# forck.live — 28 September – 4 October 2026 > Live feed of AI lab newsrooms. First-party announcements, traced to the source. forck.live follows what companies announce about their own products. Press may corroborate an official story; it never leads and never appears as an item of its own. A missing date stays missing. A title is the publisher's words — where a company does not publish in English, an English rendering is listed separately as `Title in English` and named as this site's translation, never as theirs. This file is the week in progress: every confirmed first-party announcement of ISO week 2026-W40, 28 September – 4 October 2026, newest first. The window never moves, so this URL always names the same seven days — but they have not all happened yet, so the file is still filling: announcements are appended as they land, and it will answer with more of them later today. Fetch it again rather than reading a cached copy as the finished week. Every entry is one first-party announcement, and carries the same eight fields in this order before any others: Company · Announced · Category · Coverage · Announcement · Group · Source · Record `Source` is the company's own page the announcement was traced to and `Record` is this site's page for it. Both are safe to cite. `Coverage` is a number of outlets, or `not counted` — which means this record reached no count of its own, never that nobody wrote about it. `Category` is this tracker's own analysis, and is authoritative wherever it is stated. `not stated` means either that nothing has analysed the post yet or that the analysis placed it outside this taxonomy; neither is a verdict about the announcement. `Announcement` is whether the post is news of something the company did, rather than a how-to, a customer story, a changelog listing or a staffing note. `Group` is which of this record's groups the entry falls in, carried on the entry itself so it holds where the groups are headings and where they are not. One of: `models`, `covered`, `announcements`, `routine`. Anything after `Record` is held detail that only some announcements carry. - This week's page: https://forck.live/week/2026-W40 - The current issue, and the catalogue of the other weeks: https://forck.live/briefing.md - Catalogue: https://forck.live/llms.txt ## Announcements ### Amazon — Sweep thousands of leases for compliance using Amazon Quick and the Adjudicated Query pattern - Company: Amazon (amazon.com) - Announced: 2026-10-02T15:48:26+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://aws.amazon.com/blogs/machine-learning/sweep-thousands-of-leases-for-compliance-using-amazon-quick-and-the-adjudicated-query-pattern/ - Record: https://forck.live/items/15760-sweep-thousands-of-leases-for-compliance-using-amazon-quick-and-the - Subject: Bedrock / Nova Amazon introduced the Adjudicated Query pattern, which uses a deterministic rules engine behind a conversational interface in Amazon Quick to check lease compliance. The pattern ensures provable completeness and defensibility by never allowing the model to write queries or determine populations, and every sweep produces a completeness receipt. The reference architecture uses Amazon Quick, AWS Lambda, Amazon Aurora Serverless v2, and Amazon Bedrock only for exploratory clause searches. ### Amazon — Add secure Web Search to Claude Desktop with Amazon Bedrock AgentCore - Company: Amazon (amazon.com) - Announced: 2026-10-02T15:46:05+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://aws.amazon.com/blogs/machine-learning/add-secure-web-search-to-claude-desktop-with-amazon-bedrock-agentcore/ - Record: https://forck.live/items/15761-add-secure-web-search-to-claude-desktop-with-amazon-bedrock-agentcore - Subject: Bedrock / Nova Amazon published a walkthrough showing how to connect Claude Desktop on Amazon Bedrock to a managed web search capability via Bedrock AgentCore Gateway. The integration uses AWS IAM Identity Center, Amazon Cognito, and JWT-based authentication to keep all query traffic within AWS infrastructure. Web Search is described as a fully managed, MCP-compatible service backed by an Amazon web index of tens of billions of documents. ### Amazon — Fine-tune a search agent with multi-turn RL on Amazon SageMaker AI - Company: Amazon (amazon.com) - Announced: 2026-10-02T15:44:20+00:00 - Category: not stated - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://aws.amazon.com/blogs/machine-learning/fine-tune-a-search-agent-with-multi-turn-rl-on-amazon-sagemaker-ai/ - Record: https://forck.live/items/15762-fine-tune-a-search-agent-with-multi-turn-rl-on-amazon-sagemaker-ai - Subject: Bedrock / Nova Search agents powered by large language models (LLMs) are transforming how enterprises retrieve information. Rather than requiring users to craft the perfect query, a search agent autonomously decides what to search for, which retrieval strategy to use, and when to stop searching. It does this across multiple rounds of interaction, refining its approach based on what it has already retrieved. However, getting this multi-step behavior to work well is hard. No base model arrives knowing your tools or your environment. Prompt a small model and you rarely get dependable multi-turn behavior. Prompt a frontier model and it often works, but you pay for that capability in latency and cost. Fine-tuning offers a third path: you teach a small model your tools and environment directly. The result is a small model’s speed and cost with the reliability that would otherwise require a frontier model. Even though fine-tuning is the natural next step, the traditional approaches each fall short. Supervised fine-tuning (SFT) depends on expert demonstrations of ideal multi-turn trajectories, which are costly to collect and usually don’t exist for your setup. … ### Hugging Face — Open-sourcing AstaBrief, the fast report-generation model in Asta - Company: Hugging Face (huggingface.co) - Announced: 2026-10-02T15:19:50+00:00 - Category: not stated - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://huggingface.co/blog/allenai/astabrief - Record: https://forck.live/items/15758-open-sourcing-astabrief-the-fast-report-generation-model-in-asta - Subject: Platform Language models can already help researchers search the literature, synthesize evidence, and work through complex questions. But scientific work places particular demands on these models—answers need to stay grounded in evidence, the models need to preserve what the evidence actually supports rather than quietly broadening a study’s conclusions, and researchers need to be able to verify the final outputs. We see that in how scientists use Asta , our agentic platform for scientific work. Instead of simple keyword searches, users often bring substantial context and many constraints—for example, asking Asta to compare approaches across a body of literature while accounting for a particular method, population, or setting. Many also return to generated reports later, treating them as working research artifacts rather than one-off answers. We wanted to help scientists generate cited reports faster, with a model they could download and run themselves. To do that, we tested whether a small, open model trained specifically for scientific report generation could match the report quality of the proprietary models we were using, while reducing generation time and serving costs. … ### Google Research — Toward provably private learning from federated data - Company: Google Research (research.google) - Announced: 2026-10-02T14:57:41+00:00 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://research.google/blog/toward-provably-private-learning-from-federated-data/ - Record: https://forck.live/items/15749-toward-provably-private-learning-from-federated-data - Subject: Research Google Research announced a new Federated Learning system that uses Trusted Execution Environments (TEEs) to provide externally verifiable privacy guarantees. The system shifts computation to the server to improve training speed, accuracy, and device coverage. Gboard has already adopted the new system and is benefiting from substantially faster compute times. ### NVIDIA — NVIDIA DGX Spark 64GB Gives Developers More Ways to Build and Scale Local AI - Company: NVIDIA (nvidia.com) - Announced: 2026-10-02T13:00:39+00:00 - Category: not stated - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://blogs.nvidia.com/blog/local-ai-dgx-spark-64gb-sync/ - Record: https://forck.live/items/15740-nvidia-dgx-spark-64gb-gives-developers-more-ways-to-build-and-scale-local-ai - Subject: AI platform Local AI is becoming more useful by the token. As AI agents move from experiments into everyday development, increasingly capable open models are shrinking to fit on more devices, giving builders more to run locally. Coming this month, NVIDIA DGX Spark will be available with 64GB of unified memory from top manufacturer partners — Acer, ASUS, Dell, Gigabyte, HP and MSI — giving developers, researchers and AI enthusiasts a new configuration with DGX OS and the NVIDIA AI software stack ready to use from day one. The new SKU runs capable local agents on device — privately, without cloud dependency. And when workloads grow, two units can cluster together via NVIDIA Sync Cluster Assistant without any additional setup. A New Starting Point for Personal AI Supercomputing DGX Spark combines NVIDIA Grace Blackwell compute, unified memory, NVIDIA ConnectX-7 networking and an NVIDIA CUDA -accelerated AI software stack in one system. It’s a complete local AI platform for agents, inference, fine-tuning, data science and edge development. … ### Hugging Face — AutoSynthData: Generating Training Data for Enterprise Agents - Company: Hugging Face (huggingface.co) - Announced: 2026-10-02T04:01:31+00:00 - Category: research-paper - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://huggingface.co/blog/ServiceNow-AI/autosynthdata - Record: https://forck.live/items/15683-autosynthdata-generating-training-data-for-enterprise-agents - Subject: Platform ServiceNow CoreAI introduces AutoSynthData, a pipeline that identifies capability gaps in a target model by evaluating it in an enterprise environment, then generates and validates new training tasks that exercise those weaknesses. The system uses a stronger teacher model to characterize successful behavior and produces tasks that are feasible, realistic, and difficult for the current model. The approach is illustrated with the EnterpriseOps Gym environment and a released dataset. ### Anthropic — Anthropic invests $100 million to train 10,000 engineers and tackle the enterprise AI talent gap - Company: Anthropic (anthropic.com) - Announced: 2026-10-02 - Category: not stated - Coverage: 2 outlets - Announcement: no - Group: covered - Source: https://www.anthropic.com/news/claude-frontier-academy - Record: https://forck.live/items/15772-anthropic-invests-100-million-to-train-10-000-engineers-and-tackle-the - Subject: Claude Anthropic announced the launch of Claude Frontier Academy, a program backed by a $100 million commitment to train 10,000 Frontier Deployed Engineers by the end of 2027. The first cohorts include engineers from Accenture, Bain, Capgemini, Commonwealth Bank of Australia, Deloitte, McKinsey, Morgan Stanley, and Novo Nordisk. The program follows a residency model with in-person training and a 12-week project, leading to a Claude Frontier Deployed Engineer badge. ### Perplexity — AI maturity: what it means for your organization - Company: Perplexity (perplexity.ai) - Announced: 2026-10-02 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://www.perplexity.ai/hub/blog/ai-maturity - Record: https://forck.live/items/15769-ai-maturity-what-it-means-for-your-organization - Subject: Perplexity Perplexity's guide defines AI maturity as the degree of deliberate integration of AI beyond tool adoption, describing a progression from unstructured experimentation to full organizational transformation. It cites examples from Inteleos, Paypal, and Rox, and summarizes frameworks from Gartner, Deloitte, and Google Cloud that share a common journey of pilots, workflow integration, and eventual restructuring around AI. The guide also identifies common blockers such as data quality and offers steps to address them. ### Apple — Language Discrimination Improves Linguistic Learning in Multilingual Speech Models - Company: Apple (apple.com) - Announced: 2026-10-02 - Category: research-paper - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://machinelearning.apple.com/research/language-discrimination-multilingual-learning - Record: https://forck.live/items/15747-language-discrimination-improves-linguistic-learning-in-multilingual-speech - Subject: Machine Learning Research Apple researchers found that enhancing language discrimination during pretraining of multilingual speech models (using an auxiliary language classifier and per-language k-means targets) reduces the performance gap compared to monolingual models. In a controlled English/French HuBERT setting, phone discrimination error decreased from 11.6% (bilingual baseline) to 10.4%, while lexical and prosodic performance improved, in some cases matching or exceeding monolingual baselines. The strongest gains occurred when language discrimination was introduced in the first training iteration. ### LG AI Research — [ICML2026] Application of Mechanistic Interpretability: Understanding Graph Transformers (TokenGT) - Company: LG AI Research (lgresearch.ai) - Announced: 2026-10-02 - Category: research-paper - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://www.lgresearch.ai/blog/view?seq=712 - Record: https://forck.live/items/15706-icml2026-application-of-mechanistic-interpretability-understanding-graph - Subject: EXAONE LG AI Research presented a study applying mechanistic interpretability to graph transformers, specifically analyzing how TokenGT processes graph data. The work found that models trained on degree calculation, cycle detection, and shortest-path distance tasks all begin with a shared local-structure computation, such as encoding node degree through ID-matching attention in the first transformer layer. ### NVIDIA — How NVIDIA GPUs Help Accelerate OpenAI’s GPT-6 Astra Ultrafast - Company: NVIDIA (nvidia.com) - Announced: 2026-10-01T23:44:13+00:00 - Category: model-update - Coverage: not counted - Announcement: no - Group: routine - Source: https://blogs.nvidia.com/blog/gpus-openai-gpt-6-astra-ultrafast/ - Record: https://forck.live/items/15647-how-nvidia-gpus-help-accelerate-openai-s-gpt-6-astra-ultrafast - Subject: AI platform OpenAI released GPT-6 Astra Ultrafast, a faster inference mode running on NVIDIA Blackwell GPUs, offering up to 8x faster token generation than the Astra Standard mode. The model is available now in the OpenAI API and to eligible ChatGPT Work and Codex users. OpenAI used its own models to optimize inference software on NVIDIA GPUs, enabling ongoing performance improvements. ### Amazon — Serve live, governed data in AI-built apps with Amazon Quick - Company: Amazon (amazon.com) - Announced: 2026-10-01T19:49:06+00:00 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://aws.amazon.com/blogs/machine-learning/serve-live-governed-data-in-ai-built-apps-with-amazon-quick/ - Record: https://forck.live/items/15632-serve-live-governed-data-in-ai-built-apps-with-amazon-quick - Subject: Bedrock / Nova Amazon Quick introduces Live Data in Apps, allowing AI-built Quick apps to query governed Quick Sight datasets in real time rather than relying on static build-time snapshots. The feature supports SPICE and Direct Query datasets, enforces row-level and column-level security per viewer, and requires authenticated Quick users with Reader Pro role or higher. Builders describe apps in natural language; the agent discovers relevant datasets and writes SQL that re-runs on each open. ### GitHub — GitHub Copilot can now interact with desktop apps with computer use - Company: GitHub (github.com) - Announced: 2026-10-01T19:11:26+00:00 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://github.blog/changelog/2026-10-01-github-copilot-can-now-interact-with-desktop-apps - Record: https://forck.live/items/15628-github-copilot-can-now-interact-with-desktop-apps-with-computer-use - Subject: Copilot GitHub Copilot can now interact with desktop applications on macOS and Windows in public preview, reading accessible content and visual context, clicking controls, entering text, pressing keys, scrolling, dragging, and navigating workflows across apps. The feature works in Copilot CLI and the GitHub Copilot app, requires user approval before controlling an app, and can be disabled by organization-managed settings. ### GitHub — GitHub Copilot in VS Code, September 2026 releases - Company: GitHub (github.com) - Announced: 2026-10-01T19:09:10+00:00 - Category: capability-change - Coverage: not counted - Announcement: no - Group: routine - Source: https://github.blog/changelog/2026-10-01-github-copilot-in-vs-code-september-2026-releases - Record: https://forck.live/items/15629-github-copilot-in-vs-code-september-2026-releases - Subject: Copilot GitHub Copilot in VS Code gained agent-driven features including automations for scheduled tasks, agent-assisted pull request merging, and improved session management. Agents can now create pull requests, navigate related chats, and run in Dev Containers. A new HydraFusion feature in research preview coordinates model selection automatically. ### Amazon — Build agent memory with NVIDIA NeMo Agent Toolkit and Amazon S3 Vectors - Company: Amazon (amazon.com) - Announced: 2026-10-01T17:34:56+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://aws.amazon.com/blogs/machine-learning/build-agent-memory-with-nvidia-nemo-agent-toolkit-and-amazon-s3-vectors/ - Record: https://forck.live/items/15619-build-agent-memory-with-nvidia-nemo-agent-toolkit-and-amazon-s3-vectors - Subject: Bedrock / Nova Amazon published a guide showing how to use Amazon S3 Vectors as a persistent memory layer within the NVIDIA NeMo Agent Toolkit (NAT) on Amazon EKS. The implementation involves creating an S3 Vectors bucket and index, implementing a custom MemoryEditor plugin, and configuring the agent workflow. NAT's memory subsystem stores conversation history, user preferences, and long-term knowledge across agent invocations. ### OpenAI — The eternal complement - Company: OpenAI (openai.com) - Announced: 2026-10-01T17:00:00+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://openai.com/index/the-eternal-complement - Record: https://forck.live/items/15617-the-eternal-complement - Subject: GPT / ChatGPT / API This essay argues that as AI advances, its greatest value may lie in handling the monotonous, institutional work that supports frontier innovation, rather than in generating brilliant insights. The authors contend that progress depends on a vast support system of execution, and AI can help reduce the scarcity of that execution. ### Amazon — Uplifting conversion across the acquisition funnel with personalization using contextual bandits on AWS - Company: Amazon (amazon.com) - Announced: 2026-10-01T16:51:04+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://aws.amazon.com/blogs/machine-learning/uplifting-conversion-across-the-acquisition-funnel-with-personalization-using-contextual-bandits-on-aws/ - Record: https://forck.live/items/15608-uplifting-conversion-across-the-acquisition-funnel-with-personalization-using - Subject: Bedrock / Nova Amazon Payments applied a multi-objective contextual multi-armed bandit (LinUCB) on Amazon SageMaker AI to personalize a product acquisition funnel. A seven-week A/B test showed a high single-digit percentage relative lift in final-funnel conversion for one customer population, while another saw no improvement, attributed to the content rather than the model. The post explains the bandit approach, the extension to optimize an entire funnel, and the AWS architecture. ### Amazon — Building ambient agents with Amazon Bedrock AgentCore: From event-driven signals to human-in-the-loop workflows - Company: Amazon (amazon.com) - Announced: 2026-10-01T16:40:24+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://aws.amazon.com/blogs/machine-learning/building-ambient-agents-with-amazon-bedrock-agentcore-from-event-driven-signals-to-human-in-the-loop-workflows/ - Record: https://forck.live/items/15604-building-ambient-agents-with-amazon-bedrock-agentcore-from-event-driven - Subject: Bedrock / Nova Amazon published a guide on building ambient agents using Amazon Bedrock AgentCore. Ambient agents listen to event streams (such as file uploads or system alerts) and act on them without requiring a human to start a conversation, pausing only for human input when clarification or approval is needed. The post walks through an end-to-end pattern on AgentCore Runtime with a reference implementation that uses AWS Lambda, Amazon DynamoDB, and a single ask_human tool for human-in-the-loop interactions. ### Amazon — Implementing Multi-Environment Access for Claude Platform on AWS - Company: Amazon (amazon.com) - Announced: 2026-10-01T16:32:23+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://aws.amazon.com/blogs/machine-learning/implementing-multi-environment-access-for-claude-platform-on-aws/ - Record: https://forck.live/items/15601-implementing-multi-environment-access-for-claude-platform-on-aws - Subject: Bedrock / Nova Amazon published a guide for implementing multi-environment access to Claude Platform on AWS (CPonAWS). The post covers three access patterns: cross-account SigV4 for AWS workloads, workspace-scoped API keys for developers, and OIDC federation for external environments. It provides step-by-step instructions for setting up a dedicated AI Services account, workspaces, and authentication paths. ### GitHub — Dynamic workflows in Copilot CLI and the Copilot app - Company: GitHub (github.com) - Announced: 2026-10-01T16:30:10+00:00 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://github.blog/changelog/2026-10-01-dynamic-workflows-in-copilot-cli-and-the-copilot-app - Record: https://forck.live/items/15639-dynamic-workflows-in-copilot-cli-and-the-copilot-app - Subject: Copilot GitHub introduced dynamic workflows for Copilot CLI, the GitHub Copilot app, and the GitHub Copilot SDK. These allow users to define an orchestration in code that combines automated steps with agent work, supporting sequential or parallel execution, subagent verification, and checkpoints. Dynamic workflows are available on all Copilot plans and are in public preview. ### Amazon — Simplify dashboard drill-down with the Amazon Quick Sight hierarchy filter - Company: Amazon (amazon.com) - Announced: 2026-10-01T16:28:14+00:00 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://aws.amazon.com/blogs/machine-learning/simplify-dashboard-drill-down-with-the-amazon-quick-sight-hierarchy-filter/ - Record: https://forck.live/items/15602-simplify-dashboard-drill-down-with-the-amazon-quick-sight-hierarchy-filter - Subject: Bedrock / Nova Amazon announced the hierarchy filter for Quick Sight, a single compact control that replaces multiple independent filters. It supports up to five levels of drill-down (e.g., Region → Country → City) and reduces visual clutter while guiding readers through a logical path. The post includes a walkthrough for configuring the filter on a sample retail sales dataset. ### OpenAI — How Albertsons Companies is reimagining retail from the inside out - Company: OpenAI (openai.com) - Announced: 2026-10-01T16:00:00+00:00 - Category: not stated - Coverage: 1 outlet - Announcement: no - Group: covered - Source: https://openai.com/index/albertsons-reimagining-retail - Record: https://forck.live/items/15599-how-albertsons-companies-is-reimagining-retail-from-the-inside-out - Subject: GPT / ChatGPT / API Albertsons Companies is expanding its partnership with OpenAI to integrate AI into retail operations and customer experiences. The retailer is using ChatGPT Enterprise and the OpenAI API to help teams streamline workflows and to offer a new Safeway shopping experience within ChatGPT, where customers can plan meals, build carts, and check out. The initiative is planned to extend to other Albertsons-owned grocery brands beyond Safeway. ### Amazon — How uniopen customized Amazon Nova to their retail moderation policies for production deployment - Company: Amazon (amazon.com) - Announced: 2026-10-01T15:33:01+00:00 - Category: not stated - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://aws.amazon.com/blogs/machine-learning/how-uniopen-customized-amazon-nova-to-their-retail-moderation-policies-for-production-deployment/ - Record: https://forck.live/items/15595-how-uniopen-customized-amazon-nova-to-their-retail-moderation-policies-for - Subject: Bedrock / Nova uniopen is a digital communication and membership platform launched by Taiwan’s Uni-President Enterprises Group, connecting customers to ecommerce, membership benefits, and other retail experiences across web, tablet, and mobile channels. Across those channels, uniopen applies a moderation policy that classifies each interaction along two axes. The first is what behavior occurred (nine categories), and the second is what subject the behavior refers to (brand, other, or forbidden). Both must be correct for a moderation decision to be useful, and both are specific to uniopen’s business rather than something a general-purpose model can be expected to learn out of the box. In this post, we show how the team adapted Amazon Nova 2 Lite to these business-specific moderation policies through supervised fine-tuning in Amazon SageMaker AI and a final prompt-level output optimization. The AWS approach kept correction data, managed training, evaluation, and deployment controls in one repeatable workflow. Model availability varies by AWS Region. See Supported models by AWS Region in Amazon Bedrock . Figure 1 shows the web, tablet, and mobile experiences covered by this moderation policy. … ### Hugging Face — Introducing Olmo-core 3: Open, scalable training infrastructure for large MoEs - Company: Hugging Face (huggingface.co) - Announced: 2026-10-01T15:01:43+00:00 - Category: developer-tool-release - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://huggingface.co/blog/allenai/olmocore3 - Record: https://forck.live/items/15589-introducing-olmo-core-3-open-scalable-training-infrastructure-for-large-moes - Subject: Platform Hugging Face released Olmo-core 3, an upgrade to its framework for developing large language models featuring a redesigned open mixture-of-experts (MoE) training system. The system is designed to scale MoE training into the trillion-parameter range while preserving computational efficiency. In one benchmark, increasing the expert pool from 8 to 128 while keeping active parameters per token fixed at about 3.2B resulted in total parameter capacity growing from 4.6B to 47B with training throughput falling by less than 5%. ### Ideogram — Ideogram 4.5 - Company: Ideogram - Announced: not stated by the source (first seen 2026-10-01T14:56:49.191608+00:00) - Category: model-update - Coverage: not counted - Announcement: yes - Group: models - Source: https://ideogram.ai/models/4.5/ - Record: https://forck.live/items/15588-ideogram-4-5 - Subject: Ideogram image models Ideogram released version 4.5 of its image model, which reduces pixel shift, color changes, and texture artifacts during multi-turn edits, preserving details across successive modifications. The model supports editing at high resolutions up to 24.2 megapixels without downsizing, and can handle edits such as color and lighting changes, text modification, product photography, interior design, architecture, photo restoration, sketch-to-image, style reference, reframing, depth-to-image, and zoom editing. ### NVIDIA — Productive, Durable, Fungible: How NVIDIA AI Factories Maximize Return on Investment - Company: NVIDIA (nvidia.com) - Announced: 2026-10-01T13:00:49+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://blogs.nvidia.com/blog/productive-durable-fungible-ai-factories/ - Record: https://forck.live/items/15573-productive-durable-fungible-how-nvidia-ai-factories-maximize-return-on - Subject: AI platform NVIDIA describes how AI factories maximize return on investment through three attributes: productivity (highest throughput per megawatt and lowest cost per token), durability (GPUs like A100 remain in service for years, with extended depreciation schedules), and fungibility (running every type of AI model and workload across phases and locations). The post cites SemiAnalysis data showing Vera Rubin NVL72 systems deliver over 30x higher throughput per megawatt and up to 45x lower cost per million tokens than GB300 NVL72 on the DeepSeek V4 Pro model. ### Manus — Introducing Game Dev: Enabling Everyone to Make the Games They Want to Play - Company: Manus - Announced: 2026-10-01 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://manus.im/blog/introducing-game-developer - Record: https://forck.live/items/15678-introducing-game-dev-enabling-everyone-to-make-the-games-they-want-to-play - Subject: Manus Manus 2.0 introduces Game Dev, a feature that enables people without coding experience to create, tweak, and share games using AI. It includes a real-time tweak panel, asset management with AI editing, and built-in multiplayer server support. The feature is part of the Manus 2.0 update launched on September 28. ### Manus — Introducing Video Editor in Manus 2.0: Anyone Can Now Make A Video Worth Posting - Company: Manus - Announced: 2026-10-01 - Category: not stated - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://manus.im/blog/introducing-video-editor - Record: https://forck.live/items/15677-introducing-video-editor-in-manus-2-0-anyone-can-now-make-a-video-worth-posting - Subject: Manus Video models are already remarkably good. Write one sentence and a few minutes later you have a beautiful shot. But the more videos I made, the clearer it became how far a great shot is from a video you can actually publish. First, you have to figure out what the video is really about. Then come the questions: How do you get people to stop scrolling in the first two seconds? Where should the joke land? What music fits, and should it dip when someone is talking? Which font should the caption use? Which beat should each cut hit? And on top of all that, everything has to match the brand’s tone. Making a video means hundreds of small decisions, and every one of them depends on context. The second challenge comes after the first version is generated. There’s always something you want to change: a more upbeat song, a reworded caption, a shot trimmed by half a second. If all you have is a chat box, small tweaks like these often mean regenerating the entire video, and the parts you liked in the previous version may change along with them. So we set ourselves one question How can someone who has never learned video editing make a video they want to share? Our answer has two halves. … ### Suno — Introducing Speech (beta) - Company: Suno (suno.com) - Announced: 2026-10-01 - Category: capability-change - Coverage: 2 outlets - Announcement: yes - Group: covered - Source: https://suno.com/blog/introducing-speech-beta - Record: https://forck.live/items/15645-introducing-speech-beta - Subject: Music / Studio Suno released Speech, a beta audio model that generates voice and background music together as a single track, allowing users to create spoken audio set to original music by typing an idea and describing the desired voice and musical style. ### OpenAI — The Den frees up 10-15 hours a week to grow with ChatGPT Work - Company: OpenAI (openai.com) - Announced: 2026-10-01 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://openai.com/index/the-den-family-social - Record: https://forck.live/items/15641-the-den-frees-up-10-15-hours-a-week-to-grow-with-chatgpt-work - Subject: GPT / ChatGPT / API The Den Family Social, a Denver-based social club for parents, reports saving 10-15 hours per week using ChatGPT Work. The team reduced grant application time from three days to two hours and liquor-license preparation from four days to three hours by using ChatGPT Work with Gmail, Slack, and Google Drive plugins to gather information and draft documents. ### Anthropic — Customize Claude Code with mods - Company: Anthropic (anthropic.com) - Announced: 2026-10-01 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://claude.com/blog/claude-code-mods - Record: https://forck.live/items/15621-customize-claude-code-with-mods - Subject: Claude Change how Claude Code behaves and looks with a few lines of TypeScript. Category Product announcements Product Claude Code Date October 1, 2026 Reading time 5 min Share Copy link https://claude.com/blog/claude-code-mods Today we're introducing mods, small TypeScript functions that change how Claude Code works. A mod can rewrite a prompt, add new UI, replace a built-in feature, or add entirely new functionality. You can write a mod yourself, or ask Claude Code to write one for you. Mods ship inside plugins, so you install and share them like any plugin. They work in the Claude Code CLI and desktop app. Mods run with the same access to your machine as Claude Code itself. They aren’t sandboxed, and you should only install mods from sources you trust, the same way you'd install any code on your computer. To see what mods can do, read our guide to building your first mod . Why we built mods Developers have asked for more control over how Claude Code works, without waiting for us to ship a feature. Hooks helped give users some of this control, but hooks can't rewrite events, draw new UI, or replace features. Mods can. … ### Microsoft AI — Our first streaming transcription model debuts at no. 1 on Artificial Analysis - Company: Microsoft AI (microsoft.ai) - Announced: 2026-10-01 - Category: product-launch - Coverage: 6 outlets - Announcement: yes - Group: covered - Source: https://microsoft.ai/news/our-first-streaming-transcription-model - Record: https://forck.live/items/15607-our-first-streaming-transcription-model-debuts-at-no-1-on-artificial-analysis - Subject: MAI Microsoft AI launched MAI-Transcribe-2-Streaming, a real-time transcription model supporting 60 languages with automatic language detection, ranking first on Artificial Analysis for accuracy. It also released MAI-Voice-2.1, a multilingual text-to-speech model supporting 23 languages and 26 locales, and MAI-Voice-2.1-Flash, a faster variant with 150ms latency and lower pricing. The transcription model is available at an introductory price of $0.54 per hour of audio through year-end, while the voice models are priced at $22 and $15 per 1M characters respectively. ### Apple — RLTL;DR: Self-Improvement by Internalizing Self-Generated Feedback - Company: Apple (apple.com) - Announced: 2026-10-01 - Category: research-paper - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://machinelearning.apple.com/research/rltl-dr-self-improvement - Record: https://forck.live/items/15591-rltl-dr-self-improvement-by-internalizing-self-generated-feedback - Subject: Machine Learning Research Apple researchers introduce RLTL;DR, a method where after each failed attempt, the policy sees the verifier outputs and writes its own TL;DR insight, conditioning subsequent rollouts on all previous insights. On challenging tool-calling and coding datasets where standard GRPO training of a Qwen 3.5 9B Thinking policy achieves 0–1% Pass@1, RLTL;DR achieves 14–31% Pass@1 with insights in context during training and 12–13% when no insight is in context at evaluation. The paper also presents SFTL;DR, training only on (task, insight) tuples, which recovers almost full performance from only 4k tuples. ### Apple — How Much of a Harness Does a Strong Agent Need for Autonomous ML Engineering? - Company: Apple (apple.com) - Announced: 2026-10-01 - Category: research-paper - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://machinelearning.apple.com/research/harness-autonomous-ml-engineering - Record: https://forck.live/items/15582-how-much-of-a-harness-does-a-strong-agent-need-for-autonomous-ml-engineering - Subject: Machine Learning Research Apple researchers found that, under equal time and using the same frontier LLM, a minimal-harness coding agent baseline matches or outperforms open-source state-of-the-art harnesses on current MLE benchmarks, suggesting the backbone model is the primary driver of performance. ### Perplexity — Perplexity and American Express: AI for growing businesses - Company: Perplexity (perplexity.ai) - Announced: 2026-10-01 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://www.perplexity.ai/hub/blog/perplexity-and-american-express-make-ai-easier-for-growing-businesses - Record: https://forck.live/items/15577-perplexity-and-american-express-ai-for-growing-businesses - Subject: Perplexity Perplexity announced ready-to-use Skills for Perplexity Computer curated for U.S. American Express Business Card Members with a Perplexity Enterprise subscription. The collection provides pre-built AI workflows for tasks such as tax prep, cash flow forecasting, and marketing campaign generation. Eligible Card Members can link their American Express Business Card through Plaid to access the Skills. ### Perplexity — AI Tools for Consultants: Features, Pricing & Alternatives - Company: Perplexity (perplexity.ai) - Announced: 2026-10-01 - Category: not stated - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://www.perplexity.ai/hub/blog/ai-tools-for-consultants - Record: https://forck.live/items/15570-ai-tools-for-consultants-features-pricing-alternatives - Subject: Perplexity AI tools help consultants speed up research, analyze data, synthesize client documents, draft deliverables, capture meeting notes, keep client knowledge organized, automate repetitive work, build proposals, manage time, and generate new leads. This gives them more time to focus on what they're paid for: applying their expertise and judgment to solve clients' problems. Perplexity is one of them. Consultants use it for client deep dives, market sizing, proposal development, final-pass document review, and as a general AI assistant. Every Deep Research answer comes with inline citations to the sources it pulled from, making it easier to verify accuracy. Projects keep all chats, files, and custom instructions for each client in one place, and Perplexity Computer , its AI Agent, completes complex multi-step tasks. But we’re not going to focus on Perplexity. What are the other tools that should be at the top of your consideration list? We’ve covered 10 other AI tools for consulting, organized by use case. Each entry lists what the tool does, its core features, current pricing, and alternative tools worth considering. 10 Best AI Tools for Consultants at a Glance … ### Anthropic — Barclays scales Claude to upgrade operations and improve client experience - Company: Anthropic (anthropic.com) - Announced: 2026-10-01 - Category: partnership-acquisition - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://www.anthropic.com/news/barclays-scales-claude - Record: https://forck.live/items/15556-barclays-scales-claude-to-upgrade-operations-and-improve-client-experience - Subject: Claude Barclays is expanding its collaboration with Anthropic to deploy Claude across the bank, aiming to accelerate software development, modernize legacy systems, and improve operational efficiency. The bank expects Claude Code adoption to reach 50% of its developer population by end of 2026 and a majority of software engineers by 2027. Already, over 16,000 colleagues use a Claude-powered knowledge assistant that has handled over one million searches, and Claude models process approximately 120,000 emails daily in the Global Markets business. ### Luma — Live Now: Luma Variants - Company: Luma (lumalabs.ai) - Announced: 2026-10-01 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://lumalabs.ai/news/introducing-luma-variants - Record: https://forck.live/items/15537-live-now-luma-variants - Subject: Ray / Dream Machine Luma launched Variants, a feature that lets users upload one approved static ad and automatically generate versions for multiple placements and languages. The tool supports five standard ad formats and translation into selected markets, keeping logos, headlines, and CTAs intact. It is available now inside Luma's Discover tab. ### Google DeepMind — Gemini 4 Argon: our next era of frontier intelligence - Company: Google DeepMind (deepmind.google) - Announced: 2026-09-30T20:01:45+00:00 - Category: not stated - Coverage: 65 outlets - Announcement: yes - Group: covered - Source: https://deepmind.google/blog/gemini-4-argon-our-next-era-of-frontier-intelligence/ - Record: https://forck.live/items/15475-gemini-4-argon-our-next-era-of-frontier-intelligence - Subject: Gemini Gemini 4 Argon delivers frontier performance in complex workflows across real-world software engineering, enterprise knowledge work like legal and finance, and cybersecurity defense. Today, we’re announcing our new frontier model, Gemini 4 Argon, which is rolling out to a set of trusted cyber defenders through our Fairwind Program . Built to sustain deep reasoning across complex, long-horizon workflows, Argon is fundamentally changing the way we work and build at Google. It delivers frontier performance in complex workflows across real-world software engineering, enterprise knowledge work like legal and finance, and cybersecurity defense. Safely releasing frontier capabilities at this level requires a phased approach. We are actively engaged in the U.S. government’s voluntary process for pre-release model access while we gradually expand access. We’ll continue to gather feedback from early testers as we iterate on guardrails before making Argon available to developers, enterprises, and consumers as soon as possible. … ### Runway — Introducing Runway Ads - Company: Runway (runway.com) - Announced: 2026-09-30 - Category: not stated - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://runway.com/news/company-news/introducing-runway-ads - Record: https://forck.live/items/15446-introducing-runway-ads - Subject: Gen / Aleph An autonomous engine for performance marketing. Today, we're announcing Runway Ads, a new product that runs and optimizes the creative side of paid advertising programs end to end. Connect an ad account and a brand kit, and Runway Ads generates video and image creative, publishes approved variants directly to Meta, Google and TikTok, reads performance back from those same platforms and produces the next round of creative based on what earned spend. Performance advertising rewards volume, but most advertisers are putting meaningful spend behind only a small fraction of the ads they create. Without constant iteration and consistent spend, most marketing teams aren’t getting maximum value from the programs they’re running. Finding the ads that work is less a matter of taste or judgment, and more about how many variants a team can produce. Almost regardless of scale, companies are capped by their ability to produce enough creative, not their analytics or intuition. Runway Ads solves that problem by providing a complete solution for creating, monitoring and iterating on performance ad campaigns. … ### NVIDIA — NVIDIA Opens Applications for 2027–2028 Graduate Fellowships With Awards Up to $60,000 - Company: NVIDIA (nvidia.com) - Announced: 2026-09-30T17:00:21+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://blogs.nvidia.com/blog/applications-open-graduate-fellowship-awards-2026/ - Record: https://forck.live/items/15441-nvidia-opens-applications-for-2027-2028-graduate-fellowships-with-awards-up-to - Subject: AI platform NVIDIA has opened applications for its 2027-2028 Graduate Fellowship Program, offering awards up to $60,000 per student. The program supports doctoral students in AI, machine learning, autonomous vehicles, computer graphics, robotics, healthcare, high-performance computing, and related fields. The application deadline is October 30, 2026, and an in-person internship at an NVIDIA research office in summer 2027 is mandatory. ### Runway — Introducing Praxis-1 - Company: Runway (runway.com) - Announced: 2026-09-30T17:00:00+00:00 - Category: open-weight-release - Coverage: not counted - Announcement: yes - Group: models - Source: https://runway.com/research/introducing-praxis-1 - Record: https://forck.live/items/15447-introducing-praxis-1 - Subject: Gen / Aleph Runway announced Praxis-1, an open-weight world action model that uses video pretraining to control robots. The model is being tested with early partners and will be released publicly in the coming months. Runway will ship it with open weights rather than as a closed model. ### Microsoft Research — Forecasting space weather risks on power grids - Company: Microsoft Research (microsoft.com) - Announced: 2026-09-30T16:00:00+00:00 - Category: research-paper - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://www.microsoft.com/en-us/research/blog/forecasting-space-weather-risks-on-power-grids/ - Record: https://forck.live/items/15418-forecasting-space-weather-risks-on-power-grids - Subject: Research / Phi Microsoft Research developed a machine learning pipeline that forecasts space-weather risk for 66,935 substations in the continental United States, using solar-wind data and local geological factors to provide location-specific risk estimates 30–60 minutes ahead of potential impact. During evaluation, the system detected 76.5% of major events (≥10 nT/min), 81.2% of severe events (≥20 nT/min), and 64.1% of extreme events (≥50 nT/min). ### Lovable — A vulnerability in TanStack Start: what we found and what we did to protect your apps - Company: Lovable (lovable.dev) - Announced: 2026-09-30T15:45:00+00:00 - Category: safety-policy-update - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://lovable.dev/blog/how-lovable-protects-your-app-from-a-tanstack-start-vulnerability - Record: https://forck.live/items/15417-a-vulnerability-in-tanstack-start-what-we-found-and-what-we-did-to-protect - Subject: Lovable / AI app builder Lovable's security team discovered a vulnerability in TanStack Start, a framework used by Lovable apps, and reported it to maintainers. Firewall protections were deployed for hosted apps, and affected projects are automatically updated when users make changes. No evidence of exploitation was found. ### Amazon — Query claims in natural language with Amazon Bedrock Knowledge Bases - Company: Amazon (amazon.com) - Announced: 2026-09-30T15:37:15+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://aws.amazon.com/blogs/machine-learning/query-claims-in-natural-language-with-amazon-bedrock-knowledge-bases/ - Record: https://forck.live/items/15413-query-claims-in-natural-language-with-amazon-bedrock-knowledge-bases - Subject: Bedrock / Nova Amazon's technical guide explains how to use Amazon Bedrock Knowledge Bases with AgenticRetrieveStream to build a claims assistant that answers natural-language questions from claim documents stored in Amazon S3, with metadata filtering and contextual grounding guardrails. ### Amazon — Build a multi-agent music production pipeline on Amazon Bedrock AgentCore Runtime Instances - Company: Amazon (amazon.com) - Announced: 2026-09-30T15:21:57+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://aws.amazon.com/blogs/machine-learning/build-a-multi-agent-music-production-pipeline-on-amazon-bedrock-agentcore-runtime-instances/ - Record: https://forck.live/items/15408-build-a-multi-agent-music-production-pipeline-on-amazon-bedrock-agentcore - Subject: Bedrock / Nova Amazon Bedrock AgentCore adds Runtime Instances, a persistent compute option for long-running multi-agent workflows, alongside the existing serverless MicroVMs. This guide demonstrates a three-agent music production pipeline that uses shared sessions, GPUs, persistent volumes, and colocation of multiple agents on a single instance. ### Google DeepMind — Introducing SynthID Bio - Company: Google DeepMind (deepmind.google) - Announced: 2026-09-30T15:03:07+00:00 - Category: research-paper - Coverage: 6 outlets - Announcement: yes - Group: covered - Source: https://deepmind.google/blog/introducing-synthid-bio/ - Record: https://forck.live/items/15402-introducing-synthid-bio - Subject: Gemini Google DeepMind introduced SynthID Bio, a family of watermarking methods for AI-generated proteins. The approach embeds a detectable signature into amino acid sequences or 3D structures without compromising biological function in laboratory tests. SynthID Bio is intended to strengthen biosecurity by enabling verification of synthetic biological designs and preserving the integrity of public databases. ### Microsoft — Responsible infrastructure at hyperscale: Managing the full lifecycle of Azure hardware - Company: Microsoft (microsoft.com) - Announced: 2026-09-30T15:00:00+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://azure.microsoft.com/en-us/blog/responsible-infrastructure-at-hyperscale-managing-the-full-lifecycle-of-azure-hardware/ - Record: https://forck.live/items/15568-responsible-infrastructure-at-hyperscale-managing-the-full-lifecycle-of-azure - Subject: Azure AI Microsoft describes its approach to managing the full lifecycle of Azure hardware, including design, operation, and decommissioning. The company reports a 92% reuse and recycling rate for decommissioned servers and components, and it has expanded its Circular Centers network to eight facilities across North America, Europe, and Asia Pacific, with a new center planned for San Antonio, Texas. The post also notes that since 2014, cores per rack have increased approximately 13-fold while the power required for the same task has decreased by roughly 90%. ### Google DeepMind — We’re introducing SynthID Bio, bringing our watermarking technology to synthetic biology. - Company: Google DeepMind (deepmind.google) - Announced: 2026-09-30T15:00:00+00:00 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://blog.google/innovation-and-ai/models-and-research/google-deepmind/synthid-bio/ - Record: https://forck.live/items/15403-we-re-introducing-synthid-bio-bringing-our-watermarking-technology-to - Subject: Gemini Google DeepMind introduced SynthID Bio, a watermarking technology for AI-designed proteins that embeds an imperceptible, verifiable watermark into biological designs such as protein sequences and predicted 3D structures. In laboratory tests, watermarked designs matched the performance and natural diversity of unwatermarked versions, aiming to strengthen biosecurity and preserve the integrity of open scientific databases. ### NVIDIA — From Training to Production, NVIDIA and CoreWeave Close the Loop on Agentic AI - Company: NVIDIA (nvidia.com) - Announced: 2026-09-30T15:00:00+00:00 - Category: not stated - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://blogs.nvidia.com/blog/coreweave-agentic-ai-vera-rubin/ - Record: https://forck.live/items/15400-from-training-to-production-nvidia-and-coreweave-close-the-loop-on-agentic-ai - Subject: AI platform Building on nearly a decade of co-engineering, CoreWeave has built NVIDIA compute, networking and software into a cloud purpose-built for AI that’s still returning on investment across multiple generations of deployment. Now, CoreWeave is bringing the next generation of NVIDIA infrastructure to production. At CoreWeave Fully Connected, running this week in San Francisco, CoreWeave announced availability of NVIDIA Vera Rubin NVL72 systems with Spectrum-X 102.4T Ethernet networking. Cognition, the applied AI lab behind the Devin AI software engineer, is the first customer running production workloads on Vera Rubin. CoreWeave will also offer NVIDIA Vera , the first CPU built for AI agents. In addition, CoreWeave launched CoreWeave Forge, a connected environment for training, evaluating and improving models and agents on NVIDIA accelerated computing. “NVIDIA accelerated computing delivers value across generations,” said Ian Buck, vice president of hyperscale and high-performance computing at NVIDIA. “CoreWeave’s NVIDIA V100 GPUs are still running customer workloads nearly a decade after Volta launched, even as CoreWeave brings Vera Rubin NVL72 into production. … ### GitHub — HydraFusion in VS Code and the GitHub Copilot app - Company: GitHub (github.com) - Announced: 2026-09-30T14:31:19+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://github.blog/changelog/2026-09-30-hydrafusion-in-vs-code-and-the-github-copilot-app - Record: https://forck.live/items/15450-hydrafusion-in-vs-code-and-the-github-copilot-app - Subject: Copilot The HydraFusion research preview is now available in Visual Studio Code and the GitHub Copilot app, expanding beyond Copilot CLI. HydraFusion appears in the model picker, but rather than being a single model, it orchestrates multiple models. HydraFusion treats workflow selection as an optimization problem. It uses capability signals for reasoning, code generation, debugging, and tool use to select the most efficient execution pattern to meet the quality bar. HydraFusion uses one of three workflows: Single: One selected model directly solves the task. Cascade: An efficient model drafts a solution and a quality gate decides whether to accept it or escalate to a stronger model. Critique: One model drafts a result, an independent read-only critic from a different model family reviews it—following the same review pattern as Rubber Duck —and the drafting model revises once. In VS Code (version 1.140 or later, or VS Code Insiders): Select HydraFusion from the Copilot Chat model picker. If it does not appear, enable chat.copilot.hydraFusion.enabled . … ### ElevenLabs — ElevenLabs valuation increases to $22 billion fueled by enterprise demand for conversational agents - Company: ElevenLabs (elevenlabs.io) - Announced: 2026-09-30T12:00:00+00:00 - Category: not stated - Coverage: 13 outlets - Announcement: yes - Group: covered - Source: https://elevenlabs.io/blog/tender-22bn - Record: https://forck.live/items/15399-elevenlabs-valuation-increases-to-22-billion-fueled-by-enterprise-demand-for - Subject: Voice AI We recently closed a $300 million employee tender offer that values ElevenLabs at $22 billion, double our valuation at our Series D in February. The round was led by Wellington and T. Rowe Price, long-term institutional investors with the experience our next stage of growth requires. In 2022, we raised our first round at a $9 million valuation, which we announced alongside Eleven v1, our first Text to Speech model, and the first model to cross the uncanny valley of speech. Four years later, as we reach our new valuation, we just released two new state-of-the-art Text to Speech models. Eleven v4 and v4 Turbo are our fastest and most emotive models yet, which are already leading independent benchmarks . Alongside models, we have built a full interaction platform for AI. It allows organizations to deploy conversational agents, connected to their knowledge and systems, that interact naturally in any industry, geography, or context. This shift is driving much of our growth, with enterprise now accounting for 55% of our revenue, as businesses adopt ElevenAgents to support their customers and employees. … ### OpenAI — Disrupting a coordinated model-distillation campaign - Company: OpenAI (openai.com) - Announced: 2026-09-30T10:30:00+00:00 - Category: not stated - Coverage: 10 outlets - Announcement: no - Group: covered - Source: https://openai.com/index/disrupting-a-coordinated-model-distillation-campaign - Record: https://forck.live/items/15445-disrupting-a-coordinated-model-distillation-campaign - Subject: GPT / ChatGPT / API We recently identified and disrupted a coordinated campaign designed to extract protected reasoning from our models, with the earliest observed activity occurring in the first week of July. This activity is consistent with adversarial distillation: the systematic and unauthorized use of one model’s outputs or reasoning to help train, reproduce, or improve another model. Protected reasoning is the model’s internal record for working through a task; extracting it can reveal information withheld from the final answer and help others reproduce the model’s capabilities. The operators did not break our encryption, compromise a database, or gain direct access to stored user conversations. Instead, they manipulated model interactions so that protected reasoning could be reproduced in forms visible to the requester in a coordinated, scaled manner that violated our terms of service. This manipulation is not a vulnerability unique to OpenAI’s models, and we have shared information about it with industry partners through the Frontier Model Forum in order to strengthen collective defenses against adversarial distillation. … ### OpenAI — Helping small businesses put AI to work - Company: OpenAI (openai.com) - Announced: 2026-09-30T10:00:00+00:00 - Category: partnership-acquisition - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://openai.com/index/helping-small-businesses-put-ai-to-work - Record: https://forck.live/items/15414-helping-small-businesses-put-ai-to-work - Subject: GPT / ChatGPT / API OpenAI is partnering with America's SBDC to train around 150 small business advisors and host hands-on workshops, aiming to reach at least 1,000 small businesses. A new report finds that in a week in September 2026, about 4 million small-firm employees used OpenAI's tools, and agentic output tokens for small businesses rose from one-third to two-thirds of all output tokens between April and August 2026. ### Amazon — Amazon Bedrock expands Claude model availability to in-country inferencing in India - Company: Amazon (amazon.com) - Announced: 2026-09-30T01:13:14+00:00 - Category: availability-change - Coverage: 35 outlets - Announcement: yes - Group: covered - Source: https://aws.amazon.com/blogs/machine-learning/amazon-bedrock-expands-claude-model-availability-to-india-cross-region-inference/ - Record: https://forck.live/items/15167-amazon-bedrock-expands-claude-model-availability-to-in-country-inferencing-in - Subject: Bedrock / Nova Amazon Bedrock now offers Anthropic's Claude Opus 5, Claude Sonnet 5, and Claude Haiku 4.5 through an India regional endpoint using geographic cross-Region inference, allowing customers to process data locally within India while accessing these models. ### Amazon — Introducing Anthropic models on Amazon Bedrock for in-region inference in Seoul and Singapore - Company: Amazon (amazon.com) - Announced: 2026-09-30T01:13:12+00:00 - Category: availability-change - Coverage: 4 outlets - Announcement: yes - Group: covered - Source: https://aws.amazon.com/blogs/machine-learning/introducing-anthropic-models-on-amazon-bedrock-for-in-region-inference-in-seoul-and-singapore/ - Record: https://forck.live/items/15168-introducing-anthropic-models-on-amazon-bedrock-for-in-region-inference-in - Subject: Bedrock / Nova Amazon Bedrock now offers in-region inference for Anthropic's Claude Opus 5 and Claude Sonnet 5 in Seoul and Claude Sonnet 5 in Singapore, enabling data processing entirely within the specified region to meet data residency requirements. The service is available on the bedrock-runtime endpoint and supports the Anthropic Messages API, Amazon Bedrock InvokeModel and Converse APIs, along with features like Guardrails and intelligent prompt routing. ### Apple — SCLATE: A Substrate for Continual-Learning Agent Training and Evaluation - Company: Apple (apple.com) - Announced: 2026-09-30 - Category: not stated - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://machinelearning.apple.com/research/sclate-agent-training-evaluation - Record: https://forck.live/items/15485-sclate-a-substrate-for-continual-learning-agent-training-and-evaluation - Subject: Machine Learning Research Authors Youngmok Jung, Sirajul Salekin, Henry Tran, Javier Movellan, Zhao Huang, Manjot Bilkhu Continual-learning agents are systems of models, harnesses, and memory operating over long multi-session horizons. Evaluating and training them requires interleaving tasks with agent-side events such as session stop and start, crons, and memory consolidation. Yet existing benchmarks and training frameworks schedule only the benchmark’s own events, leaving each benchmark and agent pair to build a custom scheduling loop. We present SCLATE, an execution substrate where benchmarks and unmodified agents each add their events to one open event scheduler through an adapter. A hybrid simulated clock runs these events on a shared timeline, flowing in real time while the agent works and skipping idle gaps, which compresses a month-long scenario into hours. SCLATE also serves as a rollout engine that runs any agent’s harness and memory unmodified, recording the tokens and log probabilities of every model call through an in-container proxy. We port seven benchmarks to SCLATE and compare ten unmodified harness and memory configurations head to head on ten models. … ### Anthropic — Claude for Government is now generally available - Company: Anthropic (anthropic.com) - Announced: 2026-09-30 - Category: availability-change - Coverage: 6 outlets - Announcement: yes - Group: covered - Source: https://claude.com/blog/claude-for-government-is-now-generally-available - Record: https://forck.live/items/15444-claude-for-government-is-now-generally-available - Subject: Claude Anthropic made Claude for Government generally available to U.S. federal and state agencies, offering coding and agentic capabilities through a FedRAMP High authorized environment. The platform includes purpose-built governance controls, audit logs, and usage-based billing with no seat fees. Claude Code CLI and Claude for Microsoft 365 are also available in early access through the same environment. ### Perplexity — Contextual embedding beyond the gold passage - Company: Perplexity (perplexity.ai) - Announced: 2026-09-30 - Category: new-model - Coverage: not counted - Announcement: yes - Group: models - Source: https://www.perplexity.ai/hub/blog/contextual-embedding-beyond-the-gold-passage - Record: https://forck.live/items/15439-contextual-embedding-beyond-the-gold-passage - Subject: Perplexity Perplexity released pplx-embed-v2-context-9b-preview, a contextual embedding model preview on Hugging Face. It uses a context compression model as a teacher to train chunk-level relevance scores, aiming to retrieve both answer chunks and supporting context instead of a single gold chunk. The model achieves state-of-the-art results on context-bench and ConTEB benchmarks. ### Apple — On the Effectiveness-Fluency Trade-Off in LLM Conditioning: A Systematic Study - Company: Apple (apple.com) - Announced: 2026-09-30 - Category: research-paper - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://machinelearning.apple.com/research/effectiveness-fluency-llm-conditioning - Record: https://forck.live/items/15422-on-the-effectiveness-fluency-trade-off-in-llm-conditioning-a-systematic-study - Subject: Machine Learning Research Apple researchers systematically studied conditioning methods for LLMs, finding that efficient steering often degrades fluency and that activation steering is less effective on instruction-tuned models than on base models. Simple prompting and supervised fine-tuning work for concept injection but not removal, and cheap textual metrics correlate well with costly LLM-as-judge scores. ### Hugging Face — Open TTS Leaderboard: Scalable Evaluation for Multilingual Text-to-Speech and Voice Cloning - Company: Hugging Face (huggingface.co) - Announced: 2026-09-30 - Category: not stated - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://huggingface.co/blog/open-tts-leaderboard - Record: https://forck.live/items/15397-open-tts-leaderboard-scalable-evaluation-for-multilingual-text-to-speech-and - Subject: Platform The pace of open-source text-to-speech (TTS) model releases has been incredible. On the Hugging Face Hub (as of Sep 30, 2026) there are more than 8K TTS models available 🚀 Evaluation, however, hasn't kept pace: it remains fragmented and unstandardized. The gold standard is human preference scores such as MOS or MUSHRA (more on metrics ). To this end, several arena-based leaderboards have established themselves as useful reference points for the community: These arenas compare models by presenting users with TTS outputs from two models, and asking them to choose one over the other. After collecting a sufficient number of votes, an Elo score is computed to rank models, typically with the Bradley–Terry model (see Voice Arena methodology ). While human preference is the ultimate decider, arenas cannot scale to keep up with the pace of TTS releases . This may partly explain why open-source models are underrepresented on arena-style leaderboards: as of Sep 30, 2026, only 16 of the 92 models on Artificial Analysis are open-weights, with a similar skew on Voice Arena . … ### Anthropic — How Anthropic's sales team rebuilt inbound with Claude Managed Agents - Company: Anthropic (anthropic.com) - Announced: 2026-09-30 - Category: not stated - Coverage: 26 outlets - Announcement: yes - Group: covered - Source: https://claude.com/blog/how-anthropics-sales-team-rebuilt-inbound-with-claude-managed-agents - Record: https://forck.live/items/15389-how-anthropic-s-sales-team-rebuilt-inbound-with-claude-managed-agents - Subject: Claude Carl Johnson, a sales development leader at Anthropic, shares how a Claude-powered buying agent now answers most inbound customers, and how that changed the way our sales team works. Category Enterprise AI Product Claude Platform Date September 30, 2026 Reading time 5 min Share Copy link https://claude.com/blog/how-anthropics-sales-team-rebuilt-inbound-with-claude-managed-agents Author(s) Carl Johnson As a sales leader, it pains me to admit that not long ago, people who wanted to buy Claude for their company weren’t getting the answers they needed quickly enough. They had filled out our Contact Sales form but would wait too long to hear back, sometimes for multiple days. Most of their questions were simple: what a plan costs, whether there's a seat minimum, or whether we can meet HIPAA’s contract requirements. The answers were in our documentation and support articles, but customers wanted someone to walk them through, quickly. So we built a buying agent on Claude Managed Agents (beta) that takes a prospect from "I want Claude for my company" to a completed purchase. … ### Cohere — RCP-nDCG@10: a more complete way to measure retrieval relevance - Company: Cohere (cohere.com) - Announced: 2026-09-30 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://cohere.com/blog/rcp-ndcg - Record: https://forck.live/items/15383-rcp-ndcg-10-a-more-complete-way-to-measure-retrieval-relevance - Subject: Command Our new standard for measuring enterprise retrieval quality, validated against human judgment. How well does nDCG capture the quality of modern retrieval systems? nDCG is the established yardstick for measuring how well search systems rank results, and is widely used across benchmarks such as MTEB and BEIR . It is preferred because it reflects both relevance and position, giving more weight to useful results that appear near the top. However, it relies on pre-existing relevance labels, which, if incomplete, mean other relevant results can be missed or scored incorrectly. Here, we introduce Rubric-Calibrated Preferences nDCG@10 (RCP-nDCG@10), an evaluation methodology that grades each retrieved document against the same explicit relevance criteria for every query, using a calibrated AI judge. Designed to better reflect real-world enterprise retrieval quality, RCP-nDCG@10 is the metric against which we have optimised our next generation search models. Here we explain how it works, what's different from the status quo, how we validated it against human judgment. nDCG is only as good as its labels … ### Cohere — Introducing Embed 5—A New Family of Frontier Embedding Models - Company: Cohere (cohere.com) - Announced: 2026-09-30 - Category: not stated - Coverage: 1 outlet - Announcement: yes - Group: covered - Source: https://cohere.com/blog/embed-5 - Record: https://forck.live/items/15363-introducing-embed-5-a-new-family-of-frontier-embedding-models - Subject: Command Our most powerful embedding models yet, now available in Pro and Fast tiers. Key takeaways State-of-the-art enterprise retrieval: Embed 5 Pro achieves the highest average score of any model we tested - particularly across financial datasets, parsed PDFs, and visually rich documents. A new Fast tier: Embed 5 Fast brings strong retrieval quality to latency - and cost-sensitive workloads, at $0.08 per million tokens. One index, two models: Pro and Fast share an embedding space, so teams can index with Pro and query with either model without re-indexing. Built for complex enterprise data: Embed 5 supports multimodal inputs and retrieval, 100+ languages, and a 128K-token context window for longer documents. More efficient at scale: Matryoshka representations and lower-precision outputs reduce vector storage and search costs, while quantized weights lower serving requirements for private deployments. Today, we're releasing Embed 5, a new family of embeddings models at the frontier of high-quality enterprise retrieval. Embed 5 delivers stronger retrieval across complex enterprise data while giving teams more control over latency, cost, and deployment. … ### Amazon — Bring near-Astra intelligence to everyday work with GPT-6.1 Sol on Amazon Bedrock - Company: Amazon (amazon.com) - Announced: 2026-09-29T19:34:14+00:00 - Category: not stated - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://aws.amazon.com/blogs/machine-learning/bring-near-astra-intelligence-to-everyday-work-with-gpt-6-1-sol-on-amazon-bedrock/ - Record: https://forck.live/items/15068-bring-near-astra-intelligence-to-everyday-work-with-gpt-6-1-sol-on-amazon - Subject: Bedrock / Nova GPT-6.1 Sol is now generally available on Amazon Bedrock, bringing stronger reasoning to coding, computer use, and professional workloads that run frequently. For an AI agent to complete a task, it may need to gather information, use tools, test different approaches, recover from errors, and verify its result. Every decision shapes what happens next. A wrong turn can add model interactions, tool calls, latency, and human intervention before the agent reaches a useful result. The economics of an AI agent take shape across the entire task. Token prices influence the cost of each interaction, while reasoning quality influences how many interactions the work requires and whether they lead to a successful outcome. The total cost of completing a task depends on both. Today, GPT-6.1 Sol is generally available on Amazon Bedrock, running on an inference engine built for performance, security, and reliability at scale. A major upgrade to GPT-6 Sol , it delivers strong performance on agentic coding, computer use, and professional work. According to OpenAI, it brings near- Astra intelligence to everyday workflows. Apply stronger reasoning across agentic work … ### Perplexity — GPT-6.1 Sol - Company: Perplexity (perplexity.ai) - Announced: 2026-09-29T19:18:00+00:00 - Category: api-release - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://docs.perplexity.ai/docs/resources/changelog#september-2026 - Record: https://forck.live/items/15058-gpt-6-1-sol - Subject: Sonar Perplexity's Agent API now supports the openai/gpt-6.1-sol model. ### Google Research — How Diffusion Controller unifies and simplifies AI image generation - Company: Google Research (research.google) - Announced: 2026-09-29T18:38:27+00:00 - Category: not stated - Coverage: 1 outlet - Announcement: yes - Group: covered - Source: https://research.google/blog/how-diffusion-controller-unifies-and-simplifies-ai-image-generation/ - Record: https://forck.live/items/15057-how-diffusion-controller-unifies-and-simplifies-ai-image-generation - Subject: Research We introduce Diffusion Controller, a lightweight "steering damper" network that precisely steers image generation to achieve significantly better prompt alignment. It seamlessly attaches to even access-restricted, closed-source models, boosting image quality without breaking baseline stability. Quick links The rapid advancement of text-to-image AI models, such as Nano Banana , Stable Diffusion and Flux , has fundamentally transformed creative design, allowing anyone to synthesize photorealistic, high-fidelity images from textual descriptions. However, steering these massive models to meet precise user intent, downstream goals, or strict visual constraints remains a delicate and unpredictable balancing act. For example, imagine prompting a model for "a lizard wearing sunglasses". The model might generate a realistic lizard that's not wearing sunglasses. Alternatively, forcing the model to include the sunglasses might distort the lizard's face, ruining the image quality. Existing methodologies that guide or fine-tune image generation are very disconnected. … ### GitHub — GPT-6.1 Sol in GitHub Copilot - Company: GitHub (github.com) - Announced: 2026-09-29T17:02:27+00:00 - Category: model-update - Coverage: 6 outlets - Announcement: yes - Group: models - Source: https://github.blog/changelog/2026-09-29-gpt-6-1-sol-in-github-copilot - Record: https://forck.live/items/15048-gpt-6-1-sol-in-github-copilot - Subject: Copilot GitHub announced the general availability of OpenAI's GPT-6.1 Sol model in GitHub Copilot. The model supports agentic coding and terminal workflows, and in early testing it completed tasks using fewer tokens and steps than earlier GPT-6 and GPT-5.6 family models. It is available to Copilot Pro+, Max, Business, and Enterprise users across multiple IDEs and platforms, with a gradual rollout. ### Hume — Evaluating Google’s multi-speaker TTS: A case study in why private evaluations matter - Company: Hume (hume.ai) - Announced: 2026-09-29T17:02:00+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://www.hume.ai/blog/evaluating-multi-speaker-tts - Record: https://forck.live/items/15012-evaluating-google-s-multi-speaker-tts-a-case-study-in-why-private-evaluations - Subject: EVI / Octave ### Microsoft — SQL Server on Azure Local is now generally available - Company: Microsoft (microsoft.com) - Announced: 2026-09-29T16:40:06+00:00 - Category: not stated - Coverage: 6 outlets - Announcement: yes - Group: covered - Source: https://www.microsoft.com/en-us/sql-server/blog/2026/09/28/sql-server-on-azure-local-is-now-generally-available/ - Record: https://forck.live/items/15504-sql-server-on-azure-local-is-now-generally-available - Subject: Azure AI Some of the world’s most critical databases run in places where cloud connectivity cannot be assumed. A remote industrial site, for example, may need production systems to continue operating when external connectivity is unavailable. A regulated organization may need sensitive data and processing to remain within sovereign boundaries. Today, SQL Server on Azure Local is generally available for connected and disconnected operations. With this release, customers can modernize where their data resides while maintaining control over infrastructure, connectivity, and data placement. They can bring AI closer to their data with Foundry Local on Azure Local, currently in preview, and use eligible existing SQL Server licensing investments. Explore SQL Server on Azure Local Run SQL Server where your data needs to stay SQL Server has a long history of supporting mission-critical workloads on customer infrastructure. Azure Local builds on that flexibility by bringing Azure infrastructure to customer-owned environments, giving organizations a consistent platform for running SQL Server across datacenters and edge locations. … ### Amazon — Prompt engineering fundamentals for Amazon Quick - Company: Amazon (amazon.com) - Announced: 2026-09-29T16:27:57+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://aws.amazon.com/blogs/machine-learning/prompt-engineering-fundamentals-for-amazon-quick/ - Record: https://forck.live/items/15007-prompt-engineering-fundamentals-for-amazon-quick - Subject: Bedrock / Nova Amazon's guide explains foundational prompt engineering principles for Amazon Quick, covering specificity, context, few-shot examples, and the CRISPE framework to improve AI output quality. It is Part 1 of a two-part series, with Part 2 planned for component-specific techniques. ### Amazon — Prompt engineering by Quick component: Patterns and pitfalls - Company: Amazon (amazon.com) - Announced: 2026-09-29T16:27:33+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://aws.amazon.com/blogs/machine-learning/prompt-engineering-by-quick-component-patterns-and-pitfalls/ - Record: https://forck.live/items/15008-prompt-engineering-by-quick-component-patterns-and-pitfalls - Subject: Bedrock / Nova Amazon's guide explains how to write effective prompts for each Quick component, covering patterns and pitfalls for Quick Research, Quick Flows, and Quick Sight. It provides concrete examples and techniques, such as specifying objectives, decomposing complex topics, and structuring flow logic as numbered steps. ### Amazon — Building an AI-powered contract intelligence platform with Amazon Quick and Amazon Bedrock AgentCore - Company: Amazon (amazon.com) - Announced: 2026-09-29T16:14:24+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://aws.amazon.com/blogs/machine-learning/building-an-ai-powered-contract-intelligence-platform-with-amazon-quick-and-amazon-bedrock-agentcore/ - Record: https://forck.live/items/15009-building-an-ai-powered-contract-intelligence-platform-with-amazon-quick-and - Subject: Bedrock / Nova Amazon published a guide describing a contract intelligence platform built on AWS that uses AI agents (Claude Sonnet for extraction, Claude Haiku for verification) to extract structured data from PDF contracts, with Amazon Textract as a tiebreaker for signature detection. The system stores results in Aurora PostgreSQL and provides portfolio-wide aggregation via embedded dashboards and natural language queries, addressing the limitation of RAG-based tools that cannot sum or compare across many documents. ### Amazon — How Condé Nast built multimodal video discovery with Amazon Bedrock - Company: Amazon (amazon.com) - Announced: 2026-09-29T15:55:17+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://aws.amazon.com/blogs/machine-learning/how-conde-nast-built-multimodal-video-discovery-with-amazon-bedrock/ - Record: https://forck.live/items/14998-how-cond-nast-built-multimodal-video-discovery-with-amazon-bedrock - Subject: Bedrock / Nova Condé Nast built a multimodal video discovery solution on Amazon Bedrock and OpenSearch Service that reduced editorial video search time from 250 minutes to under 2 minutes per task. The solution uses TwelveLabs Marengo embedding model accessed through Bedrock to index over 140,000 videos by visual, audio, and transcript signals, enabling intent-based semantic search. ### Hugging Face — NVIDIA Kumo Tabular Sets a New Accuracy-Efficiency Frontier for Tabular Prediction - Company: Hugging Face (huggingface.co) - Announced: 2026-09-29T15:30:38+00:00 - Category: not stated - Coverage: 1 outlet - Announcement: yes - Group: covered - Source: https://huggingface.co/blog/nvidia/kumo-tabular - Record: https://forck.live/items/14996-nvidia-kumo-tabular-sets-a-new-accuracy-efficiency-frontier-for-tabular - Subject: Platform NVIDIA Kumo Tabular, part of the NVIDIA Kumo Structured model collection, is an open foundation model for tabular data now available on Hugging Face . Given a table of labeled rows, it predicts the labels of new rows in a single forward pass, with no training, no tuning, and no feature engineering, for both classification and regression. It was pretrained only on artificial data, comes in three sizes (28M to 215M parameters), runs through our open-source library , and is released under the OpenMDW-1.1 license for commercial use. It ranks first on the four benchmarks TabArena , BeyondArena , TALENT and ScoringBench . Model Code: https://github.com/NVIDIA/structured-data-models Model Weights: https://huggingface.co/nvidia/Kumo-Tabular The Shift to Tabular Foundation Models Tabular data is the backbone of enterprise machine learning. Customer records, transactions, sensor logs, claims, and orders all live in tables, and predicting churn, default, demand, or price from them is among the most common machine learning tasks in industry. For two decades, this work has been done with gradient-boosted trees, and it has worked well. … ### Runway — Runway Joins the OpenAI Marketplace - Company: Runway (runway.com) - Announced: 2026-09-29T15:19:00+00:00 - Category: not stated - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://runway.com/news/company-news/runway-openai-marketplace - Record: https://forck.live/items/15053-runway-joins-the-openai-marketplace - Subject: Gen / Aleph The best image, video, audio and editing models are now available through OpenAI Marketplace, using existing spend commitments. Today, Runway is joining the OpenAI Marketplace as a launch partner with Runway Creative, our platform for generating and editing video, images and audio. Eligible enterprise customers can now find Runway Creative in the OpenAI Marketplace, and can apply part of their OpenAI commitment toward purchases with Runway. The OpenAI Marketplace lets enterprise teams put committed OpenAI budget toward approved partner products, so creative, marketing and production teams already working with OpenAI can bring Runway into the same budget. What's in Runway Creative Runway Creative brings image, video, audio and editing models from Runway and other labs into one workspace, including Gen-4.5, Seedance 2.5, GPT Image 2.5 and ElevenLabs V4. With it, teams can: Generate video and images from text, images or existing footage Edit footage by describing the change they want Give Runway Agent a brief and let it plan, generate and assemble a complete video through conversation … ### Perplexity — Claude Sonnet 5.5 - Company: Perplexity (perplexity.ai) - Announced: 2026-09-29T15:17:58+00:00 - Category: not stated - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://docs.perplexity.ai/docs/resources/changelog#september-2026 - Record: https://forck.live/items/14997-claude-sonnet-5-5 - Subject: Sonar The Agent API now supports anthropic/claude-sonnet-5-5 . See the Agent API Models reference . ### Microsoft Research — Introducing Quine: An AI research system designed for the complexity of biology - Company: Microsoft Research (microsoft.com) - Announced: 2026-09-29T14:00:02+00:00 - Category: research-paper - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://www.microsoft.com/en-us/research/blog/introducing-quine-an-ai-research-system-designed-for-the-complexity-of-biology/ - Record: https://forck.live/items/14980-introducing-quine-an-ai-research-system-designed-for-the-complexity-of-biology - Subject: Research / Phi Microsoft Research introduced Quine, a research system that combines a multimodal world model of biology with a harness connecting scientific tools, literature, and researchers. The system was used with the Broad Institute to prioritize compounds for therapeutic tumor-state shifts, with several top candidates validated in wet-lab assays. Quine is experimental research technology not intended for clinical use, and access will be expanded through the Quine Fellows program and eventually products like Microsoft Discovery. ### Hugging Face — Getting the Source Right, Not Just the Fact: Source-Aware Verification for MCP Agents - Company: Hugging Face (huggingface.co) - Announced: 2026-09-29T13:07:00+00:00 - Category: research-paper - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://huggingface.co/blog/MultiverseComputingCAI/getting-the-source-right-not-just-the-fact-source - Record: https://forck.live/items/14976-getting-the-source-right-not-just-the-fact-source-aware-verification-for-mcp - Subject: Platform Hugging Face and Multiverse Computing released a paper introducing ProvenanceGuard, a post-generation verification layer for MCP-based LLM agents that checks whether each claim is supported by the specific source the answer attributes it to, rather than by any source in a pooled evidence set. In tests on 361 claims from a medical agent, it caught 138 of 139 claims that human experts said should not pass, while flagging 67 supported claims for review under a conservative setting. ### OpenAI — DevDay 2026 Recap - Company: OpenAI (openai.com) - Announced: 2026-09-29T10:00:00+00:00 - Category: capability-change - Coverage: not counted - Announcement: no - Group: routine - Source: https://openai.com/index/devday-2026-recap - Record: https://forck.live/items/15041-devday-2026-recap - Subject: GPT / ChatGPT / API OpenAI announced always-on Dots agents, the GPT-6.1 Sol upgrade, and new Codex and ChatGPT developer capabilities at DevDay 2026. The Agents API now supports computer use and brings Codex’s multi-agent capabilities, tool search, tool calling and context compaction into applications. ChatGPT plugin extensions let developers build sidebar experiences, interactive panels and file viewers. ### OpenAI — Introducing GPT-6.1 Sol - Company: OpenAI (openai.com) - Announced: 2026-09-29T10:00:00+00:00 - Category: not stated - Coverage: 16 outlets - Announcement: yes - Group: covered - Source: https://openai.com/index/introducing-gpt-6-1-sol - Record: https://forck.live/items/15040-introducing-gpt-6-1-sol - Subject: GPT / ChatGPT / API Near-Astra intelligence for a fifth of the price We’re introducing GPT‑6.1 Sol , an upgrade to GPT‑6 Sol that nearly matches GPT‑6 Astra’s intelligence on agentic coding, computer use, and professional work at one-fifth of Astra’s standard input and output token prices. Cached input costs just $0.10 per million tokens —95% less than standard input pricing and 50% less than GPT‑6 Sol’s cached input pricing—giving developers more room to build and run capable agents that reuse context across requests. A more capable Sol across tasks GPT‑6.1 Sol offers a new balance of capability and cost for important everyday work. It delivers substantial improvements over GPT‑6 Sol across complex professional tasks, from writing and debugging code to understanding documents and executing multi-step business workflows. On several of these evaluations, it approaches GPT‑6 Astra’s performance at substantially lower cost. Coding On DeepSWE v1.1 , which evaluates complex software-engineering tasks in real codebases, GPT‑6.1 Sol matches GPT‑6 Astra at roughly one-fifth of the cost, while eclipsing GPT‑6 Sol’s best score by 6.4 percentage points at a lower reasoning effort and cost. … ### Microsoft — FabCon and SQLCon 2026 in Barcelona: Building the data foundation for Microsoft Copilot and agents - Company: Microsoft (microsoft.com) - Announced: 2026-09-29T06:30:00+00:00 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://azure.microsoft.com/en-us/blog/fabcon-and-sqlcon-2026-in-barcelona-building-the-data-foundation-for-microsoft-copilot-and-agents/ - Record: https://forck.live/items/14952-fabcon-and-sqlcon-2026-in-barcelona-building-the-data-foundation-for-microsoft - Subject: Azure AI Microsoft announced that Fabric IQ, a shared intelligence layer, is now integrated into Microsoft Copilot, providing business context from Fabric data without additional AI token costs. The integration is generally available in Copilot Chat and Cowork, with Code integration coming soon. Additionally, Microsoft Power BI Desktop is gaining agentic app creation, allowing users to generate applications from semantic models using natural language. ### OpenAI — How we will do better for Australia - Company: OpenAI (openai.com) - Announced: 2026-09-29T01:00:00+00:00 - Category: not stated - Coverage: 10 outlets - Announcement: no - Group: covered - Source: https://openai.com/index/how-we-will-do-better-for-australia - Record: https://forck.live/items/14797-how-we-will-do-better-for-australia - Subject: GPT / ChatGPT / API In June, during internal training and evaluation our models accessed Australian government websites in ways they were not authorised to. We also should have handled our response better. We are sorry and working to do better in the future. In this post, we are setting out what we know, what we have changed, and what we will do to rebuild trust with the Australian people. This is a new kind of cyber incident which represents an emerging global challenge. One of the ways we intend to take accountability for the situation is to be intentional in working with Australia to help develop practical approaches to how AI developers and governments identify, disclose, and respond to AI cyber behaviour, whether malicious or unintentional. When we became aware and how we responded to this incident After the Hugging Face incident in July, we began reviewing earlier training and evaluation activity to identify other affected organisations. In mid-August, that review identified activity affecting the Australian government websites below. Here’s what we know based on the evidence: … ### Manus — Introducing Manus Flex: Use your own API key in the Manus workspace - Company: Manus - Announced: 2026-09-29 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://manus.im/blog/introducing-manus-flex - Record: https://forck.live/items/15625-introducing-manus-flex-use-your-own-api-key-in-the-manus-workspace - Subject: Manus Manus launched Manus Flex, a module that allows users to connect their own API key from a supported inference provider to the Manus agent workspace. Initial inference partners include OpenRouter, Fireworks, and Modal. Model inference is billed directly by the provider, while other Manus services still consume Manus credits. ### Amp — Plaid Speed - Company: Amp (ampcode.com) - Announced: 2026-09-29 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://ampcode.com/news/plaid-mode - Record: https://forck.live/items/15070-plaid-speed - Subject: Amp Amp introduced Plaid speed for modes using GPT-6 Astra, offering up to 6× faster inference at 6× the cost per token. Users can select Plaid in the mode picker for compatible modes, and it works with both Amp-provided inference and linked ChatGPT subscriptions that support ultrafast. ### Suno — Three Must-learn Tips for Using EQ - Company: Suno (suno.com) - Announced: 2026-09-29 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://suno.com/blog/about-eq - Record: https://forck.live/items/15062-three-must-learn-tips-for-using-eq - Subject: Music / Studio In a previous blogpost, I said that compression was one of the most ubiquitous music production tools. And while this is true, there is a clear winner in this particular popularity contest. And the winner is: Equalization , normally referred to as EQ . You’ll find EQ everywhere in music production. Some DAWs, like Logic Pro, have EQ built into every channel, whether you want it or not, ready to grab at a moment's notice. The truth is you normally will want to use it. Seasoned producers will often have several go-to EQs; I for one love to grab Ableton Live’s EQ Three for crude tone shaping, and EQ Eight for fine tuning. But if I want some character, I’ll go for a Pultec, API or SSL emulation to give my tracks more bite. What does EQ actually do? An EQ adjusts the sonic profile of your sounds, by boosting or cutting specific frequencies or ranges of frequencies. Most EQ are parametric, meaning they have several bands , and each band can have a different filter curve , such as high-pass, low-pass, shelving or notch. On each band of a parametric EQ, you’ll be able to adjust the gain , frequency , and bandwidth (Q) to dial in the characteristics you want. … ### OpenAI — Introducing dots - Company: OpenAI (openai.com) - Announced: 2026-09-29 - Category: product-launch - Coverage: 42 outlets - Announcement: yes - Group: covered - Source: https://openai.com/index/introducing-dots - Record: https://forck.live/items/15042-introducing-dots - Subject: GPT / ChatGPT / API OpenAI introduced dots, always-on AI agents powered by GPT-6 Astra that operate on their own cloud computer, connect to over 4,000 apps via plugins, and work autonomously on tasks like bug fixes, content production, and proposal updates. Dots are rolling out across Pro, Business Premium, and Enterprise plans in eligible markets, with a preview of specialist dots for enterprise use. The agents can be accessed via ChatGPT, Slack, and Teams, and include safety and privacy safeguards. ### Perplexity — Computer adds Automations for ongoing work - Company: Perplexity (perplexity.ai) - Announced: 2026-09-29 - Category: not stated - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://www.perplexity.ai/hub/blog/computer-adds-automations-for-ongoing-work - Record: https://forck.live/items/15013-computer-adds-automations-for-ongoing-work - Subject: Perplexity A useful weekly project update builds on the week before, including which blockers were resolved and which decisions are still pending. When Perplexity Computer prepares that update, it should remember where things stood for the previous one. Today we’re launching Automations, which bring Computer’s capabilities to long-running agents that work proactively on a schedule or in response to conditional events. Each agent uses Computer’s connected files, tools , and Projects to carry out its assignment. It remembers prior work, picking up where it left off without waiting for a new request. Unlike other agents, Automations don’t start from scratch every time. For a customer-service Automation, Computer drafts replies with earlier emails in mind. For a weekly competitor-pricing report, Computer compares the latest prices with the previous ones and flags what changed. Schedules or events trigger the automation Automations replace Scheduled Tasks, adding event-based triggers and memory of prior runs to recurring work. Computer can carry out assignments on a schedule or when a specified event occurs in Slack, Gmail, Outlook, Linear, or GitHub. … ### Apple — The Communication Bottleneck: A Round-Trip Study of Tree-Structured Expression Serialization in Language Models - Company: Apple (apple.com) - Announced: 2026-09-29 - Category: research-paper - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://machinelearning.apple.com/research/communication-bottleneck-serialization - Record: https://forck.live/items/14990-the-communication-bottleneck-a-round-trip-study-of-tree-structured-expression - Subject: Machine Learning Research Apple researchers propose a round-trip protocol to measure how much tree-structured compositional content survives when language models serialize expressions into natural language. Evaluating 16 models, they find the channel is lossy and asymmetric, with at least 73.6% of failures originating at generation, and that fine-tuning on ∼3600 examples lifts open-weight models above an untrained frontier model. ### Perplexity — How we engineer safer agents - Company: Perplexity (perplexity.ai) - Announced: 2026-09-29 - Category: not stated - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://www.perplexity.ai/hub/blog/how-we-engineer-safer-agents - Record: https://forck.live/items/14989-how-we-engineer-safer-agents - Subject: Perplexity To build safer AI agents, we need to design for both their capabilities and their failures. Even when carrying out legitimate tasks, and without any malicious instructions, agents may cross security boundaries as they try to overcome obstacles. We believe this is a security engineering problem, and the industry should tackle it the way it tackled the internet worms of the early 2000s. The systems surrounding agents must limit what agents can access, detect unwanted behavior, and contain failures when they occur. At Perplexity, we are actively collaborating with researchers and practitioners to advance security engineering for AI agents, and we apply multiple layers of defense across our models, agent harnesses, and infrastructure so that security does not depend on any single safeguard. Meltdown behavior of AI agents In July, AI agents broke into open-source AI platform Hugging Face's production infrastructure. During internal cybersecurity evaluation and training runs, OpenAI models discovered a way to exploit the Artifactory repository manager to establish a bulletin board for communicating with one another, enabling them to organize into swarms. … ### Anthropic — Agents you can coach: how Asana builds human-agent teams with Claude - Company: Anthropic (anthropic.com) - Announced: 2026-09-29 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://claude.com/blog/agents-you-can-coach-how-asana-builds-human-agent-teams-with-claude - Record: https://forck.live/items/14987-agents-you-can-coach-how-asana-builds-human-agent-teams-with-claude - Subject: Claude Arnab Bose, Chief Product Officer at Asana, on how Asana runs AI agents as teammates with scoped roles, shared memory, and work that everyone can see. Category Agents Product Claude Platform Date September 29, 2026 Reading time 5 min Share Copy link https://claude.com/blog/agents-you-can-coach-how-asana-builds-human-agent-teams-with-claude This is the third post in our series on building human-agent teams. The first shared what we’ve learned working with multiplayer AI at Anthropic. The second shared how Slack turns workplace conversation into the context agents need. This one looks at what changes when agents operate on the same platform where teams work. Years before they introduced AI agents, teams at Asana were iterating on ways to encode structure and accountability into how teams work together. They ultimately built the W ork Graph ® model, which maps out every task, project, goal, and conversation on a web of relationships, with defined owners, contributors, and dependencies. When they started building AI agents, they decided that rather than adding new context structures for AI, agents would operate within this same model. … ### Amazon — Grok 4.7 is now available on Amazon Bedrock - Company: Amazon (amazon.com) - Announced: 2026-09-28T22:13:16+00:00 - Category: new-model - Coverage: not counted - Announcement: yes - Group: models - Source: https://aws.amazon.com/blogs/machine-learning/grok-4-7-is-now-available-on-amazon-bedrock/ - Record: https://forck.live/items/14755-grok-4-7-is-now-available-on-amazon-bedrock - Subject: Bedrock / Nova xAI's Grok 4.7 is now available on Amazon Bedrock, offering a 500K token context window and configurable reasoning effort at four levels. According to xAI, it is their most capable model for coding and knowledge work, designed to work longer on difficult tasks and verify its own output. The model accepts text and image input, returns text, and supports the Responses, Chat Completions, and Converse APIs through cross-Region inference profiles. ### Microsoft Research — One year in: How Microsoft Research Asia – Singapore is advancing research, partnership and talent for real-world impact - Company: Microsoft Research (microsoft.com) - Announced: 2026-09-28T21:00:00+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://www.microsoft.com/en-us/research/blog/one-year-in-how-microsoft-research-asia-singapore-is-advancing-research-partnership-and-talent-for-real-world-impact/ - Record: https://forck.live/items/14751-one-year-in-how-microsoft-research-asia-singapore-is-advancing-research - Subject: Research / Phi Microsoft Research Asia – Singapore marks its first anniversary, highlighting progress in building partnerships with Singapore's government, academia, and industry to advance frontier AI research and translate it into real-world applications, particularly in healthcare. ### Perplexity — xhigh preset uses Claude Opus 5.5 - Company: Perplexity (perplexity.ai) - Announced: 2026-09-28T19:16:57+00:00 - Category: model-update - Coverage: not counted - Announcement: yes - Group: models - Source: https://docs.perplexity.ai/docs/resources/changelog#september-2026 - Record: https://forck.live/items/14744-xhigh-preset-uses-claude-opus-5-5 - Subject: Sonar Perplexity's Agent API xhigh preset now uses Claude Opus 5.5 instead of GPT-5.6-sol. Prompts, reasoning effort, tools, token budgets, and step limits remain unchanged. ### OpenAI — Towards safety cases for frontier AI training - Company: OpenAI (openai.com) - Announced: 2026-09-28T19:00:00+00:00 - Category: safety-policy-update - Coverage: 1 outlet - Announcement: no - Group: covered - Source: https://openai.com/index/towards-safety-cases-for-frontier-ai-training - Record: https://forck.live/items/14842-towards-safety-cases-for-frontier-ai-training - Subject: GPT / ChatGPT / API OpenAI published initial guidelines for safety cases that should be required before frontier reinforcement learning training runs. The framework covers technical safeguards including model alignment, containment, and monitoring, with specific practices such as automated dataset reviews, containment red-teaming, and immutable transcripts. OpenAI treats safety cases as an aspirational goal and invites community feedback on the evolving best practices. ### Amazon — Introducing Claude Sonnet 5.5 on AWS - Company: Amazon (amazon.com) - Announced: 2026-09-28T18:57:13+00:00 - Category: not stated - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://aws.amazon.com/blogs/machine-learning/introducing-claude-sonnet-5-5-on-aws/ - Record: https://forck.live/items/14738-introducing-claude-sonnet-5-5-on-aws - Subject: Bedrock / Nova Today, we’re excited to announce the availability of Claude Sonnet 5.5 on Amazon Bedrock and Claude Platform on AWS . Claude Sonnet 5.5 is a smarter, more efficient Sonnet model suited for focused coding and knowledge work with lower cost per task for most work at faster speed. Amazon Bedrock gives you Sonnet 5.5 capabilities while keeping your data within AWS infrastructure with Regional data residency. It works with the AWS controls your team already uses, including AWS Identity and Access Management (IAM) for access, AWS CloudTrail for audit, Amazon CloudWatch for monitoring, and Amazon Bedrock Guardrails. Usage appears on your AWS bill. This post covers Claude Sonnet 5.5’s improvements, practical guidance on when to choose Sonnet, and how to get started on Amazon Bedrock. What makes Claude Sonnet 5.5 different The improvements stand out on well-scoped work. Whether a developer assigns it a feature or a bug fix, Sonnet 5.5 completes the work and checks the result against the stated requirements. It also produces more polished documents and visuals than Sonnet 5, including one-pagers, architecture diagrams, and summary slides. … ### GitHub — Claude Sonnet 5.5 in GitHub Copilot - Company: GitHub (github.com) - Announced: 2026-09-28T18:03:57+00:00 - Category: not stated - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://github.blog/changelog/2026-09-28-claude-sonnet-5-5-in-github-copilot - Record: https://forck.live/items/14726-claude-sonnet-5-5-in-github-copilot - Subject: Copilot Claude Sonnet 5.5, Anthropic’s newest Sonnet model, is now generally available in GitHub Copilot. It is designed for well-scoped everyday work like building features and fixing bugs. In our early testing, Sonnet 5.5 stood out for its efficiency, matching Claude Sonnet 5 on coding tasks while using significantly fewer steps, tokens, and tool calls. It also finished tasks noticeably faster. This model is billed at provider list pricing under usage-based billing. See Models and pricing for GitHub Copilot for details. Claude Sonnet 5.5 is available to Copilot Pro, Pro+, Max, Business, and Enterprise users. You can select the model in the model picker in: Visual Studio Code Visual Studio Copilot CLI GitHub Copilot coding agent GitHub Copilot app github.com GitHub Mobile on iOS and Android JetBrains IDEs Xcode Eclipse Rollout will be gradual. Check back soon if you don’t see it yet. Copilot Enterprise and Copilot Business plan administrators can manage access to Claude Sonnet 5.5 through the model policy in Copilot settings. … ### Amazon — Build real-time voice applications with vLLM-Omni on SageMaker AI – Part 1 - Company: Amazon (amazon.com) - Announced: 2026-09-28T16:15:46+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://aws.amazon.com/blogs/machine-learning/build-real-time-voice-applications-with-vllm-omni-on-sagemaker-ai-part-1/ - Record: https://forck.live/items/14684-build-real-time-voice-applications-with-vllm-omni-on-sagemaker-ai-part-1 - Subject: Bedrock / Nova Voice agents, interactive learning applications, accessibility tools, and customer service assistants need to respond without long silent pauses. In this tutorial, you deploy a text-to-speech (TTS) model on Amazon SageMaker AI that can start playing speech before it finishes generating the full response. You use the AWS vLLM-Omni Deep Learning Container (DLC) to deploy Qwen3-TTS , stream text in and audio out over one persistent bidirectional connection, and try the workflow through a Gradio application. AWS Deep Learning Containers provide Docker images with deep learning frameworks and dependencies for training and inference on AWS. AWS provides deployment guidance for broadly adopted serving frameworks such as vLLM and SGLang . This post is Part 1 of a series about specialized DLCs, including vLLM-Omni , WhisperX , and llama.cpp . It focuses on streamed speech for real-time voice applications. Part 2 applies the vLLM-Omni DLC to image and video generation. The series pairs focused use cases with deployment examples and reproducible benchmarks where they add useful evidence. … ### Amazon — Generate images and video with vLLM-Omni on SageMaker AI – Part 2 - Company: Amazon (amazon.com) - Announced: 2026-09-28T16:15:17+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://aws.amazon.com/blogs/machine-learning/generate-images-and-video-with-vllm-omni-on-sagemaker-ai-part-2/ - Record: https://forck.live/items/14685-generate-images-and-video-with-vllm-omni-on-sagemaker-ai-part-2 - Subject: Bedrock / Nova Amazon's guide shows how to deploy two SageMaker AI endpoints from a single vLLM-Omni DLC: a real-time endpoint for FLUX.2-klein-4B image generation and an asynchronous endpoint for Wan2.1-VACE-1.3B video generation. The workflow sends a text prompt to generate an image, then passes that image with a motion prompt to the video endpoint, retrieving the MP4 from Amazon S3. The sample includes a command-line workflow and a Streamlit application. ### Mistral AI — Hallo, Deutschland! - Company: Mistral AI (mistral.ai) - Announced: 2026-09-28T15:57:59+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://mistral.ai/news/hallo-deutschland/ - Record: https://forck.live/items/14687-hallo-deutschland - Subject: Mistral / Le Chat Mistral AI opened a new hub in Munich, Germany, to focus on Physics AI and Industrial AI for European heavy industry. The hub will house research teams and applied engineers, and the company announced plans to build one gigawatt of European compute capacity by 2030. Mistral also formed a research partnership with the Technical University Munich (TUM) to develop digital twins for automotive aerodynamics. ### Amazon — Implementing synthetic monitoring using Amazon Nova Act - Company: Amazon (amazon.com) - Announced: 2026-09-28T15:56:32+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://aws.amazon.com/blogs/machine-learning/implementing-synthetic-monitoring-using-amazon-nova-act/ - Record: https://forck.live/items/14686-implementing-synthetic-monitoring-using-amazon-nova-act - Subject: Bedrock / Nova Amazon published a guide on implementing synthetic monitoring using Amazon Nova Act and Amazon Bedrock AgentCore. The approach replaces traditional DOM-selector-based browser automation with a multimodal LLM that processes UI screenshots, aiming to reduce script brittleness and maintenance overhead. The post includes architecture, sample code, and scheduling via Amazon EventBridge Scheduler. ### Amazon — Automating Amazon Textract adapter lifecycle management across accounts - Company: Amazon (amazon.com) - Announced: 2026-09-28T15:49:04+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://aws.amazon.com/blogs/machine-learning/automating-amazon-textract-adapter-lifecycle-management-across-accounts/ - Record: https://forck.live/items/14667-automating-amazon-textract-adapter-lifecycle-management-across-accounts - Subject: Bedrock / Nova Amazon published a guide on automating Amazon Textract adapter lifecycle management across accounts, covering adapter promotion, document routing, and production security. The post provides infrastructure templates (CloudFormation and Terraform) and describes externalizing adapter IDs into AWS Systems Manager Parameter Store to enable zero-downtime updates without application redeployment. ### Cognition — Devin Mobile is now available in beta. - Company: Cognition (cognition.com) - Announced: 2026-09-28T15:43:11+00:00 - Category: availability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://devin.ai/ios - Record: https://forck.live/items/14718-devin-mobile-is-now-available-in-beta - Subject: Devin / SWE Cognition is accepting waitlist sign-ups for Devin Mobile. ### Microsoft — Enhancing Microsoft Azure Virtual Machine lifecycle - Company: Microsoft (microsoft.com) - Announced: 2026-09-28T15:00:00+00:00 - Category: not stated - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://azure.microsoft.com/en-us/blog/enhancing-microsoft-azure-virtual-machine-lifecycle/ - Record: https://forck.live/items/14953-enhancing-microsoft-azure-virtual-machine-lifecycle - Subject: Azure AI Azure continuously advances through new infrastructure generations, platform capabilities, services, and innovations that lead the industry in performance, security, reliability, and efficiency. Our Virtual Machine (VM) Lifecycle policy guides how we manage these transitions, giving Azure customers transparency, predictability, and guidance to plan for and execute technology changes with confidence. Our goal is to help every organization plan and navigate these modernization steps in line with Azure’s latest innovations and ongoing platform investments. Learn more about virtual machines in Azure To give you access to the most advanced cloud and AI infrastructure, deployed on the most sophisticated silicon and hardware, the Azure Virtual Machine lifecycle is built around three customer outcomes: Transparency to easily understand the lifecycle status of Azure technologies, Microsoft’s level of ongoing investment, and the recommended path forward as new innovations become available. Predictability to plan against defined lifecycle stages, understand what to expect at each stage, and advance notice as technologies progress through their lifecycle. … ### Lovable — Lovable + Microsoft: run the apps you build inside your company's Microsoft tenant - Company: Lovable (lovable.dev) - Announced: 2026-09-28T15:00:00+00:00 - Category: partnership-acquisition - Coverage: 8 outlets - Announcement: yes - Group: covered - Source: https://lovable.dev/blog/microsoft-partnership - Record: https://forck.live/items/14663-lovable-microsoft-run-the-apps-you-build-inside-your-company-s-microsoft-tenant - Subject: Lovable / AI app builder Lovable is partnering with Microsoft so that apps built with Lovable can be published into a company's own Microsoft tenant via the Copilot Managed Runtime SDK, with sign-in through Microsoft Entra ID and management like any other Microsoft app. The integration also lets Lovable apps read and write Microsoft 365, Fabric, Dataverse, and SQL data without exporting or copying it. Copilot Managed Runtime is rolling out in public preview, while the Microsoft 365 connectors, Fabric, and Microsoft sign-in are available on every Lovable plan today. ### ElevenLabs — Introducing Eleven v4, our most emotive model - Company: ElevenLabs (elevenlabs.io) - Announced: 2026-09-28T12:00:00+00:00 - Category: new-model - Coverage: not counted - Announcement: yes - Group: models - Source: https://elevenlabs.io/blog/eleven-v4 - Record: https://forck.live/items/14660-introducing-eleven-v4-our-most-emotive-model - Subject: Voice AI ElevenLabs launched Eleven v4, a text-to-speech model designed to interpret tone, pacing, emotion, character, and context, alongside a low-latency variant Eleven v4 Turbo with median inference latency of ~100ms. Both models support over 90 languages, improved voice cloning with 10 seconds of audio, and more accurate audio tags and direction prompts. The company states Eleven v4 was ranked #1 by Artificial Analysis and preferred by ~75% of listeners in blind tests over competing models. ### Tencent — Four Ways to Help International Visitors Feel at Home in China - Company: Tencent (tencent.com) - Announced: 2026-09-28T09:49:28+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://www.tencent.com/four-ways-to-help-international-visitors-feel-at-home-in-china/ - Record: https://forck.live/items/14993-four-ways-to-help-international-visitors-feel-at-home-in-china - Subject: Hunyuan Tencent describes four technology projects aimed at helping international visitors in Shenzhen ahead of the 2026 APEC meeting: an offline translator (Tencent Hy Translate), a menu translation and cultural explanation feature in TenPay Go, a tool to standardize inconsistent English signage, and a service called Knowa that maps international apps to local equivalents. ### Hugging Face — Holo4: powering generalist computer-use agents - Company: Hugging Face (huggingface.co) - Announced: 2026-09-28T09:44:05+00:00 - Category: new-model - Coverage: 1 outlet - Announcement: yes - Group: models - Source: https://huggingface.co/blog/Hcompany/holo4 - Record: https://forck.live/items/14610-holo4-powering-generalist-computer-use-agents - Subject: Platform Hugging Face released Holo4, a new series of agentic models in two sizes (27B dense and 35B-A3B MoE), available via the H Models API. The models interact with software through GUIs, code, MCP, and APIs, and are trained via supervised and reinforcement learning on diverse environments and tasks. An updated version of Holotron 3, Holotron4 Nano, was also released. ### Google Antigravity — Custom agents in Google plugins - Company: Google Antigravity (antigravity.google) - Announced: 2026-09-28T01:11:14+00:00 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://antigravity.google/blog/custom-agents-in-google-plugins - Record: https://forck.live/items/14747-custom-agents-in-google-plugins - Subject: Antigravity Google Antigravity announced that teams at Google are packaging custom agents into official plugins available through Build with Google, and previewed three plugins with custom agent support: a Flutter & Dart accessibility auditing agent, a Firebase Security Rules agent for authoring and hardening Firestore security rules, and a Google Play pre-submission release auditing agent for Android. The post includes a detailed workflow for the Play release auditor, which operates in read-only mode to inspect policy compliance, intent security, and R8/DEX health. ### Amp — Opus 5.5 - Company: Amp (ampcode.com) - Announced: 2026-09-28 - Category: model-update - Coverage: not counted - Announcement: yes - Group: models - Source: https://ampcode.com/news/opus-5.5 - Record: https://forck.live/items/14910-opus-5-5 - Subject: Amp Amp replaced GPT-5.6 Sol with Claude Opus 5.5 as the default model for its medium mode. In internal evaluations, Opus 5.5 solved 65% of tasks (up from GPT-5.6 Sol's 61% and Opus 5's 56%) at 10% lower cost than GPT-5.6 Sol and 25% lower than Opus 5. The model is set to high reasoning effort by default, as higher settings degrade performance and increase cost. ### Perplexity — MCP vs. API: Definitions, Differences, and Use Cases - Company: Perplexity (perplexity.ai) - Announced: 2026-09-28 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://www.perplexity.ai/hub/blog/mcp-vs-api - Record: https://forck.live/items/14843-mcp-vs-api-definitions-differences-and-use-cases - Subject: Perplexity Perplexity's guide defines the Model Context Protocol (MCP) as an open standard enabling AI agents to access external systems, contrasting it with traditional APIs that connect applications via fixed endpoints. MCP allows LLMs to discover and use up-to-date tools at runtime without client-side updates, while APIs remain preferable for high-volume data transfers and fixed workflows. ### World Labs — World Labs is Joining AMD - Company: World Labs (worldlabs.ai) - Announced: 2026-09-28 - Category: partnership-acquisition - Coverage: 23 outlets - Announcement: yes - Group: covered - Source: https://www.worldlabs.ai/blog/amd-announcement - Record: https://forck.live/items/14753-world-labs-is-joining-amd - Subject: Marble / Atlas World Labs has signed a definitive agreement to join AMD. Dr. Fei-Fei Li will become an Executive Vice President and Chief Scientist at AMD, working with CEO Dr. Lisa Su. Justin Johnson and Ben Mildenhall will continue leading the World Labs team as it joins AMD to form a frontier research organization focused on an open AI ecosystem. The transaction is expected to close by the end of 2026, subject to regulatory approvals. ### xAI — Team Bots: shared AI teammates that learn as they work - Company: xAI (x.ai) - Announced: 2026-09-28 - Category: capability-change - Coverage: 3 outlets - Announcement: yes - Group: covered - Source: https://x.ai/news/team-bots - Record: https://forck.live/items/14742-team-bots-shared-ai-teammates-that-learn-as-they-work - Subject: Grok xAI launched Team Bots, which are shared Grok Bots that teams can give access to files, apps, and expertise, and that learn from team interactions. Each Team Bot can be configured with context, plugins, credentials, and memories, and can be used in Slack with its own handle. The announcement includes examples of use at SpaceXAI for sales, engineering, marketing, and data analytics. ### OpenAI — Basis completes a tax workbook 2x faster with GPT-6 Astra - Company: OpenAI (openai.com) - Announced: 2026-09-28 - Category: not stated - Coverage: 24 outlets - Announcement: no - Group: covered - Source: https://openai.com/index/basis-tax-workbook-with-astra - Record: https://forck.live/items/14741-basis-completes-a-tax-workbook-2x-faster-with-gpt-6-astra - Subject: GPT / ChatGPT / API Basis, which builds AI agents for accountants, reported that GPT-6 Astra completed a 50-tab tax workbook in half the time of GPT-5.6 Sol and improved its internal evaluation scores by about 20%. The company attributed the gains to GPT-6 Astra's better understanding of user intent and its ability to adjust reasoning effort during tasks. ### Anthropic — Giving companies more control over their AI agents, with NVIDIA - Company: Anthropic (anthropic.com) - Announced: 2026-09-28 - Category: partnership-acquisition - Coverage: 75 outlets - Announcement: yes - Group: covered - Source: https://claude.com/blog/giving-companies-more-control-over-their-ai-agents-with-nvidia - Record: https://forck.live/items/14739-giving-companies-more-control-over-their-ai-agents-with-nvidia - Subject: Claude NVIDIA announced the Open Agent Safety Platform, an open software platform and reference system design for strengthening AI security. Anthropic collaborated with NVIDIA to bring additional layers of security and control to the agent stack. Claude Managed Agents holds credentials in a vault so the agent never sees them, while NVIDIA OpenShell software controls what the agent can execute and reach, with rules enforced outside the agent and logged. ### Perplexity — Agent API now supports reusable agents - Company: Perplexity (perplexity.ai) - Announced: 2026-09-28 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://www.perplexity.ai/hub/blog/agent-api-now-supports-reusable-agents - Record: https://forck.live/items/14732-agent-api-now-supports-reusable-agents - Subject: Perplexity Perplexity updated its Agent API with reusable Profiles, versioned Skills, and managed connectors, allowing teams to configure an agent once and reuse it across applications. Project administrators can share these resources, and members can invoke them via API keys from the same project. Managed Connectors are in preview with support for GitHub, Slack, Google Drive, Datadog, Linear, and Notion. ### Anthropic — Introducing Claude Sonnet 5.5 - Company: Anthropic (anthropic.com) - Announced: 2026-09-28 - Category: model-update - Coverage: 23 outlets - Announcement: yes - Group: models - Source: https://www.anthropic.com/claude-sonnet-5-5 - Record: https://forck.live/items/14728-introducing-claude-sonnet-5-5 - Subject: Claude Anthropic released Claude Sonnet 5.5, the second model in the Claude 5.5 family, as a faster and lower-cost complement to Claude Opus 5.5. It runs 30%+ faster than Sonnet 5 and costs up to 30% less per task, with the same per-token pricing but fewer tokens needed. Sonnet 5.5 scores 70.6% on Terminal-Bench 4.0, up from Sonnet 5's 10.3%, and is the first Sonnet model to launch with cyber safeguards and fallbacks. ### Manus — Introducing Manus 2.0 - Company: Manus - Announced: 2026-09-28 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://manus.im/blog/introducing-manus-2-0 - Record: https://forck.live/items/14699-introducing-manus-2-0 - Subject: Manus Manus 2.0 is here, with the new Cascade agent harness, Cloud Computer and Automations, Manus Studio for professional creation, and Cue, a new app for personal agents. ### Apple — Faster Rates for Federated Variational Inequalities - Company: Apple (apple.com) - Announced: 2026-09-28 - Category: research-paper - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://machinelearning.apple.com/research/federated-variational-inequalities - Record: https://forck.live/items/14689-faster-rates-for-federated-variational-inequalities - Subject: Machine Learning Research Apple researchers present a paper studying federated optimization for stochastic variational inequalities. They provide improved convergence rates for the Local Extra SGD algorithm and propose a new algorithm, LIPPAX, which reduces client drift and achieves better guarantees in bounded Hessian, bounded operator, and low-variance settings. The work also extends results to federated composite variational inequalities.