intelligence showed GENTO, a linkage-driven 21-degree-of-freedom dexterous robot hand, on static display at CoRL in Austin. It also claims Nvidia demonstrated GR00T 2 on a custom Marvin Pro platform.
MemryX announced the MX3+ inference accelerator, targeting up to 10x the peak AI compute and nearly 10x the on-chip memory of MX3 within the same low-power envelope.
Lenovo launched ThinkSystem servers with 6th Gen AMD EPYC SP7 processors, storage, interconnects and expanded AI factory services spanning customer data centers, hybrid environments and colocation.
Google's latest Pixel Drop lets Gemini make phone calls on users' behalf and brings Rambler dictation tech to Pixel 10 devices.
Why it matters: Two large cross-border AI hardware deals collapsing highlight growing tech decoupling and supply-chain risk.
Oracle cancelled a $7 billion AI chip deal with Tencent, while China blocked SoftBank's $5.4 billion takeover of ABB Robotics, raising supply-chain questions.
Why it matters: Puts reasoning models directly on self-driving hardware, a key step for physical AI.
Nvidia released Alpamayo 2 Edge weights for autonomous driving, aimed at running AV reasoning models within the compute and latency limits of onboard vehicle hardware instead of datacenters.
Why it matters: A landmark gas-power commitment shows frontier labs securing energy directly for compute.
Anthropic signed a $3.4 billion deal for gas-powered data centers to secure AI autonomy ahead of its IPO, even as voters in 10 states rejected data centers.
Why it matters: An open tactile foundation model could advance robotic manipulation and physical AI.
Robbyant, an embodied AI company under Ant Group, open-sourced LingBot-Touch, a tactile foundation model trained on a 5,000-hour dataset to give robots a shared sense of touch and better dexterity.
Why it matters: Signals that demand for rented GPU capacity remains strong despite broader AI-spending jitters.
CoreWeave topped quarterly revenue estimates and lifted its annual forecast as AI compute demand defies slowdown worries, and said it aims for profitability in 2028.
Mosaic Intelligence Labs raised $47M in seed funding to develop spatial intelligence sensing for surgical robots, emerging from stealth.
Why it matters: A frontier robotics model reaching near-production reliability on a real assembly line.
Google DeepMind detailed Gemini Robotics 2.5, a physical-intelligence model that sorted metal car parts with 96% success in a Mercedes-Benz assembly pilot after tackling 200 tasks.
Why it matters: Open Kubernetes tooling for agents and GPU workloads lowers AI infrastructure friction.
At KubeCon + CloudNativeCon, Nvidia donated OpenShell and released tools including an agent substrate for Kubernetes, a GPU workload snapshot feature, and NVIDIA Kubernetes Engine for managing GPU clusters.
Why it matters: A mainstream phone maker entering screen-free AI gadgets signals a new device category.
Nothing shared teasers of its first AI device, showing a metal body, LED lights and a large physical red record button with a retro look and no screen.
Why it matters: Aims to industrialize post-training for physical AI models.
Hedra's Model Factory is an autonomous platform to train, deploy and improve real-world AI models, cranking up production volume for physical AI post-training and inference.
Why it matters: Tests whether an AI-first OS can challenge Windows on emerging PC silicon.
KORA is a free Linux-based desktop OS built for AI models and agents, targeting users who rely on AI and bound for Qualcomm Snapdragon PCs as an alternative to Windows.
Marble, founded by former BitGo and Galaxy Digital executives, launched with an $11M Series Seed led by Polychain Capital. It builds private-credit infrastructure for AI compute operators alongside an institutional product.
Why it matters: Ring extends AI security cameras into autonomous home robots, a new physical AI form factor.
Ring Robotics debuts Ring Drone Cam and Ring Rover Cam, mobile proactive home security robots; the $250 drone patrols and responds to events.
Why it matters: Signs of strain in AI infrastructure financing could slow the datacenter buildout.
Carlyle's Thomas warns AI infrastructure financing carries risks as the Fed tightens, Nvidia-backed Nscale considers delaying a $35B listing after an AI sell-off, and Gulf sovereign funds are hit.
Amp is a frontier coding agent and development environment. This update ships as an npm-installable CLI, giving developers a terminal interface to the agent.
Alconost launched Localize.io, a localization service that uses AI agents for translation and localization workflows while keeping human review in the loop. It targets teams wanting automation without losing quality control.
VoiceInk is a voice-to-text app that transcribes speech into text. Version 2.27 is out for macOS, Windows and Linux.
Typeless is an AI voice dictation app that turns speech into polished text. Version 2.11.0 is available for macOS, Windows and Linux.
More than 20,000 US veterans and transitioning service members have claimed a free year of ChatGPT Plus for career preparation.
Version 26.51106.51428 of OpenAI's Codex VS Code extension, a coding agent that works inside the editor and is included in ChatGPT Plus, Pro, Business, Edu and Enterprise plans.
A Hugging Face Space runs an uncensored version of the Qwen-Image-2.1 image model for generating and editing images, using a LoRA on the official Diffusers weights with ZeroGPU.
DIGITIMES Intelligence argues local LLMs and AI agents are reshaping OS competition, with AI-enabled devices running mixture-of-experts models potentially favoring Linux over Windows.
Brain4Agent is a Python memory module designed to give mainstream AI agents long-term memory tuned for Chinese, Japanese and Korean text; the repo drew 100 stars in its first week.
GitHub documents the server configuration reference for its Managed Agent offering, giving developers a guide to setting up agent servers. It is documentation, not a new product release.
Block released Buzz 1.0, a team chat platform with plugins for running Claude, Codex, Goose and Pi agents locally, labeled beta in its repository.
lot-map is an open-source TypeScript tool that puts a live map beside the Claude Code chat, lighting up tool calls, model steps, hooks, skills and subagents as they happen. Runs locally in five languages.
Korean art-tech startup yourNEST launched Rheina 2.0, an emotion-aware AI recommendation engine, plus the MeM emotion-recording service and an art-commerce store, guiding users from mood input to artwork and art-tour recommendations.
A guide to using Google Maps inside AI Studio: what the no-credit-card sandbox allows and what changes, including billing and quotas, when you move to production.
Zum launched Zoe, a conversational AI assistant embedded in its CMX platform that lets district operators trace late routes, coverage gaps and pay anomalies in plain language.
GitHub Copilot app is a native desktop client for GitHub Copilot. Version 1.2.0 is out for macOS, Windows and Linux.
pydantic-ai is a Python framework for building AI agents, realtime voice, image generation and embeddings with typed interfaces across models. Version 2.67.0 adds SubAgents support under durable execution and input sanitization fixes.
Cursor rolled out a beta multiplayer upgrade to Projects, its system for directing fleets of coding agents; teams can share a workspace where people and agents coordinate tasks, context and handoffs.
Lexroom put its Lex legal AI assistant on WhatsApp, giving the messaging app's 2B users access to legal help through chat.
Why it matters: Adobe is embedding agentic AI into mainstream creative and marketing software.
Adobe announced AI agents and tools for creatives and deeper integration with Google and Microsoft, and will add a CX Enterprise Coworker to GenStudio content workflows.
The University of Michigan launched an AI app store for its community, Spaces4Learning reported. The store gives students and staff a curated place to find AI tools.
Wispr Flow released version 1.6.1457 of its voice-to-text dictation app with AI auto-editing, for macOS, Windows and Linux.
Gemini CLI, Google's open-source terminal agent that brings Gemini into the command line, hit v0.68.0 with assorted fixes and changelog items. The same round includes Codex CLI 0.167.1 and litellm 1.109.0.
An open-source Python project that turns a book into an Agent Skill, using a dual-engine pipeline, pluggable genre plugins and word-level fidelity verification.
Version 1.59.0 of Goose, an open-source extensible AI agent that goes beyond code suggestions, is out for macOS, Windows and Linux.
Andreas Spiess built Embedded AI Harness, a hardware-in-the-loop framework where an AI writes ESP32 firmware and then tests it on real hardware through linked test scripts.
Darling, an open-source translation layer that runs macOS binaries on Linux, is now using AI in its development work, Phoronix reports.
Latent Space interview with Resend's Zeno Rocha covers answer engine optimization versus agent experience, including docs that help Claude Code pick an email API and tools agents use to send.
Granola is an AI-powered notepad for meetings that transcribes and summarizes notes. Version 7.774.2 is a minor update.
Perplexity is an AI answer engine with a Personal Computer agent. Version 26.44.0 is a minor update for macOS, Windows and Linux.
A Hugging Face Space runs an uncensored version of the Qwen-Image-2.1 image model for generating and editing images, using a LoRA on the official Diffusers weights with ZeroGPU.
H3-Image is an open-weight text-to-image and editing model from MiniMax, capable of 16-megapixel/4K images and run in six steps; weights are on Hugging Face.
Propaganda_Eval is a Hugging Face dataset of 803 prompts on topics China censors, each with required facts; answers are scored on factuality and adoption of state narratives.
Mirage is an Apache-2.0 image-text-to-text model that describes scenes, detects objects, reads text, segments images and extracts structured data. Weights are on Hugging Face with a Python package.
Arete is a Hugging Face dataset of prompt-response pairs on seven values topics, each answered by an Inkling teacher model and the Qwen 3.8-27B checkpoint, with category tags and verification hashes.
GGUF quantized versions of the Gemma-4-31B-it-ARA-heretic model, for running locally with tools like llama.cpp. Static and imatrix quants are available.
A new paper introduces MMDM, a benchmark measuring massive multitask decision making for System-1 decision models that make structured probabilistic decisions in a single forward pass, to test generalization.
A new paper introduces DexForesight, a framework that learns spatial foresight implicitly by injecting dynamics-aware world representations into the latent space of a dexterous manipulation policy instead of predicting futures explicitly.
LongCat-2.5 is a 1.6-trillion-parameter multimodal AI model released by Meituan on Hugging Face, available for download and use.
Brain4Agent is a Python memory module designed to give mainstream AI agents long-term memory tuned for Chinese, Japanese and Korean text; the repo drew 100 stars in its first week.
lot-map is an open-source TypeScript tool that puts a live map beside the Claude Code chat, lighting up tool calls, model steps, hooks, skills and subagents as they happen. Runs locally in five languages.
pydantic-ai is a Python framework for building AI agents, realtime voice, image generation and embeddings with typed interfaces across models. Version 2.67.0 adds SubAgents support under durable execution and input sanitization fixes.
Gemini CLI, Google's open-source terminal agent that brings Gemini into the command line, hit v0.68.0 with assorted fixes and changelog items. The same round includes Codex CLI 0.167.1 and litellm 1.109.0.
An open-source Python project that turns a book into an Agent Skill, using a dual-engine pipeline, pluggable genre plugins and word-level fidelity verification.
ComfyUI, an open-source node-graph GUI, API and backend for running diffusion models locally, released v0.45.0 with updated inference features.
A TypeScript MCP server that gives Claude, Cursor, VS Code and Codex read-only access to Yandex Metrika analytics: reports, period comparisons, time series and exports.
An open-source TypeScript collection of Claude-style skills that run in the terminal: company-specific mock interviews, online assessments with hidden tests and resume critiques covering more than 470 companies.
numbrone is an open-source shell tool that acts as a chief of staff for coding agents, routing requests to per-project manager agents that run parallel workers in git worktrees.
Goose, an open-source extensible AI agent that installs, executes, edits and tests with any LLM, releases v1.58.0 adding context relevance, a context window viewer and new providers.
Jarvis is an autonomous, local-first on-device AI agent framework for Android that uses full accessibility control and continuous voice to automate workflows on the phone. It hit 100 stars in its first week.
A free Python workbench for AI security operations that handles alert triage, incident response and vulnerability remediation, connecting Atlas, models, tools and workflows. It drew 100 GitHub stars in its first week.
roblox-studio-mcp is an open-source Model Context Protocol server that lets AI assistants like Claude, Cursor and Cline read Roblox Studio object trees, automate Lua scripting and run playtests safely. It gained 279 stars in its first week.
VoiceInk is a voice-to-text app that transcribes speech into text. Version 2.27 is out for macOS, Windows and Linux.
Typeless is an AI voice dictation app that turns speech into polished text. Version 2.11.0 is available for macOS, Windows and Linux.
More than 20,000 US veterans and transitioning service members have claimed a free year of ChatGPT Plus for career preparation.
Korean art-tech startup yourNEST launched Rheina 2.0, an emotion-aware AI recommendation engine, plus the MeM emotion-recording service and an art-commerce store, guiding users from mood input to artwork and art-tour recommendations.
Zum launched Zoe, a conversational AI assistant embedded in its CMX platform that lets district operators trace late routes, coverage gaps and pay anomalies in plain language.
Lexroom put its Lex legal AI assistant on WhatsApp, giving the messaging app's 2B users access to legal help through chat.
Adobe announced AI agents and tools for creatives and deeper integration with Google and Microsoft, and will add a CX Enterprise Coworker to GenStudio content workflows.
The University of Michigan launched an AI app store for its community, Spaces4Learning reported. The store gives students and staff a curated place to find AI tools.
Wispr Flow released version 1.6.1457 of its voice-to-text dictation app with AI auto-editing, for macOS, Windows and Linux.
Granola is an AI-powered notepad for meetings that transcribes and summarizes notes. Version 7.774.2 is a minor update.
Perplexity is an AI answer engine with a Personal Computer agent. Version 26.44.0 is a minor update for macOS, Windows and Linux.
Wispr Flow is an AI voice keyboard and speech-to-text app that turns talk into polished writing. Version 2.0.0 for iPhone and iPad tunes the experience for smoother dictation.
Leaker Aaron Perris found evidence Apple is building a Mac app called Harbor that makes it easy to run local AI models across one or more Macs; Claris is reportedly testing it.
Google's Gemini now displays everything it remembers about a user and lets them delete it. The Information also details using Gemini to assist with RFP responses.
HomeRoll is a native iOS 26 app that looks and behaves like Apple Photos but stores photos on a user's own Immich server, built with Claude to replace iCloud Photos.
A developer built a web page that uses a webcam to show a skull, brain and muscles tracking the user's face and hands in real time, made with Claude Fable 5.2.
BoltAI is a desktop AI chat client for macOS, Windows and Linux; version 3.1.0 is now available.
Blender 5.3 is out with Nvidia DLSS 4.5 Ray Reconstruction support and Vulkan as the default backend. DLSS 4.5 adds new presets users should know about.
Hostinger Credits gives users one balance usable across AI tools, even outside Hostinger's own platform, the company announced.
Cherry Studio is a desktop client that connects to multiple LLM providers in one app. Version 2.2.0 is out for macOS, Windows and Linux.
Siri AI is now available to all developers for full app integration in the latest iOS betas, alongside a Siri camera upgrade rolling out for iPad ahead of public release.
VOYA is an AI travel concierge letting eligible Club Wyndham owners go from describing a vacation to booking an eligible stay in one conversation, available in the app and website.
Nod is a free, open-source personal AI agent that runs in a private sandbox on your own computer, with its own browser, and can carry out tasks in your apps and on websites. Research preview from n8n Labs.
OpenAI Academy is offering free, hands-on ChatGPT workshops to caregivers in five U.S. cities to help them build practical AI skills for everyday life.
Claude for Mac is Anthropic's official Claude AI desktop app. Versions 4.23118.1 and 5.1231.0/5.1231.2 are available for macOS, Windows and Linux.
Amp is a frontier coding agent and development environment. This update ships as an npm-installable CLI, giving developers a terminal interface to the agent.
Alconost launched Localize.io, a localization service that uses AI agents for translation and localization workflows while keeping human review in the loop. It targets teams wanting automation without losing quality control.
Version 26.51106.51428 of OpenAI's Codex VS Code extension, a coding agent that works inside the editor and is included in ChatGPT Plus, Pro, Business, Edu and Enterprise plans.
DIGITIMES Intelligence argues local LLMs and AI agents are reshaping OS competition, with AI-enabled devices running mixture-of-experts models potentially favoring Linux over Windows.
GitHub documents the server configuration reference for its Managed Agent offering, giving developers a guide to setting up agent servers. It is documentation, not a new product release.
Block released Buzz 1.0, a team chat platform with plugins for running Claude, Codex, Goose and Pi agents locally, labeled beta in its repository.
A guide to using Google Maps inside AI Studio: what the no-credit-card sandbox allows and what changes, including billing and quotas, when you move to production.
GitHub Copilot app is a native desktop client for GitHub Copilot. Version 1.2.0 is out for macOS, Windows and Linux.
Cursor rolled out a beta multiplayer upgrade to Projects, its system for directing fleets of coding agents; teams can share a workspace where people and agents coordinate tasks, context and handoffs.
Gemini CLI, Google's open-source terminal agent that brings Gemini into the command line, hit v0.68.0 with assorted fixes and changelog items. The same round includes Codex CLI 0.167.1 and litellm 1.109.0.
Version 1.59.0 of Goose, an open-source extensible AI agent that goes beyond code suggestions, is out for macOS, Windows and Linux.
Andreas Spiess built Embedded AI Harness, a hardware-in-the-loop framework where an AI writes ESP32 firmware and then tests it on real hardware through linked test scripts.
Darling, an open-source translation layer that runs macOS binaries on Linux, is now using AI in its development work, Phoronix reports.
Latent Space interview with Resend's Zeno Rocha covers answer engine optimization versus agent experience, including docs that help Claude Code pick an email API and tools agents use to send.
Databricks says Unity Catalog Metric Views are now available in Power BI, giving organizations one governed semantic layer so reporting and AI share a single trusted definition of business metrics like revenue.
A 14-year-old created Autark, a tool that lets AI agents band together like a team and act without constant prompting, though users must enforce their own security boundaries.
Microsoft Research detailed a major upgrade to Copilot in Visual Studio; the source gives no concrete version, feature list or availability details.
A Show HN project offers tooling to get coding agents off ~2019 JavaScript and onto modern web primitives, helping agents write up-to-date browser code.
VictoriaStack is an open-source observability stack combining VictoriaMetrics' metrics, logs and distributed-tracing products with AI-assisted troubleshooting, letting teams expand telemetry without replacing their monitoring setup.
A community mod for Claude Code visualizes what the coding agent is doing as a live map: the agent sits in the center with stations for files, shell, network, browser, subagents, skills and hooks, shown in a pane beside the chat.
Chatbox, a desktop app for chatting with GPT-4 and GPT-3.5 via the OpenAI API, updated to 1.24.3 on macOS, Windows and Linux. Kiro released version 1.4.154.
An Arduino C++ library for the Texas Instruments TCAN4550 and TCAN4551 CAN-FD controllers, wrapping TI's low-level C driver in an Arduino interface for car and robot projects. Runs on real hardware.
AWS Machine Learning · 15h
#
AWS outlined a centralized observability data lake aggregating Amazon Bedrock invocation logs across accounts to see prompts, token counts and model choice, plus an agent evaluation playbook.
RainFocus launched an Anthropic connector letting AI agents access event data.
OpenCode Desktop is a desktop client for an AI coding agent. Version 2.2.7 is available for macOS, Windows and Linux.
