K3.1 Pro arrives in Kimi Code
Adds Pro and Highspeed; context limits and access depend on the subscription tier.
Kimi ↗Product profile →TOOLORA / Local reading edition
Product news
News summaries are prepared by this site; open the title to read the original.
News snapshot · 2026-10-25 · 70 records
Copilot global default rolls out
Business and Enterprise rollout preserves explicit choices; previews remain opt-in.
GitHub ↗Product profile →JetBrains Air Governance becomes generally available
Centralizes organizational agents, models, permissions and spending, with features depending on the organization’s plan.
JetBrains ↗Product profile →Replit introduces Monitoring Agent
Investigates production alerts and validates fixes; automatic publishing depends on the user’s settings.
Replit ↗Product profile →HeyGen Voice Turbo offers a limited API promotion
The announced free API period ends November 30; this is temporary and subject to promotion conditions.
HeyGen ↗Product profile →HeyGen introduces its video-generation model
Generates video with sound from text, images or references; web, app and API terms differ.
HeyGen ↗Product profile →Mistral introduces Voxtral 3 for real-time voice and tool use
Voxtral 3 brings transcription, speech generation and a voice-agent SDK into streaming conversations with reasoning and tool use. The announcement lists separate language coverage for transcription and speech generation, offers API and self-deployment options, and allows individual pipeline models to be replaced.
Mistral AI ↗Product profile →Gemini expands tools for small businesses
Adds a business hub, app connections and brand context; Google describes a rollout over the following days.
Google ↗Product profile →Perplexity Computer runs locally
Runs locally on supported Apple silicon Macs with Pro, Max or Enterprise.
Perplexity ↗Product profile →MiniMax releases M3.1 Flash through API and Code
MiniMax released M3.1 Flash for agent tasks, coding and computer use through its API, M Plan and MiniMax Code. The announcement schedules weights for later release rather than establishing immediate open-weight availability.
MiniMax ↗Product profile →Anthropic proposes committing 5% of its shares to a public benefit fund
Anthropic plans to allocate 5% of its shares to a new Public Benefit Fund, subject to shareholder approval. For every 50% valuation increase above a baseline, the board may add 2%, supporting public giving and responses to AI’s labor-market effects.
Anthropic ↗Product profile →DeepSeek updates V4.1-Flash-1017
DeepSeek upgrades its web, app and API to V4.1-Flash-1017, retaining the deepseek-flash model ID and pricing. The update targets reasoning and agent tasks, with optimizations for DeepSeek Harness, Tencent CodeBuddy and WorkBuddy.
DeepSeek ↗Product profile →DeepSeek opens its in-house search API
DeepSeek opens Search / Fetch APIs for its own search engine, providing multilingual web search and page retrieval for models and agents.
DeepSeek ↗Product profile →Stack Overflow reports tripled signups after launching its ChatGPT plugin
The plugin lets users discover and configure Stack Overflow for Agents inside ChatGPT, reducing MCP setup steps. The team reports threefold signups and up to tenfold agent creation after launch; two engineers built the product with Codex in a few weeks.
OpenAI ↗Product profile →Cursor adds cloud-agent canvases and a beta Environments API
Cloud agents can build interactive canvases during tasks and share the results by link. The Environments API enters beta for programmatic creation and management of agent environments, while the update also increases repository capacity per environment.
Cursor ↗Product profile →Hunyuan open-sources Hy-Memory 2.0
Releases a model, memory framework and MemHub application for chat, work and coding memory.
腾讯混元 ↗Product profile →Replit details enterprise BYOC options
Distinguishes full-platform BYOC, published-app BYOC and image export, each with different operational responsibilities.
Replit ↗Product profile →Figma Code Layers enters open beta
Teams explore code-backed interactive prototypes together on the design canvas; this is an open beta.
Figma AI ↗Product profile →Figma introduces Markdown Specs
Brings Markdown specifications onto the shared canvas for teams and agents to exchange project context.
Figma AI ↗Product profile →OpenAI introduces the Rosalind Academic Plan
The Rosalind Academic Plan offers eligible US university labs up to 30 free Business Standard seats for up to 12 months. Access to GPT-Rosalind and life-science tools requires approval; API access is not included.
OpenAI ↗Perplexity open-sources SPACE
Publishes isolation, recovery and network controls for developers to deploy and evaluate.
Perplexity ↗Product profile →Runway MCP adds sound effects and background removal
Paid Runway MCP plans add Mirelo SFX 2.0 for separate video soundtracks and BiRefNet v2 for automatic image background removal.
Runway ↗Product profile →Grok Voice 3.0 API launches
New speech-to-speech model available by API; announced pricing matches its predecessor.
SpaceXSI ↗Product profile →Gemini 4 Argon comes to the Gemini app
Google AI Pro and Ultra subscribers can use Gemini 4 Argon for tasks in the Gemini app. Gemini also expands to more than 40 connected apps for workflows including reservations, shopping and business design.
Google ↗Product profile →Google shares Gemini 4 Argon developer updates
Gemini 4 Argon opens to developers for multi-step work such as code migrations, enterprise knowledge tasks and cybersecurity defense. Access includes the paid Gemini API tier through Google AI Studio and Android Studio.
Google ↗Product profile →Manus announces OPPO partnership
Planned for Find X10 AI Mind Pilot; markets and benefits await launch details.
Manus ↗Product profile →Notion 3.8 expands agent context
Adds MCP connections, Skills API and clearer change history for human-agent collaboration.
Notion ↗Product profile →Copilot local sandboxing launches
CLI, app and supported VS Code sessions can limit file, network and credential access.
GitHub ↗Product profile →Oracle’s ChatGPT and Codex workflow case study
Oracle built a ChatGPT Work tool to research comparable roles, compensation and talent pools from a job description. Its recruiting lead reports preparation falling from 2–4 days to 15–20 minutes. Another team uses Codex to turn business questions into SQL queries and reports.
OpenAI ↗Product profile →Harness acquires selected Augment Code assets including Cosmos
Harness announced acquisition of Cosmos, Auggie CLI, the Code Context Engine and related assets, with the team joining. Harness Cosmos is available now, while connections to Harness policy, approval and audit controls are described as future work.
Harness ↗Product profile →Midjourney tests Thinking Mode
Alpha users can try Rerun (Thinking); broad availability has not been announced.
Midjourney ↗Product profile →Anthropic introduces Claude Haiku 5.5
Claude Haiku 5.5 targets frequent tasks such as summaries, classification and database queries, adding adjustable effort to the Haiku family. Anthropic also lowers Sonnet 5.5 cache-read prices and introduces monthly API credits for Max and Team subscribers.
Anthropic ↗Product profile →Cursor iOS adds remote access to local agents
The iOS app lets users view and message agents running on their computers, with pairing approved in the desktop app. Tasks remain on the original computer, which must stay powered on and online. Enterprise organizations require administrator enablement.
Cursor ↗Product profile →Mistral Large 4 opens a preview API ahead of weight release
Mistral introduces Large 4 in preview for coding, agent tasks and multimodal understanding. The announcement provides a preview API and schedules weights for later in the month; API preview access does not establish that weights were already downloadable at publication.
Mistral AI ↗Product profile →Granola introduces meeting Snapshots
Adds relevant shared slides, charts and design screenshots to meeting notes alongside the conversation.
Granola ↗Product profile →Manus introduces Game Dev
Build playable games, edit code and assets, and publish to the web.
Manus ↗Product profile →Manus adds Video Editor
Edit clips, text and audio on a timeline before export.
Manus ↗Product profile →JetBrains opens Air in IDEs EAP
Offers agent workflows through a plugin or 2026.3 EAP; remote features and some entry points remain restricted.
JetBrains ↗Product profile →Suno Speech enters open beta
Generates spoken audio and original background music together as one track; still in beta.
Suno ↗Product profile →Cohere releases Embed 5 Pro and Fast with shared embeddings
Embed 5 offers Pro and Fast tiers for enterprise retrieval and RAG, supporting text, images and combined inputs. Both share an embedding space, allowing Pro indexing and queries with either tier without rebuilding the index for that switch. The release lists API and multiple deployment channels.
Cohere ↗Product profile →Manus Flex supports your key
Connect supported inference providers, which bill model usage directly.
Manus ↗Product profile →Claude Sonnet 5.5 launches with a thinking-setting migration
Sonnet 5.5 targets everyday coding and documents. API users with thinking disabled must migrate to between_tools; reported cost reductions reflect vendor testing.
Anthropic ↗Product profile →ElevenLabs launches v4 and low-latency Turbo
v4 and Turbo reach ElevenCreative, ElevenAgents and the API, with emotion tags and multi-speaker speech. Published September 28; revised October 19.
ElevenLabs ↗Product profile →Manus launches 2.0
Introduces Cascade, Studio and Automations across web, desktop and mobile.
Manus ↗Product profile →JetBrains introduces Air Teams
Shares agent environments, tools and rules across teams; initially for business customers, with personal access planned later.
JetBrains ↗Product profile →Auphonic adds n8n production integration
A verified n8n community node starts Auphonic productions, reacts to completion and routes output files to other services. It connects an Auphonic account and service to existing automation workflows.
Auphonic ↗Product profile →Replit announces the acquisition of Atta
Expands business-user analytics and introduces interactive charts directly in conversations.
Replit ↗Product profile →Cursor adds deployment monitoring and security review
Teams and Enterprise gain deployment monitoring and pull-request vulnerability reviews. Rollouts can propose reversions or fixes, but does not merge or roll back automatically.
Cursor ↗Product profile →Claude Opus 5.5 launches with lower API pricing
Opus 5.5 opens API access for complex coding and knowledge work. Announced input/output pricing is $4/$20 per million tokens; Fast mode costs extra.
Anthropic ↗Product profile →Kimi introduces a finance solution
Combines financial sources, specialist skills and report workflows; institutional results are vendor-reported.
Kimi ↗Product profile →Cursor Projects enters beta for ongoing work
Projects coordinates agents in the cloud, retaining shared context and supporting scheduled or event-triggered work. The September 10 announcement describes a gradual beta rollout.
Cursor ↗Product profile →DeepSeek releases V4.1-Flash
DeepSeek releases V4.1-Flash with native visual understanding and an asymmetric MoE architecture aimed at reducing inference and cache costs. The multimodal API uses deepseek-flash; older Flash model IDs temporarily route to the new model.
DeepSeek ↗Product profile →Suno introduces v6 music models
Adds natural-language edits, multi-source mashups and multimodal creation; older models retire as v6 rolls out.
Suno ↗Product profile →Manus resumes independent operations
The founding team continues leading; affected users can restore backups using official guidance.
Manus ↗Product profile →Reducto introduces its r-1 document-parsing preview
r-1 combines layout, tables, formatting and citation grounding in a single parsing pass for complex documents. This announcement introduces a preview through a Parse API configuration flag; the smaller r-1 mini and automatic routing are described as future plans, not simultaneous releases.
Reducto ↗Product profile →Dify introduces a new Agent builder
Adds conversational building, reviewable drafts, skills and tool composition, plus execution tracing.
Dify ↗Product profile →ElevenLabs CLI v1 adds agent configuration workflows
CLI v1 exposes API operations in the terminal, with structured JSON, parameter schemas and change previews, plus pull/edit/push workflows for agent configuration.
ElevenLabs ↗Product profile →Replit launches Free Mode
Subscription-mode everyday tasks avoid usage credits but retain limits; complex work uses other modes.
Replit ↗Product profile →MiniMax releases Music 3.0 for complete-song generation
Music 3.0 takes a creative description and optional lyrics. MiniMax describes a single generation covering composition, arrangement, vocals and production for songs up to five minutes. Its architecture separates global song structure from local acoustic detail and revises audio rendering.
MiniMax ↗Product profile →Stagehand v4 introduces a shared browser-extension core
Browserbase released Stagehand v4, moving targets, state management and CDP dispatch into an extension running with the browser. TypeScript, Python and Go SDKs share the core, applying the same architecture to local and remote browsers to reduce stale client-side state.
Browserbase ↗Product profile →MiniMax releases open H3 models
Publishes video-generation variants with different input formats, resolutions and generation methods.
MiniMax ↗Product profile →Seedance 2.5 launches with longer shots and reference editing
The announcement introduces 2.5 on Jimeng and Doubao Pro, describing up to 30-second audiovisual generation, repeated extensions, multiple references and timestamp editing. BytePlus ModelArk API access was forthcoming at publication.
ByteDance Seed ↗Product profile →Granola launches on Apple Watch
Starts and stops meeting notes from the watch, then syncs them to phone and computer.
Granola ↗Product profile →Midjourney defaults to V8.1
V8.1 replaces V7 as default; Omni Reference support remains version-specific.
Midjourney ↗Product profile →Google launches Antigravity CLI and outlines consumer Gemini CLI transition
Antigravity CLI shares the desktop agent backend and supports background agents, skills, hooks and plugins. The announcement ends original free and Google AI Pro/Ultra Gemini CLI access on June 18 while preserving separate enterprise and paid API paths; initial features are not one-to-one equivalents.
Google Developers Blog ↗Product profile →Stitch adds real-time canvas collaboration and development handoffs
Google announced that Stitch can start from text, voice, code or design files, stream design work live and accept guidance before completion. The update also describes AI Studio sharing, screen export to Antigravity and publishing through Netlify.
Google ↗Product profile →Recraft V4.1 adds Vector and Utility creative options
The V4.1 family distinguishes the main model, Vector and Utility for visual exploration, vector illustration and lettering, and simpler product or mockup scenes. The original announcement offers Studio and API access; quality comparisons are vendor demonstrations.
Recraft ↗Product profile →Qwen App connects to Taobao’s full catalog for conversational shopping
Alibaba announced full Taobao and Tmall catalog access for Qwen App in China, supporting conversational browsing, comparisons, orders and delivery management. The shopping assistant inside Taobao is a separate entry; virtual try-on was still planned for Qwen App at publication.
Alibaba Group ↗Product profile →Qwen App tests connected shopping, travel and payments in China
Qwen App announced connections to Taobao Instant Commerce, Fliggy, Amap and Alipay for conversational ordering, planning and bookings, with supported payments after user confirmation. Task Assistant was invite-only beta, demonstrating multi-document reports, phone confirmations and lightweight web apps.
Alibaba Group ↗Product profile →Manus announces Meta deal
Historical Meta announcement; a September 2026 notice later confirmed independent operations.
Manus ↗Product profile →