Why it matters: Major labs joining a state biosecurity lab signals coordination on AI bio-risk.
The UK announced £100 million ($134 million) for an AI Biosecurity Lab supported by OpenAI, Anthropic and Google DeepMind, among measures to tackle AI risks.
FinBursa 360 is an AI-native platform that integrates investment management workflows, deal execution, fundraising and investor relations for private markets firms. FinBursa was founded in 2024 by ex-BlackRock director Ismail Badereldine.
Simon Willison shares a Claude prompt that builds an artifact where you open a photo and it quickly paints a version using a canvas element and simulated paintbrush.
Tairex announced a clinical AI framework that lets doctors coordinate specialist agents, with three doctor-agent workflows and planned staged tests at partner hospitals.
Why it matters: Moves a major design platform onto its own image model instead of third-party ones.
Canva released Craft, its first AI image model built in-house; BetaNews reports it ranked fourth of 20 models in a comparison.
Why it matters: Training voice models only on real conversations could improve naturalness over read-speech datasets.
Jack Dorsey's Block released Dessa, a voice AI model trained solely on conversational speech rather than read text, aimed at sounding more natural and human-like.
Why it matters: Consolidates Anthropic's enterprise AI offerings into one product tier.
Anthropic is combining Claude Chat and Cowork into a single offering for enterprise plans, simplifying how businesses access its assistant and collaborative agent features.
TD Bank U.S. launched Credit Assist, an AI-powered tool it says cuts the time needed for commercial underwriting.
Microsoft plans to add AI-powered, context-aware phishing detection to Teams messages starting in early November 2026, targeting malicious links and messages.
Why it matters: Expands Grok's enterprise ecosystem through a large partner network.
SpaceX has launched the SpaceX Super Intelligence Partner Network, enlisting 130 tech leaders to help businesses deploy its Grok AI model. The program offers training, tools and support for companies building products and services with Grok.
Tellia is a voice AI assistant for agronomists and growers; it now opens to independent users at $20 a month.
Napier AI released an index tracking money laundering activity in the UK, using AI to surface financial-crime patterns.
iProov launches a biometric human authorization service built on the Human Approval and Presence Specification (HAPS), giving enterprises a way to verify a human approves an AI agent's high-risk actions.
BILL launched agentic AI capabilities aimed at transforming financial operations for small and midsize businesses.
ShopOS launches Sloosh.ai, an AI creative platform that generates product images and videos for e-commerce sellers.
VoxNexus offers on-device voice artificial intelligence supporting over 100 languages without internet, aimed at affordable Android devices.
Tipalti is pitching an AI agent that handles the full accounts payable workflow for finance teams, according to Accounting Today.
AX1's Console 2.0 is a platform where AI agents perform onchain tasks and earn real USDC payments, with weekly prizes.
Cloud Wars reports Industry Copilot now ships pre-built legal skills and connectors to third-party data sources, extending its AI assistant into legal workflows.
ZoomInfo has integrated Exa's search API so AI agents can pull web search results alongside ZoomInfo's company and contact data.
Why it matters: Large AI-drug-development round shows continued investor appetite in AI biotech.
AI drug development startup Pathos AI closed a $237M investment and is teaming up with Bristol Myers Squibb and Tempus AI to advance drug development.
Evervault's Browser Tokens let AI agents browse and act in a user's browser session without exposing credentials or sensitive data, aimed at securing agentic workflows.
Girikon.AI built Koyal, a voice AI model aimed at natural, multilingual conversations, targeting more fluid spoken interactions across languages.
Chargeflow released Prevent 2.0, an AI model for Shopify merchants aimed at preventing chargebacks and payment disputes before they happen.
A developer released CreatorOS Studio, open ComfyUI workflows that recreate hosted AI video one-click reels, 25 camera moves and re-shoots using open-weights models on a local GPU, with no credits or subscription.
Why it matters: Open-sourcing AI-for-AI-research models could accelerate automated model development.
Recursive, valued at $4.65B, released its first open-source model: an 'AI-building-AI' sample reportedly made in four months, aimed at AI research automation.
Why it matters: Walmart's guardrails show mainstream retail adopting agentic commerce.
Walmart unveiled controls for AI shopping agents, including spending limits, letting chatbots shop and buy for users as retailers rethink the purchase decision.
Indian AI lab Chariot says its new speech model outperforms Google, ElevenLabs and global rivals across the 10 most spoken Indian languages.
Hong Kong Exchanges and Clearing launched the HKEX Artificial Intelligence Index and the HKEX Semiconductor Index on Friday.
Why it matters: A major Chinese lab's new frontier model keeps pressure on global model competition.
Moonshot AI released Kimi K3.1 Pro, its first upgrade since its DeepSeek moment, with reporting pointing to 3D code generation focus; the Kimi app ships as 3.2.20.
Why it matters: Extends agentic AI payments from consumer checkout into B2B invoice settlement.
Visa announced that its AI agents can now pay business invoices, extending agentic payments into business-to-business bill settlement.
Why it matters: Oracle's agent platform adds another enterprise vendor pushing agents into core business software.
Oracle launched an AI agent platform, AD HOC NEWS reports, while Mizuho maintained its $320 price target on Oracle stock.
AudioStack has integrated ElevenLabs' v4 voice models, bringing the latest speech generation to its audio production platform across more voices and markets.
Why it matters: Signals Anthropic pushing into services and enterprise deployment in Europe, a new front beyond model sales.
Anthropic and private-equity firm Hg are forming a company to embed AI engineers inside European enterprises, according to reports, with Financial Times describing it as a consultancy launch.
Functionize released an MCP server that lets AI coding agents connect to its testing platform to create and run automated tests.
LitPivot is a tool that helps researchers refine early research ideas through literature-supported critiques and revisions as proposals evolve. It is available via Hugging Face.
Mistral updated its speech models and added multilingual voice application tools; it also released Timestral, an AI that forecasts numbers without a GPU, and Mistral Large 4 licensing drew scrutiny.
MOVO AI is a long-form video agent that plans, creates and refines videos with consistent characters and scenes from your phone or the web, running in the background.
Ovren lets you describe a change in plain words; it writes code on a safe copy of your project and shows a live preview of frontend changes for approval and merge.
Agent Hunt is a live leaderboard ranking tools AI agents actually use, with positions earned by real agent calls made through Monid.
Alloop launched Trio, a modular three-in-one wearable built around its personal AI agent. The device combines multiple form factors and is positioned around an onboard assistant.
A basic flip phone for the UK market with a built-in AI companion called Mary, building on the HMD 2660 Flip launched at MWC 2025.
Striding AI presented a physical AI solution for round-the-clock retail at APRCE 2026, targeting automated retail operations using embodied AI systems.
Chery has launched sales of humanoid robots intended for home use, moving the automaker into consumer robotics alongside its vehicle business.
Procore is integrating Meta's AI smart glasses into its construction workflows, letting site workers use wearable AI on the job.
Agile ONE SN is a humanoid robot that bundles Nvidia's robotics stack, combining the hardware with Nvidia's software and compute tools for physical AI development.
Metro Redux's next-gen update includes Nvidia DLSS 5 support, using neural rendering to modernize old characters. Nvidia says DLSS 5 is now twice as fast.
Soulog's Sense is a compact AI wearable that captures real-world context, stores it in a persistent personal memory system and generates usable outputs, positioned as a pocket AI agent rather than a plain recorder.
Why it matters: Shows AI compute buildout spreading to Central Asia alongside robotaxi plans.
Kazakhstan is expanding accelerated computing across its AI stack, with Nvidia detailing GPU deployments. Separately, inDrive plans 2027 robotaxi tests in Central Asia with Pony.ai and Nvidia. Huang cites a $10B Kazakhstan AI push.
Getac has launched new notebooks designed to provide edge AI processing power for field and industrial use.
Why it matters: Shows foundries stretching mature nodes to serve edge AI and automotive demand.
Samsung's foundry business will launch a second-generation 8LPU+ platform in 2026, extending its eight-year-old 8nm process to target automotive electronics, robots and edge AI chips.
RoboParty RP1 (ROBOTO 01) is a full-stack open-source humanoid robot with up to 30 degrees of freedom; the team released full PCB files, though Chinese data laws apply.
Aptiv expanded its partnership with Nvidia to offer an air-cooled Thor platform for automotive AI systems.
Why it matters: Open chiplet standard could ease multi-die AI chip design across vendors.
Tenstorrent released the beta of its Open Chiplet Atlas Harness, an open-source framework turning its chiplet interoperability spec into reusable implementations, and raised a $1.5 billion Series E round.
AGIBOT released a dataset aimed at teaching robots to decide what a situation needs before picking their next move, tackling tasks like restocking a minibar where grasping succeeds but ordering steps fail.
Why it matters: Robot import restrictions reshape the humanoid supply chain and US competitiveness.
IDC says the FCC robot ban cuts US humanoid robot revenue 58% and sends Chinese robots to Europe. The FCC granted Intel a waiver on foreign-made robotic devices to prevent chip disruptions.
Lenovo's new Yoga laptops are aimed at AI creators, pairing Nvidia RTX Spark graphics with OLED displays, per NDTV Profit.
Why it matters: Adds a lighter, production-targeted European humanoid and a dedicated training gym to the physical AI race.
Unicorn NEURA Robotics launched its 4NE1 Gen 4 humanoid, 20% lighter and heading toward mass production, alongside a €17M TUM RoboGym training facility in Munich for European robotics.
HOHOTIME is crowdfunding an all-in-one AI strength-training station with four independently adjustable resistance arms, AI coaching and a library of 860 exercises for home gyms.
OPPO launched the Find X10 series with triple 200MP cameras, agentic AI features and bigger batteries, alongside a new tablet and wireless earbuds for its connected ecosystem.
Why it matters: Airport service marks another expansion of driverless ride-hailing into high-traffic hubs.
Waymo robotaxis started pickup and drop-off service at Las Vegas airport. Elsewhere, HTC launched Vivee, an egg-shaped personal AI companion, and researchers showed a wearable AI device for emergency allergy treatment.
Why it matters: Investor nerves over AI buildout debt are starting to hit the neocloud companies renting out GPUs.
Short sellers have pocketed about $6 billion in unrealized profits as neocloud stocks fell on concerns about AI-related debt. Zadara also won Nvidia AI Cloud Ready validation, and Crusoe planned an AI cloud deployment at Paraguay's Itaipú Dam.
Why it matters: Shows concrete industrial humanoid deployments taking shape in South Korea.
South Korean robot makers take different routes into industrial humanoids: Tesollo builds dexterous hands, AeiROBOT prepares bipedal systems for shipyards, and HD Hyundai Robotics develops a dual-arm platform; Nvidia's Microduck robots appeared at GTC Berlin.
Gasgoo Daily reports AiMOGA unveiled its Mobo humanoid robot, alongside news that Audi named Jeff Zhang its first Chinese president and BAIC and CATL deepened a battery-swap partnership.
qBraid unveiled qBraid OS and launched hosted GH200 GPU simulators at Nvidia GTC Berlin for quantum computing development.
Antigravity is an agent orchestration platform, now at version 2.26.0, available for macOS, Windows and Linux.
The Gemini Android app now has a redesigned side panel and simpler settings, following the Spark-to-Task update.
Frustrated by intrusive ads, slow interfaces and tracking, smart TV owners are using AI coding tools and open-source projects to remove unwanted software and build custom home screens, The Verge reports.
Purdue University Northwest is providing first-year students with Google Gemini licenses to support AI literacy instruction.
Simon Willison shares a Claude prompt that builds an artifact where you open a photo and it quickly paints a version using a canvas element and simulated paintbrush.
Codex is OpenAI's coding agent that works alongside you in your editor; the VS Code extension is included in ChatGPT Plus, Pro, Business, Edu and Enterprise plans.
Wispr Flow is voice-to-text dictation software with AI-powered auto-editing; version 1.6.1253 is available for macOS, Windows and Linux.
Granola is an AI-powered notepad for meetings. Version 7.668.3 is out for macOS, Windows and Linux; no further changes were detailed in the release listing.
A dataset of 1,342 selected AI-generated 3D examples with about 1,500 output files (~39 GB), including previews and source inputs, released under CC BY 4.0 for training and research.
Devin Desktop is an agentic IDE with an AI agent command center. Version 3.11.212 is available for macOS, Windows and Linux.
The official TypeScript SDK for building Model Context Protocol servers and clients gets version 2.4.0, updating its client, server, core, legacy and codemod packages.
Nvidia's developer blog explains how the Global Performance Tuner scales TensorRT optimization beyond a single build route, treating deployment as a compilation problem with many valid answers.
AWS published a guide on orchestrating SageMaker AI endpoint auto scaling with AWS CDK and a hands-on guide to multi-turn reinforcement learning on SageMaker AI.
Lifehacker reviewed ChatGPT's new Slides tool, which generates presentation decks from prompts, and compared its output against Google Gemini's slides feature.
Amazon Quick Sight layer maps now accept Point and LineString geometry from a single GeoJSON workflow, so users can plot store locations and the routes between them. The post covers dataset structure and viewport techniques.
Copilot Autofix in Azure DevOps now covers dependency and third-party security alerts, extending automated fix suggestions beyond first-party code.
A University of Michigan professor created Jac, a Python-like programming language for AI agents meant to replace polyglot stacks and avoid costly AI-driven rewrites, The Register reports.
A Hugging Face dataset for BlenderWorld, a benchmark of 223 realistic 3D creation tasks across architecture, products, environments and scientific visualization, with text, image, video and spec inputs.
A developer describes running GGUF Q4 models paged into WebGL2 compute buffers via GLSL inside an HTML page, a vibecoded local inference engine with a hand-built tokenizer.
Crush is a terminal-based AI coding assistant from Charm that gives developers intelligent coding help directly in their shell. Version 0.98.2 is available via npm.
A Hugging Face Space that runs Qwen-Image-2.1 image generation and editing with an uncensored LoRA, letting users create and modify images in the browser via ZeroGPU.
decks is an open-source HTML slide-deck project where you comment under a slide and a coding agent like Claude Code or Codex makes the edit and replies inline. It gained 100 stars in its first week.
ZoomInfo has integrated Exa's search API so AI agents can pull web search results alongside ZoomInfo's company and contact data.
Stalemap is a TypeScript tool that diffs types and changelogs for dependencies updated after a coding agent's training cutoff, then writes a short note into AGENTS.md or CLAUDE.md.
The Verge reports a surge of viral projects running retro games like Halo in browsers, Wind Waker on iPhone and Super Smash Bros. in mixed reality, driven by new AI tools.
Perplexity is a free AI answer app with cited sources. Version 26.42.0 for iPhone and iPad adds Duo support plus bug fixes.
Linus Torvalds commented that AI is acceptable for competent developers but criticized low-quality 'drive-by' patches, as kernel maintainers stop asking who used AI.
Brave's Leo AI assistant now offers OpenAI's GPT 5.6 and xAI's Grok 4.6 as model options alongside existing choices.
Version 1.64.1 of the iPhone and iPad app adds a one-tap 'Create new' button that makes a notebook without opening the source upload pop-up, plus bug fixes. Also now available as a Connected App.
lessloc is a JavaScript tool that constrains AI coding agents such as Claude Code, Codex, GitHub Copilot, Gemini CLI and Cursor to the least code that does the job and still works. The new repo passed 100 stars in its first week.
A dataset of 1,342 selected AI-generated 3D examples with about 1,500 output files (~39 GB), including previews and source inputs, released under CC BY 4.0 for training and research.
A Hugging Face dataset for BlenderWorld, a benchmark of 223 realistic 3D creation tasks across architecture, products, environments and scientific visualization, with text, image, video and spec inputs.
A Hugging Face Space that runs Qwen-Image-2.1 image generation and editing with an uncensored LoRA, letting users create and modify images in the browser via ZeroGPU.
Carbon-X is an open DNA language model that lets researchers chat with genetic sequences to predict how changes affect gene activity and health, and it can run locally via llama.cpp.
A Hugging Face paper proposes factor-aware predictive learning that makes multimodal LLM representations actionable for video diffusion transformers, enabling sparse control of character, scene, action and camera in cinematic video generation.
A ComfyUI-ready diffusion model that turns a prompt plus ordered image, video and audio references into video with synchronized audio. Ships in BF16 and INT8 ConvRot versions based on FastH3-Omni-8-Step-V1.
A paper presents IQuest-Q1, a sparse Mixture-of-Experts agentic foundation model with 320B total parameters (15B activated), trained via stable pre-training with Stable SplitNorm, Muon and muP plus a two-stage curriculum.
WeVisOne-VL (WeChat Vision All-in-One) is a vision-language model that unifies object- and pixel-level perception through structured next-token prediction, handling detection, referring localization, counting, OCR, segmentation and matting.
The official TypeScript SDK for building Model Context Protocol servers and clients gets version 2.4.0, updating its client, server, core, legacy and codemod packages.
decks is an open-source HTML slide-deck project where you comment under a slide and a coding agent like Claude Code or Codex makes the edit and replies inline. It gained 100 stars in its first week.
Stalemap is a TypeScript tool that diffs types and changelogs for dependencies updated after a coding agent's training cutoff, then writes a short note into AGENTS.md or CLAUDE.md.
lessloc is a JavaScript tool that constrains AI coding agents such as Claude Code, Codex, GitHub Copilot, Gemini CLI and Cursor to the least code that does the job and still works. The new repo passed 100 stars in its first week.
Dify is an open-source platform for building agentic workflows and RAG pipelines with many AI models and tools, deployable on cloud, VPC or self-hosted. Version 1.18.0 is out.
Hugging Face released diffusers v0.42.0 for image and video generation models, and an AWS guide explains how to check whether a model runs on Trainium chips.
goose is an open-source, extensible AI agent that installs, executes, edits and tests code with any LLM; version 1.56.0 fixes a Windows build error.
A GitHub repo laying out a beginner's path through foundations, LLM engineering, agents, evaluation and production, with verified free resources and a mini-project per layer. It drew 101 stars in its first week.
LocalAI is an open-source engine for running LLMs, vision, voice, image and video models locally, no GPU required. Version 4.12.0 redesigns the web UI and adds streaming voice responses, decision routing, image upscaling and memory controls.
The Gemini Android app now has a redesigned side panel and simpler settings, following the Spark-to-Task update.
Frustrated by intrusive ads, slow interfaces and tracking, smart TV owners are using AI coding tools and open-source projects to remove unwanted software and build custom home screens, The Verge reports.
Purdue University Northwest is providing first-year students with Google Gemini licenses to support AI literacy instruction.
Wispr Flow is voice-to-text dictation software with AI-powered auto-editing; version 1.6.1253 is available for macOS, Windows and Linux.
Granola is an AI-powered notepad for meetings. Version 7.668.3 is out for macOS, Windows and Linux; no further changes were detailed in the release listing.
Lifehacker reviewed ChatGPT's new Slides tool, which generates presentation decks from prompts, and compared its output against Google Gemini's slides feature.
The Verge reports a surge of viral projects running retro games like Halo in browsers, Wind Waker on iPhone and Super Smash Bros. in mixed reality, driven by new AI tools.
Perplexity is a free AI answer app with cited sources. Version 26.42.0 for iPhone and iPad adds Duo support plus bug fixes.
Brave's Leo AI assistant now offers OpenAI's GPT 5.6 and xAI's Grok 4.6 as model options alongside existing choices.
Version 1.64.1 of the iPhone and iPad app adds a one-tap 'Create new' button that makes a notebook without opening the source upload pop-up, plus bug fixes. Also now available as a Connected App.
Tech-insider.org compares Claude, ChatGPT, Gemini and Grok usage, reporting Claude at 56.4% share in 2026.
ShogunAI is a personal AI assistant for Mac that remembers your work and helps with meeting prep, search, writing and follow-ups. A 7-day free trial is available.
Gemini Notebook now lets users review slides and videos before generating them. Google launched the connected app with a global rollout that excludes the EEA.
Raycast is a keyboard-driven launcher that runs commands, tools and extensions without leaving the keyboard. Version 2.9.0.0 ships for macOS, Windows and Linux.
Gauth, an AI study companion app for iPhone and iPad covering all subjects and grades, releases version 2.11.0 with small user-experience optimizations.
Korean coverage examines cases where AI systems gave wrong results and consumers had trouble obtaining refunds or redress, highlighting gaps in AI accountability for buyers.
X Chat added access to xAI's Grok Bot, which can now read and reply to X DMs, use iPhone calendars, reminders and contacts, and run tasks three times faster.
Google's Gemini assistant in Android Auto can now pull answers from a user's Gmail and Photos while driving, raising questions about privacy for passengers in the car.
Otter AI, a meeting assistant that transcribes in real time and writes notes, updates its iPhone and iPad app to 3.110.1, adding support for iPhone Duo.
Microsoft is retiring an Outlook feature that lets Copilot assign priority to emails, removing the AI-powered inbox sorting capability.
Google made Gemini-built web apps easier to share and expanded Search's agentic capabilities with information agents and Universal Cart; Gemini's Personal Intelligence is unavailable in Spain.
boost.ai released controls letting customer-experience teams configure and manage their AI agents, per MarTech Cube.
Discord's latest update renders Markdown correctly instead of showing raw symbols, a change that helps people reading agent instruction files and context shared in chat.
Nvidia's new driver targets PC game stutter and Microsoft cloud shader delivery rolls out, but the Modern Warfare 4 driver is already drawing early user complaints.
MacWhisper, a speech recognition tool that transcribes audio locally, released version 15.8 for macOS, Windows and Linux.
Antigravity is an agent orchestration platform, now at version 2.26.0, available for macOS, Windows and Linux.
Simon Willison shares a Claude prompt that builds an artifact where you open a photo and it quickly paints a version using a canvas element and simulated paintbrush.
Codex is OpenAI's coding agent that works alongside you in your editor; the VS Code extension is included in ChatGPT Plus, Pro, Business, Edu and Enterprise plans.
Devin Desktop is an agentic IDE with an AI agent command center. Version 3.11.212 is available for macOS, Windows and Linux.
Nvidia's developer blog explains how the Global Performance Tuner scales TensorRT optimization beyond a single build route, treating deployment as a compilation problem with many valid answers.
AWS Machine Learning · 7h
#
AWS published a guide on orchestrating SageMaker AI endpoint auto scaling with AWS CDK and a hands-on guide to multi-turn reinforcement learning on SageMaker AI.
AWS Machine Learning · 7h
#
Amazon Quick Sight layer maps now accept Point and LineString geometry from a single GeoJSON workflow, so users can plot store locations and the routes between them. The post covers dataset structure and viewport techniques.
Copilot Autofix in Azure DevOps now covers dependency and third-party security alerts, extending automated fix suggestions beyond first-party code.
A University of Michigan professor created Jac, a Python-like programming language for AI agents meant to replace polyglot stacks and avoid costly AI-driven rewrites, The Register reports.
A developer describes running GGUF Q4 models paged into WebGL2 compute buffers via GLSL inside an HTML page, a vibecoded local inference engine with a hand-built tokenizer.
Crush is a terminal-based AI coding assistant from Charm that gives developers intelligent coding help directly in their shell. Version 0.98.2 is available via npm.
ZoomInfo has integrated Exa's search API so AI agents can pull web search results alongside ZoomInfo's company and contact data.
Linus Torvalds commented that AI is acceptable for competent developers but criticized low-quality 'drive-by' patches, as kernel maintainers stop asking who used AI.
Functionize released an MCP server that lets AI coding agents connect to its testing platform to create and run automated tests.
At DevSparks Chennai 2026, EY GDS's Arvind Purushothaman argued that moving AI agents into production is fundamentally an engineering challenge, not a model challenge, saying autonomy and control go together.
Diginomica argues that while AI agents make coding easy, business leaders must control development processes and decide what to ship.
Alibaba's Qoder coding agent now runs on Loongson LoongArch64 as well as Linux ARM64, with builds for UnionTech UOS Desktop V20 and Galaxy Kylin Desktop V10.
Ovren lets you describe a change in plain words; it writes code on a safe copy of your project and shows a live preview of frontend changes for approval and merge.
Agent Hunt is a live leaderboard ranking tools AI agents actually use, with positions earned by real agent calls made through Monid.
Context Engine is a shared memory that Claude, ChatGPT, Cursor and Claude Code save to as they work, while teammates read, query and add files. Free to start.
Monish lets AI agents running in Claude Code, Codex, Cursor or n8n show live progress on the iPhone lock screen and approve or deny actions from notifications. Free, no account.
Clairic is Cloudflare's AI model that extracts typed values like names, dates and amounts from messy text or images without generating tokens, with confidence scores. It is in preview on Workers AI at $0.10 per million.
South Korea's Hancom will unveil its AI workforce agent Nomadian in beta in the US in December and pursue an AWS Marketplace listing. The company positions Nomadian to rival Meta's Muse.
Ollama released versions 0.40.9 and 0.40.10 of its tool for running large language models locally on macOS, Windows and Linux.
Factory is a native AI agent interface for building, managing and shipping software. Version 0.205.0 ships for macOS, Windows and Linux.