# forck.live — 31 August – 6 September 2026

> AI lab announcements, news reports and podcasts.

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 one issue of the weekly archive: every confirmed
first-party announcement of ISO week 2026-W36, 31 August – 6 September 2026, newest
first. The window never moves, so this URL always names the same seven days —
but a post published inside them and detected later is added when it arrives,
so a closed week can still gain entries.

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-W36
- The current issue, and the catalogue of the other weeks: https://forck.live/briefing.md
- Catalogue: https://forck.live/llms.txt

## Announcements

### OpenAI — An Alien Mind

- Company: OpenAI (openai.com)
- Announced: 2026-09-06T09:00:00+00:00
- Category: safety-policy-update
- Coverage: 3 outlets
- Announcement: yes
- Group: covered
- Source: https://openai.com/index/an-alien-mind
- Record: https://forck.live/items/8549-an-alien-mind
- Subject: GPT / ChatGPT / API

OpenAI reflects on the challenge of AI alignment as AI becomes increasingly capable, and calls for stronger safeguards and international coordination.

### OpenAI — Research acceleration: The view inside OpenAI

- Company: OpenAI (openai.com)
- Announced: 2026-09-06T08:00:00+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://openai.com/index/research-acceleration-view-inside-openai
- Record: https://forck.live/items/8547-research-acceleration-the-view-inside-openai
- Subject: GPT / ChatGPT / API

OpenAI shared early data on how coding agents are being used internally to accelerate AI research, focusing on agent usage, experiment velocity, and task complexity.

### Amazon — Deploy a multimodal WhatsApp ordering assistant with Amazon Bedrock AgentCore

- Company: Amazon (amazon.com)
- Announced: 2026-09-04T21:45:52+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://aws.amazon.com/blogs/machine-learning/deploy-a-multimodal-whatsapp-ordering-assistant-with-amazon-bedrock-agentcore/
- Record: https://forck.live/items/8396-deploy-a-multimodal-whatsapp-ordering-assistant-with-amazon-bedrock-agentcore
- Subject: Bedrock / Nova

Amazon published a guide for deploying a multimodal WhatsApp ordering assistant using Amazon Bedrock AgentCore and Amazon Nova 2 models. The solution integrates WhatsApp Business Platform with AWS services to handle text, voice notes, and voice calls through a single backend with cross-channel memory.

### GitHub — GitHub Copilot weekly releases — August 31

- Company: GitHub (github.com)
- Announced: 2026-09-04T21:05:01+00:00
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://github.blog/changelog/2026-09-04-github-copilot-weekly-releases-august-31
- Record: https://forck.live/items/8397-github-copilot-weekly-releases-august-31
- Subject: Copilot

GitHub Copilot expands model choice with Claude Fable 5.1 and Gemini 3.8 Flash, adds content exclusions, makes the Copilot harness generally available in JetBrains, introduces Agent Merge in public preview, and adds experimental multi-root workspaces, chat backgrounds, and chat sessions.

### Perplexity — Sign in with Perplexity for the remote MCP server

- Company: Perplexity (perplexity.ai)
- Announced: 2026-09-04T20:05:35+00:00
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://docs.perplexity.ai/docs/resources/changelog#september-2026
- Record: https://forck.live/items/9601-sign-in-with-perplexity-for-the-remote-mcp-server
- Subject: Sonar

Perplexity's remote MCP server now supports OAuth authentication, allowing users to sign in with their Perplexity account instead of pasting an API key. The feature works with MCP clients that support OAuth, such as claude.ai, Claude Code, Cursor, or VS Code. API keys remain available for clients without OAuth support.

### Databricks — Achieving Extreme Efficiency through Specialized GPU Kernel Generation

- Company: Databricks (databricks.com)
- Announced: 2026-09-04T20:00:00+00:00
- Category: research-paper
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://www.databricks.com/blog/achieving-extreme-efficiency-through-specialized-gpu-kernel-generation
- Record: https://forck.live/items/17494-achieving-extreme-efficiency-through-specialized-gpu-kernel-generation
- Subject: Mosaic AI

Databricks built Proteus, a system that uses agents to generate specialized GPU kernels tailored to runtime shapes. The system produced Qwen 3.5 122B kernels that were 1.8-5.2x faster than the best implementations available in vLLM. The blog details challenges in validation and context management for agentic kernel generation.

### Runway — Introducing Team Plan

- Company: Runway (runway.com)
- Announced: 2026-09-04
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://runway.com/news/company-news/introducing-team-plan
- Record: https://forck.live/items/8385-introducing-team-plan
- Subject: Gen / Aleph

Runway launched a Team plan with shared projects, a pooled credit balance (6,900 credits per seat per month), up to 100 shared projects, Brand Kits, 1TB of storage, and collaboration features like commenting and @mentions. Standard, Pro and Max become individual plans; existing multi-member workspaces are unaffected.

### GitHub — GPT-6 Astra is generally available in GitHub Copilot

- Company: GitHub (github.com)
- Announced: 2026-09-04T18:59:51+00:00
- Category: availability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://github.blog/changelog/2026-09-04-gpt-6-astra-is-generally-available-in-github-copilot
- Record: https://forck.live/items/8394-gpt-6-astra-is-generally-available-in-github-copilot
- Subject: Copilot

GitHub announced that GPT-6 Astra from OpenAI is now generally available in GitHub Copilot. The model is designed for long-horizon, autonomous coding and agentic tasks and is available to Copilot Pro+, Max, Business, and Enterprise users.

### Amazon — Designing lifecycle policies for AgentCore memory

- Company: Amazon (amazon.com)
- Announced: 2026-09-04T17:20:04+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://aws.amazon.com/blogs/machine-learning/designing-lifecycle-policies-for-agentcore-memory/
- Record: https://forck.live/items/8388-designing-lifecycle-policies-for-agentcore-memory
- Subject: Bedrock / Nova

Amazon Web Services introduced memory lifecycle management for AI agents using AgentCore memory on Amazon Bedrock, with a deployable architecture using AWS Step Functions and Amazon Bedrock to run nightly lifecycle workflows. The solution includes TTL-based expiration, relevance decay scoring, and consolidation policies to manage agent memories.

### Amazon — Build a Physical AI model factory with NVIDIA Cosmos 3 on SageMaker HyperPod

- Company: Amazon (amazon.com)
- Announced: 2026-09-04T16:16:00+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://aws.amazon.com/blogs/machine-learning/build-a-physical-ai-model-factory-with-nvidia-cosmos-3-on-sagemaker-hyperpod/
- Record: https://forck.live/items/8382-build-a-physical-ai-model-factory-with-nvidia-cosmos-3-on-sagemaker-hyperpod
- Subject: Bedrock / Nova

NVIDIA Cosmos 3, an open omnimodal world model released under the OpenMDW-1.1 license, is used with Amazon SageMaker HyperPod to build a Physical AI model factory that runs a continuous pipeline of synthetic data generation, model post-training, and closed-loop simulation on a single persistent GPU cluster, leveraging the model's Mixture-of-Transformers architecture with per-layer joint attention.

### Amazon — Run agent-driven Amazon SageMaker HyperPod operations with InstantStart

- Company: Amazon (amazon.com)
- Announced: 2026-09-04T16:12:17+00:00
- Category: developer-tool-release
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://aws.amazon.com/blogs/machine-learning/run-agent-driven-amazon-sagemaker-hyperpod-operations-with-instantstart/
- Record: https://forck.live/items/8379-run-agent-driven-amazon-sagemaker-hyperpod-operations-with-instantstart
- Subject: Bedrock / Nova

Amazon announced HyperPod InstantStart, an open source control plane for Amazon SageMaker HyperPod that provides a web interface and an AI agent to automate multi-stage cluster operations such as bootstrap, capacity, training, inference, and storage. The system runs as a single out-of-band management container in the user's AWS account and uses MCP tools that wrap the backend's own REST APIs.

### Amazon — Customizing your knowledge base on Amazon Bedrock for large and complex documents using Amazon Textract

- Company: Amazon (amazon.com)
- Announced: 2026-09-04T16:08:10+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://aws.amazon.com/blogs/machine-learning/customizing-your-knowledge-base-on-amazon-bedrock-for-large-and-complex-documents-using-amazon-textract/
- Record: https://forck.live/items/8380-customizing-your-knowledge-base-on-amazon-bedrock-for-large-and-complex
- Subject: Bedrock / Nova

Amazon Bedrock and Amazon Textract are integrated to extract and analyze information from large, complex documents such as utility bills, enabling a Retrieval Augmented Generation (RAG) solution for customer service teams.

### Amazon — How Intuit built an agentic disaster recovery assistant with Amazon Bedrock

- Company: Amazon (amazon.com)
- Announced: 2026-09-04T16:06:01+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://aws.amazon.com/blogs/machine-learning/how-intuit-built-an-agentic-disaster-recovery-assistant-with-amazon-bedrock/
- Record: https://forck.live/items/8381-how-intuit-built-an-agentic-disaster-recovery-assistant-with-amazon-bedrock
- Subject: Bedrock / Nova

Intuit built an AI-powered disaster recovery assistant called EWOK Agent using Amazon Bedrock to automate failover decision-making for its microservices across AWS Regions.

### Google Cloud — Spanner migrations: Automating dual-write with Antigravity CLI for minimal disruption

- Company: Google Cloud (cloud.google.com)
- Announced: 2026-09-04T16:00:00+00:00
- Category: not stated
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://cloud.google.com/blog/topics/developers-practitioners/using-antigravity-cli-to-streamline-dual-write-database-migration/
- Record: https://forck.live/items/17333-spanner-migrations-automating-dual-write-with-antigravity-cli-for-minimal
- Subject: Gemini Enterprise

When Google's Finance Engineering team needed to modernize their legacy data layer, they chose Spanner , a globally distributed, strongly consistent, multi-model database with high availability capabilities.

### Perceptron — Perceptron Multilook API

- Company: Perceptron (perceptron.inc)
- Announced: 2026-09-04
- Category: api-release
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://www.perceptron.inc/blog/introducing-perceptron-multilook-api
- Record: https://forck.live/items/18609-perceptron-multilook-api
- Subject: Perceptron Isaac / Perceptron Mk1

Perceptron released the Multilook API, an endpoint that prefills a shared media context once and reuses it across up to 16 prompts in a single call. The shared context is billed at the full input rate once, with each additional prompt read at a 25% cache rate, and extra samples costing only output tokens. The API is designed for workloads where the same visual asset is queried multiple times, such as fixed query sets or in-context learning with visual examples.

### Perplexity — Fast Embeddings on GPUs

- Company: Perplexity (perplexity.ai)
- Announced: 2026-09-04
- Category: not stated
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://www.perplexity.ai/hub/blog/fast-embeddings-on-gpus
- Record: https://forck.live/items/10756-fast-embeddings-on-gpus
- Subject: Perplexity

Fast and accurate search is vital to all of Perplexity, from Search and Computer to our API Platform. Behind the scenes, the heavy lifting is done by embedding and ranking models, which help our systems identify the most relevant results for a given query.

### Microsoft AI — Pushing the quality-cost frontier with MAI-Image-2.6

- Company: Microsoft AI (microsoft.ai)
- Announced: 2026-09-04
- Category: not stated
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://microsoft.ai/news/pushing-the-quality-cost-frontier-with-mai-image-2-6
- Record: https://forck.live/items/8412-pushing-the-quality-cost-frontier-with-mai-image-2-6
- Subject: MAI

MAI-Image-2.6 is our strongest image model yet. Today, we’re bringing it to developers in Microsoft Foundry and expanding the family with MAI-Image-2.6-Flash – built to deliver that same level of quality for latency-sensitive, high-throughput production workloads.

### xAI — Setting Grok Bot loose on procurement

- Company: xAI (x.ai)
- Announced: 2026-09-04
- Category: not stated
- Coverage: 1 outlet
- Announcement: no
- Group: covered
- Source: https://x.ai/news/grok-bot-procurement
- Record: https://forck.live/items/8393-setting-grok-bot-loose-on-procurement
- Subject: Grok

xAI released the system prompt for Haggle Bot, a procurement savings agent designed to analyze vendor spend and produce evidence-backed savings recommendations.

### Amp — Desktop

- Company: Amp (ampcode.com)
- Announced: 2026-09-04
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://ampcode.com/news/desktop
- Record: https://forck.live/items/8258-desktop
- Subject: Amp

Amp added a Desktop tab to its orb threads, providing a high-resolution interactive Linux desktop. The desktop can be used for verifying agent work, developing for Linux/Android, computer use, CAD, and as a cloud desktop.

### Midjourney — Collecting ideas to improve Midjourney

- Company: Midjourney (midjourney.com)
- Announced: 2026-09-03T19:44:24+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://updates.midjourney.com/collecting-ideas-to-improve-midjourney/
- Record: https://forck.live/items/18578-collecting-ideas-to-improve-midjourney
- Subject: Midjourney V8

Midjourney is soliciting user feedback and ideas to determine its next priorities, with a formal voting session planned within the next week or two.

### Midjourney — Alpha Changelog - 9/2/26

- Company: Midjourney (midjourney.com)
- Announced: 2026-09-03T19:43:14+00:00
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://updates.midjourney.com/alpha-changelog-9-2-26/
- Record: https://forck.live/items/18579-alpha-changelog-9-2-26
- Subject: Midjourney V8

Midjourney pushed a large update to alpha.midjourney.com, adding the v8.2 edit model with a new editor built into the lightbox. The editor accepts plain-language instructions and up to four reference images, and a new Change Style feature is being tested for exploring style space.

### GitHub — Upcoming deprecation of selected GitHub Copilot models

- Company: GitHub (github.com)
- Announced: 2026-09-03T19:40:14+00:00
- Category: deprecation
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://github.blog/changelog/2026-09-03-upcoming-deprecation-of-selected-github-copilot-models
- Record: https://forck.live/items/8080-upcoming-deprecation-of-selected-github-copilot-models
- Subject: Copilot

GitHub announced that selected models across all GitHub Copilot experiences will be deprecated on October 2nd, 2026. Administrators are advised to enable alternative models through Copilot settings before that date to ensure continued availability.

### Microsoft — Enterprise AI transformation relies on the end-to-end platform: Azure was built for this moment

- Company: Microsoft (microsoft.com)
- Announced: 2026-09-03T19:00:00+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://azure.microsoft.com/en-us/blog/enterprise-ai-transformation-relies-on-the-end-to-end-platform-azure-was-built-for-this-moment/
- Record: https://forck.live/items/8486-enterprise-ai-transformation-relies-on-the-end-to-end-platform-azure-was-built
- Subject: Azure AI

Microsoft highlights Azure's integrated platform approach for enterprise AI, citing recognition as a Leader in both the 2026 Gartner Magic Quadrant for Strategic Cloud Platform Services and the Forrester Wave: Public Cloud Platforms, Q3 2026.

### GitHub — Gemini 3.8 Flash is now available in GitHub Copilot

- Company: GitHub (github.com)
- Announced: 2026-09-03T18:50:38+00:00
- Category: availability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://github.blog/changelog/2026-09-03-gemini-3-8-flash-is-now-available-in-github-copilot
- Record: https://forck.live/items/8077-gemini-3-8-flash-is-now-available-in-github-copilot
- Subject: Copilot

GitHub Copilot now includes Google's Gemini 3.8 Flash model, which performed well on complex terminal-based coding tasks in early testing. The model is available to Copilot Pro, Pro+, Max, Business, and Enterprise users at introductory provider pricing through December 31, 2026.

### Google Research — Transfer learning for genomic prediction in underrepresented populations

- Company: Google Research (research.google)
- Announced: 2026-09-03T18:20:31+00:00
- Category: research-paper
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://research.google/blog/transfer-learning-for-genomic-prediction-in-underrepresented-populations/
- Record: https://forck.live/items/8074-transfer-learning-for-genomic-prediction-in-underrepresented-populations
- Subject: Research

Google Research evaluated transfer learning for cross-population genetic risk prediction using UK Biobank and Biobank Japan data. The study found that while transfer learning from European cohorts improves prediction in small underrepresented populations, it degrades accuracy once target cohort sample sizes grow beyond 15,000, especially for traits with population-specific genetic architectures.

### Microsoft — GPT-6 Astra: Frontier intelligence for work, now generally available in Microsoft Foundry

- Company: Microsoft (microsoft.com)
- Announced: 2026-09-03T18:15:00+00:00
- Category: availability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://azure.microsoft.com/en-us/blog/gpt-6-astra-frontier-intelligence-for-work-now-generally-available-in-microsoft-foundry/
- Record: https://forck.live/items/8496-gpt-6-astra-frontier-intelligence-for-work-now-generally-available-in
- Subject: Azure AI

Microsoft announced the general availability of OpenAI's GPT-6 Astra in Microsoft Foundry. The model is designed for complex enterprise tasks such as deliberate planning, polished output, and execution across applications with computer-use capabilities. It is available with Standard and Provisioned Throughput deployment options in Global and US Data Zone geographies.

### Microsoft — GPT-6 Astra: Frontier intelligence for work, now available in Microsoft Foundry

- Company: Microsoft (microsoft.com)
- Announced: 2026-09-03T18:15:00+00:00
- Category: new-model
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://azure.microsoft.com/en-us/blog/gpt-6-astra-frontier-intelligence-for-work-now-available-in-microsoft-foundry/
- Record: https://forck.live/items/8487-gpt-6-astra-frontier-intelligence-for-work-now-available-in-microsoft-foundry
- Subject: Azure AI

Microsoft announced the availability of OpenAI's GPT-6 Astra frontier model through the Microsoft Foundry Limited Access Program, designed for complex enterprise work including planning, decision support, and multi-step task execution across applications. The model features computer-use capabilities and is available with Standard Global and U.S. Data Zone deployments on a consumption-based pricing model.

### Amazon — AI-driven development lifecycle using Amazon Bedrock AgentCore

- Company: Amazon (amazon.com)
- Announced: 2026-09-03T16:16:28+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://aws.amazon.com/blogs/machine-learning/ai-driven-development-lifecycle-using-amazon-bedrock-agentcore/
- Record: https://forck.live/items/8057-ai-driven-development-lifecycle-using-amazon-bedrock-agentcore
- Subject: Bedrock / Nova

Amazon Bedrock AgentCore is used in two reference implementations that demonstrate AI-DLC construction phase patterns: one generates Mermaid ER diagrams from SQL schemas, and the other provides automated code security analysis through a multi-agent architecture.

### Amazon — Migrate agentic workloads to Amazon Bedrock AgentCore

- Company: Amazon (amazon.com)
- Announced: 2026-09-03T16:14:12+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://aws.amazon.com/blogs/machine-learning/migrate-agentic-workloads-to-amazon-bedrock-agentcore/
- Record: https://forck.live/items/8058-migrate-agentic-workloads-to-amazon-bedrock-agentcore
- Subject: Bedrock / Nova

A LangGraph customer support agent can be migrated to Amazon Bedrock AgentCore in two stages, offloading operational burdens such as compute management, session isolation, and tool authentication. AgentCore provides Runtime, Gateway, and Memory services to host agents with managed tools and durable state.

### Amazon — Integrating Outlook with Amazon Quick for AI-powered email automation

- Company: Amazon (amazon.com)
- Announced: 2026-09-03T16:11:57+00:00
- Category: capability-change
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://aws.amazon.com/blogs/machine-learning/integrating-outlook-with-amazon-quick-for-ai-powered-email-automation/
- Record: https://forck.live/items/8059-integrating-outlook-with-amazon-quick-for-ai-powered-email-automation
- Subject: Bedrock / Nova

Amazon announced the integration of Microsoft Outlook with Amazon Quick, a generative AI assistant, enabling users to summarize email threads, draft replies, schedule meetings, and automate workflows through chat agents, Flows, and Automate capabilities.

### Amazon — Set up OpenAI ChatGPT Codex with LiteLLM on Amazon ECS and Amazon Bedrock

- Company: Amazon (amazon.com)
- Announced: 2026-09-03T16:10:39+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://aws.amazon.com/blogs/machine-learning/set-up-openai-chatgpt-codex-with-litellm-on-amazon-ecs-and-amazon-bedrock/
- Record: https://forck.live/items/8060-set-up-openai-chatgpt-codex-with-litellm-on-amazon-ecs-and-amazon-bedrock
- Subject: Bedrock / Nova

Amazon published a walkthrough for deploying a customer-operated LiteLLM gateway on Amazon ECS, connecting it to an OpenAI model on Amazon Bedrock, and configuring Codex to use the gateway's Responses API. The post explains the architecture, request flow, and trade-offs of using LiteLLM for centralized model access control.

### Amazon — Best practices for building agentic automations with Amazon Quick Automate

- Company: Amazon (amazon.com)
- Announced: 2026-09-03T16:08:28+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://aws.amazon.com/blogs/machine-learning/best-practices-for-building-agentic-automations-with-amazon-quick-automate/
- Record: https://forck.live/items/8061-best-practices-for-building-agentic-automations-with-amazon-quick-automate
- Subject: Bedrock / Nova

Amazon shared best practices for building agentic automations with Amazon Quick Automate, emphasizing that process quality determines agent quality and that agents should be designed with clear, bounded responsibilities.

### Amazon — Embed Quick Sight visuals using Cognito user authentication

- Company: Amazon (amazon.com)
- Announced: 2026-09-03T16:01:56+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://aws.amazon.com/blogs/machine-learning/embed-quick-sight-visuals-using-cognito-user-authentication/
- Record: https://forck.live/items/8062-embed-quick-sight-visuals-using-cognito-user-authentication
- Subject: Bedrock / Nova

Amazon published a guide on embedding individual Amazon QuickSight visuals into React applications with registered user authentication via Amazon Cognito. The solution uses a four-layer serverless architecture including API Gateway, AWS Lambda, Cognito User Pools, and QuickSight Embedding SDK. It details user synchronization and role-based access control to ensure each user sees only permitted data.

### NVIDIA — Sparks Fly: NVIDIA Accelerates Local AI at IFA 2026

- Company: NVIDIA (nvidia.com)
- Announced: 2026-09-03T16:00:59+00:00
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://blogs.nvidia.com/blog/local-ai-ifa-next-gen-agents-nv-pair-rtx-spark/
- Record: https://forck.live/items/8056-sparks-fly-nvidia-accelerates-local-ai-at-ifa-2026
- Subject: AI platform

NVIDIA and Microsoft announced partnerships to accelerate local AI inference on NVIDIA hardware at IFA 2026, including new optimizations for llama.cpp and vLLM, and simplified local model setup in Hermes Agent, OpenClaw, and Perplexity Portable Computer. NVIDIA also introduced the Personal AI Router tool (NVIDIA PAIR) and the RTX Spark compact Windows PC line coming in October from Lenovo and Acer.

### Meta — ZGateway: Learnings from Putting a Proxy in Front of ZippyDB

- Company: Meta (meta.com)
- Announced: 2026-09-03T16:00:20+00:00
- Category: infrastructure-release
- Coverage: 1 outlet
- Announcement: yes
- Group: covered
- Source: https://engineering.fb.com/2026/09/03/core-infra/zgateway-proxy-zippydb-meta/
- Record: https://forck.live/items/8055-zgateway-learnings-from-putting-a-proxy-in-front-of-zippydb
- Subject: Llama / infrastructure

Meta introduced ZGateway, a stateless proxy tier for ZippyDB that unifies client traffic, enables admission control, load balancing, cross-region resilience, and richer operations. It handles over 1 billion operations per second and carries about 40% of all ZippyDB traffic, projected to grow past 60%, while adding only about 6% computational overhead.

### Google Research — A connectomics milestone: Mapping the complete male fruit fly brain

- Company: Google Research (research.google)
- Announced: 2026-09-03T16:00:03+00:00
- Category: research-paper
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://research.google/blog/a-connectomics-milestone-mapping-the-complete-male-fruit-fly-brain/
- Record: https://forck.live/items/8054-a-connectomics-milestone-mapping-the-complete-male-fruit-fly-brain
- Subject: Research

Google Research, in partnership with HHMI Janelia and collaborators, published a complete wiring diagram of the male fruit fly's brain and central nervous system in Cell. The map contains over 166,000 neurons and 125 million synaptic connections, making it the largest brain map by number of neurons to date.

### OpenClaw — OpenClaw improves user onboarding with a new installer on macOS, plus easier local model setup for Windows NVIDIA RTX PCs

- Company: OpenClaw (openclaw.ai)
- Announced: 2026-09-03T16:00:00+00:00
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://openclaw.ai/blog/macos-installer-windows-local-ai/
- Record: https://forck.live/items/17758-openclaw-improves-user-onboarding-with-a-new-installer-on-macos-plus-easier
- Subject: OpenClaw

OpenClaw released a native app installer for macOS, bringing the same click-through installation UI previously available on Windows. On Windows PCs with NVIDIA RTX GPUs (at least 24GB memory), the setup now automatically detects hardware and suggests optimized local models. The macOS version also includes automatic detection of existing AI configurations (Claude, Codex, Ollama). Both platforms gain a permissioning interface that explains what the agent accesses and why.

### Harvey — Harvey Partners With Everlaw to Power Evidence-Backed Legal Work

- Company: Harvey (harvey.ai)
- Announced: 2026-09-03T15:45:00+00:00
- Category: partnership-acquisition
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://www.harvey.ai/blog/harvey-everlaw-evidence-ediscovery
- Record: https://forck.live/items/17526-harvey-partners-with-everlaw-to-power-evidence-backed-legal-work
- Subject: Harvey

Harvey announced a partnership with Everlaw to build an integration using the Model Context Protocol, allowing joint customers to query evidence stored in Everlaw directly from Harvey. The integration is expected to be available in fall 2026 and will support use cases such as early case assessment, chronology building, deposition preparation, and regulatory response.

### Google DeepMind — Introducing WeatherNext 3, our most advanced and accurate global weather AI model

- Company: Google DeepMind (deepmind.google)
- Announced: 2026-09-03T15:02:08+00:00
- Category: new-model
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://deepmind.google/blog/introducing-weathernext-3-our-most-advanced-and-accurate-global-weather-ai-model/
- Record: https://forck.live/items/8047-introducing-weathernext-3-our-most-advanced-and-accurate-global-weather-ai-model
- Subject: Gemini

Google DeepMind and Google Research introduced WeatherNext 3, which is the most advanced and accurate global weather model to date according to independent evaluations by Brightband. The model learns from real-time observations and uses raw satellite data to produce hourly forecasts in high resolution, making forecasts accessible across Google products worldwide.

### Microsoft — How Microsoft’s Physical Security Engineering Team scaled hybrid operations with Azure Arc and Azure Virtual Desktop

- Company: Microsoft (microsoft.com)
- Announced: 2026-09-03T15:00:00+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://azure.microsoft.com/en-us/blog/how-microsofts-physical-security-engineering-team-scaled-hybrid-operations-with-azure-arc-and-azure-virtual-desktop/
- Record: https://forck.live/items/8488-how-microsoft-s-physical-security-engineering-team-scaled-hybrid-operations
- Subject: Azure AI

Microsoft describes how its Physical Security Engineering Team used Azure Arc and Azure Virtual Desktop to unify management of hybrid infrastructure and provide consistent operator experiences across global datacenter operations.

### GitHub — Reopening Copilot Business and Enterprise signups

- Company: GitHub (github.com)
- Announced: 2026-09-03T14:40:43+00:00
- Category: availability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://github.blog/changelog/2026-09-03-reopening-copilot-business-and-enterprise-signups
- Record: https://forck.live/items/8067-reopening-copilot-business-and-enterprise-signups
- Subject: Copilot

GitHub is gradually reopening sign-ups for Copilot Business and Copilot Enterprise for customers paying by credit card or PayPal, with new billing changes including upfront charges for assigned seats and potential additional payment for exceeded usage. These billing updates will take effect starting October 1, 2026.

### OpenAI — Daybreak for Frontline Defenders: $1B to protect essential services

- Company: OpenAI (openai.com)
- Announced: 2026-09-03T13:15:00+00:00
- Category: not stated
- Coverage: 5 outlets
- Announcement: no
- Group: covered
- Source: https://openai.com/index/daybreak-for-frontline-defenders
- Record: https://forck.live/items/8082-daybreak-for-frontline-defenders-1b-to-protect-essential-services
- Subject: GPT / ChatGPT / API

OpenAI introduces Daybreak for Frontline Defenders, a $1 billion commitment to expand access to frontier cyber AI, training, and support for essential services.

### H Company — NeoMME: an efficient Multimodal-native and Multilingual Encoder

- Company: H Company (hcompany.ai)
- Announced: 2026-09-03T13:13:48+00:00
- Category: open-weight-release
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://huggingface.co/blog/Hcompany/neomme
- Record: https://forck.live/items/7637-neomme-an-efficient-multimodal-native-and-multilingual-encoder

Hugging Face introduces NeoMME, a family of 260M and 800M multilingual multimodal encoders trained from scratch with a masked discrete-diffusion objective. The model is fine-tuned for visual document retrieval and achieves high throughput. All model checkpoints are released under the Apache 2.0 license.

### OpenAI — Legora reviewed 41 documents in minutes with GPT-6 Astra

- Company: OpenAI (openai.com)
- Announced: 2026-09-03T12:00:00+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://openai.com/index/legora-financial-statement-review-with-astra
- Record: https://forck.live/items/8069-legora-reviewed-41-documents-in-minutes-with-gpt-6-astra
- Subject: GPT / ChatGPT / API

Legora used GPT-6 Astra to review 41 documents in minutes, find all four planted errors, and improve performance by nearly 40% in a financial-review workflow.

### OpenAI — Playco cut manual fixes 50% prototyping games with GPT-6 Astra

- Company: OpenAI (openai.com)
- Announced: 2026-09-03T12:00:00+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://openai.com/index/playco-game-prototyping-with-astra
- Record: https://forck.live/items/8068-playco-cut-manual-fixes-50-prototyping-games-with-gpt-6-astra
- Subject: GPT / ChatGPT / API

Playco used GPT-6 Astra to build three themed game prototypes from one grey box foundation and reported 50% fewer manual fixes compared to the previous model.

### NVIDIA — NVIDIA to Acquire Hugging Face

- Company: NVIDIA (nvidia.com)
- Announced: 2026-09-03T11:59:49+00:00
- Category: partnership-acquisition
- Coverage: 84 outlets
- Announcement: yes
- Group: covered
- Source: https://blogs.nvidia.com/blog/nvidia-to-acquire-hugging-face/
- Record: https://forck.live/items/7613-nvidia-to-acquire-hugging-face
- Subject: AI platform

NVIDIA has agreed to acquire Hugging Face for $12,930,300,000. Hugging Face will remain an open platform supporting open source and open weight models, multi-cloud and multi-accelerator development. NVIDIA plans to help scale Hugging Face's platform, improve infrastructure and expand AI access for developers and institutions.

### OpenAI — GPT-6 Astra: A new generation of intelligence

- Company: OpenAI (openai.com)
- Announced: 2026-09-03T11:00:00+00:00
- Category: new-model
- Coverage: 8 outlets
- Announcement: yes
- Group: models
- Source: https://openai.com/index/gpt-6-astra/
- Record: https://forck.live/items/8138-gpt-6-astra-a-new-generation-of-intelligence
- Subject: GPT / ChatGPT / API

OpenAI introduced GPT-6 Astra, described as its most intelligent and aligned model, with state-of-the-art capabilities in computer use, coding, cybersecurity, and science.

### Runway — Introducing GWM Worlds 2

- Company: Runway (runway.com)
- Announced: 2026-09-03T10:00:00+00:00
- Category: new-model
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://runway.com/research/introducing-gwm-worlds-2
- Record: https://forck.live/items/8051-introducing-gwm-worlds-2
- Subject: Gen / Aleph

Runway introduced GWM Worlds 2, a research preview of a model that generates interactive worlds in real time with 720p video at 24 fps and audio at 48 kHz, responding to user inputs. The model supports user-defined environments, subjects, visual style, physical rules, and ambience, and allows steering via text actions and camera motion.

### Alibaba — E-Commerce Bench: Long-Horizon Operations, Multi-Dimensional Evaluation

- Company: Alibaba (alibaba.com)
- Announced: 2026-09-03T02:00:00+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://qwen.ai/blog?id=e-commerce-bench
- Record: https://forck.live/items/7581-e-commerce-bench-long-horizon-operations-multi-dimensional-evaluation
- Subject: Qwen

Alibaba, in partnership with Taobao & Tmall Group, released E-Commerce Bench, a benchmark that evaluates how well a model runs online stores as a merchant in a realistic market over a simulated year. The benchmark includes real market data, a time budget, three-account settlement, storage and shipping, returns and reputation, and a deterministic negotiation kernel.

### Google DeepMind — Lyria 3.5 generally available (GA)

- Company: Google DeepMind (deepmind.google)
- Announced: 2026-09-03
- Category: model-update
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://ai.google.dev/gemini-api/docs/changelog#09-03-2026
- Record: https://forck.live/items/24003-lyria-3-5-generally-available-ga
- Subject: Gemini API

Google DeepMind has released Lyria 3.5, the next generation of its music generation model, now generally available for developers. The model generates full-length songs with improved musical coherence and natural vocals, and offers fine-grained control over duration and structure. It accepts text and image inputs and outputs high-fidelity 44.1 kHz stereo audio. Developers can find details and code samples in the Music generation guide.

### Microsoft AI — MAI-Transcribe-2 is the fastest, most accurate and cheapest speech recognition model in the world

- Company: Microsoft AI (microsoft.ai)
- Announced: 2026-09-03
- Category: new-model
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://microsoft.ai/news/mai-transcribe-2-is-the-fastest-most-accurate-and-cheapest-speech-recognition-model-in-the-world
- Record: https://forck.live/items/8413-mai-transcribe-2-is-the-fastest-most-accurate-and-cheapest-speech-recognition
- Subject: MAI

Microsoft AI announced MAI-Transcribe-2, a speech recognition model that it claims is the fastest, most accurate, and cheapest. The model features diarization, configurable transcription styles, and word-level timestamps. It ranks first on the FLEURS benchmark with an average Word-Error-Rate of 5.2% across 60 languages, and is priced at $0.10 per hour as a limited-time offer.

### xAI — Grok Bot for Enterprise

- Company: xAI (x.ai)
- Announced: 2026-09-03
- Category: availability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://x.ai/news/grok-bot-for-enterprise
- Record: https://forck.live/items/8084-grok-bot-for-enterprise
- Subject: Grok

xAI announced that Grok Bot is now available for enterprises, with new access, network, and audit controls for governance. Grok and Cursor Enterprise customers get free usage for the next two weeks and can invite their entire organization.

### OpenAI — Safety overview: GPT-6 Astra

- Company: OpenAI (openai.com)
- Announced: 2026-09-03
- Category: safety-policy-update
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://openai.com/index/safety-overview-gpt-6-astra
- Record: https://forck.live/items/8083-safety-overview-gpt-6-astra
- Subject: GPT / ChatGPT / API

GPT-6 Astra is OpenAI's most capable broadly deployed model and the first to reach the Critical level of cybersecurity capability under its Preparedness Framework.

### xAI — Designing Grok Bot for a world of persistent agents

- Company: xAI (x.ai)
- Announced: 2026-09-03
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://x.ai/news/designing-grok-bot
- Record: https://forck.live/items/8070-designing-grok-bot-for-a-world-of-persistent-agents
- Subject: Grok

xAI designed Grok Bot with persistent agents that have their own identity, memory, runtime, and tools. The interface is organized around Bots rather than conversations, and avatars show the Bot's state.

### Cohere — Automation’s Early Footprint

- Company: Cohere (cohere.com)
- Announced: 2026-09-03
- Category: research-paper
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://cohere.com/blog/automations-early-footprint
- Record: https://forck.live/items/7636-automation-s-early-footprint
- Subject: Command

Cohere Labs released the Agentic Task Ecosystem (ATE) dataset, aggregating roughly 696,000 tools from 123,000 public MCP servers. The research finds that only 2.6% of tools can fully automate an occupational task, and nearly half of U.S. occupations have no agentic tools represented. The study provides a supply-side signal of what developers judge to be automatable, complementing existing exposure and usage data.

### Hugging Face — Fine-tuning a 350M Model for Better Structured Outputs in 100 GRPO Steps

- Company: Hugging Face (huggingface.co)
- Announced: 2026-09-03
- Category: developer-tool-release
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://huggingface.co/blog/grpo-with-trl-ifstruct
- Record: https://forck.live/items/7607-fine-tuning-a-350m-model-for-better-structured-outputs-in-100-grpo-steps
- Subject: Platform

Hugging Face published an open-source recipe and notebook to fine-tune Liquid AI's LFM2.5-350M model using GRPO via the TRL library, improving structured-output compliance from 22.6% to 29.7% on the IFStruct benchmark. The guide runs on a free-tier Colab or Kaggle GPU using 500 samples and 100 training steps. Materials are available on GitHub.

### Hugging Face — Give Your Coding Agents a Memory You Own

- Company: Hugging Face (huggingface.co)
- Announced: 2026-09-03
- Category: developer-tool-release
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://huggingface.co/blog/funes
- Record: https://forck.live/items/7583-give-your-coding-agents-a-memory-you-own
- Subject: Platform

Hugging Face released funes, a durable memory layer for coding agents. It indexes agent traces locally and enables recall across sessions. The tool works with Claude Code, Codex, pi, and Hermes, and can be bound to a private Hugging Face dataset for shared memory.

### Hugging Face — Training a coding model to paint watercolours with TRL and OpenEnv

- Company: Hugging Face (huggingface.co)
- Announced: 2026-09-03
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://huggingface.co/blog/train-to-paint-with-code
- Record: https://forck.live/items/7556-training-a-coding-model-to-paint-watercolours-with-trl-and-openenv
- Subject: Platform

A project reproduced an idea to train a coding model to paint watercolours using TRL and OpenEnv. The reference pool dataset, RL environment, training scripts, and trained models were released openly. Three reward mixes were trained and compared, using HPSv3 and a pairwise judge.

### Alibaba — Qwen-Drive-1.0: An Initial Step towards a Vision-Language Foundation Model for Autonomous Driving

- Company: Alibaba (alibaba.com)
- Announced: 2026-09-03
- Category: new-model
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://qwen.ai/blog?id=qwen-drive-1.0
- Record: https://forck.live/items/7456-qwen-drive-1-0-an-initial-step-towards-a-vision-language-foundation-model-for
- Subject: Qwen

Alibaba introduces Qwen-Drive-1.0, a vision-language foundation model for autonomous driving built on Qwen3.5-4B. It unifies 3D perception, visual question answering, and motion planning by attaching a BEV perception head and a Planning Expert module. The model achieves competitive results on driving scene understanding and motion planning benchmarks.

### Perplexity — GLM 5.3 Flash

- Company: Perplexity (perplexity.ai)
- Announced: 2026-09-02T22:22:29+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/8066-glm-5-3-flash
- Subject: Sonar

Perplexity added the model perplexity/glm-5.3-flash to its Agent API and Router API, priced at $0.15 per million uncached-input tokens, $0.03 per million cached-input tokens, and $0.50 per million output tokens.

### Amazon — Accessing OpenAI models on Amazon Bedrock from Australia with global cross-Region inference

- Company: Amazon (amazon.com)
- Announced: 2026-09-02T21:22:05+00:00
- Category: availability-change
- Coverage: 1 outlet
- Announcement: yes
- Group: covered
- Source: https://aws.amazon.com/blogs/machine-learning/accessing-openai-gpt-5-6-models-on-amazon-bedrock-from-australia-with-global-cross-region-inference/
- Record: https://forck.live/items/7444-accessing-openai-models-on-amazon-bedrock-from-australia-with-global-cross
- Subject: Bedrock / Nova

Amazon announced that Australian teams can now access OpenAI GPT-5.6 Sol, Terra, and Luna through Amazon Bedrock with global cross-Region inference from Asia Pacific (Sydney) and Asia Pacific (Melbourne) AWS Regions. The models support text and image inputs, generate text, and have context windows of up to 1 million tokens.

### Anthropic — How the Claude Code team uses Claude Code

- Company: Anthropic (anthropic.com)
- Announced: 2026-09-02T19:07:15+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://www.youtube.com/watch?v=S-sYlFiGFv8
- Record: https://forck.live/items/19926-how-the-claude-code-team-uses-claude-code

The Claude Code team at Anthropic published a video in which team members Thariq Shihipar, Sid Bidasaria, and Robert Boyce discuss how they use Claude Code in their daily work. They explain why they work through Claude Tag, why they give Claude goals rather than tasks, and why they delete harness features as models outgrow them. The video traces the product's expansion from a terminal tool into a multi-surface tool incorporating auto mode, workflows, and routines.

### GitHub — Enterprise-managed settings support any default model

- Company: GitHub (github.com)
- Announced: 2026-09-02T18:51:12+00:00
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://github.blog/changelog/2026-09-02-enterprise-managed-settings-support-any-default-model
- Record: https://forck.live/items/7439-enterprise-managed-settings-support-any-default-model
- Subject: Copilot

GitHub announced that enterprise-managed settings now support setting any default model for GitHub Copilot conversations. The default can be customized by enterprise team, and it is generally available with Copilot Business and Copilot Enterprise in the Copilot app, Copilot CLI, and Visual Studio Code.

### Amazon — Modernizing and scaling support operations with generative AI on AWS

- Company: Amazon (amazon.com)
- Announced: 2026-09-02T18:26:35+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://aws.amazon.com/blogs/machine-learning/modernizing-and-scaling-support-operations-with-generative-ai-on-aws/
- Record: https://forck.live/items/7433-modernizing-and-scaling-support-operations-with-generative-ai-on-aws
- Subject: Bedrock / Nova

Amazon Web Services (AWS) described how organizations can use generative AI on AWS to modernize support operations, including automating SOP creation from training videos, applying RAG for ticket resolution, and using machine learning to optimize workload distribution and predict SLA risks. The approach aims to address scaling challenges in enterprise support operations.

### Amazon — How an AWS team detects dashboard content failures at scale using Amazon Bedrock

- Company: Amazon (amazon.com)
- Announced: 2026-09-02T18:21:20+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://aws.amazon.com/blogs/machine-learning/how-an-aws-team-detects-dashboard-content-failures-at-scale-using-amazon-bedrock/
- Record: https://forck.live/items/7434-how-an-aws-team-detects-dashboard-content-failures-at-scale-using-amazon-bedrock
- Subject: Bedrock / Nova

AWS describes a solution built using Amazon Bedrock to automatically detect dashboard content failures, reducing mean time to detection from up to 72 hours to less than 1 hour.

### Amazon — From code to diagrams: Agentic architecture documentation with Amazon Bedrock AgentCore

- Company: Amazon (amazon.com)
- Announced: 2026-09-02T18:18:47+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://aws.amazon.com/blogs/machine-learning/from-code-to-diagrams-agentic-architecture-documentation-with-amazon-bedrock-agentcore/
- Record: https://forck.live/items/7435-from-code-to-diagrams-agentic-architecture-documentation-with-amazon-bedrock
- Subject: Bedrock / Nova

Amazon Bedrock AgentCore is used to build an autonomous agent that analyzes code bases and generates architecture diagrams automatically. The solution integrates with AWS CI/CD pipelines and uses Amazon Bedrock Knowledge Bases for semantic search. A global interdealer broker has been running this solution in production since Q1 2026.

### GitHub — Content exclusions generally available in Copilot app and CLI

- Company: GitHub (github.com)
- Announced: 2026-09-02T18:14:23+00:00
- Category: availability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://github.blog/changelog/2026-09-02-content-exclusions-generally-available-in-copilot-app-and-cli
- Record: https://forck.live/items/7445-content-exclusions-generally-available-in-copilot-app-and-cli
- Subject: Copilot

GitHub made content exclusions generally available in the Copilot app and Copilot CLI, allowing enterprise, organization, and repository administrators to configure policies that prevent Copilot from using excluded files as context. The feature is available for Copilot Business and Copilot Enterprise customers.

### Amazon — Trinity: Agentic AI-powered transition planning for students with disabilities

- Company: Amazon (amazon.com)
- Announced: 2026-09-02T18:14:17+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://aws.amazon.com/blogs/machine-learning/trinity-agentic-ai-powered-transition-planning-for-students-with-disabilities/
- Record: https://forck.live/items/7436-trinity-agentic-ai-powered-transition-planning-for-students-with-disabilities
- Subject: Bedrock / Nova

University Startups, in partnership with g/d/n/a and AWS, built Trinity, a conversational AI solution on Amazon Bedrock that helps students with disabilities create personalized postsecondary transition plans. The system uses a multi-agent architecture with Claude 3.5 Sonnet and has reached educators and students across more than a dozen U.S. states, with expansion into Saudi Arabia and Kuwait underway.

### Google DeepMind — Proactive cyber defense for governments and enterprises

- Company: Google DeepMind (deepmind.google)
- Announced: 2026-09-02T16:24:24+00:00
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://deepmind.google/blog/proactive-cyber-defense-for-governments-and-enterprises/
- Record: https://forck.live/items/7385-proactive-cyber-defense-for-governments-and-enterprises
- Subject: Gemini

Google DeepMind launched the Fairwind Program, a limited access initiative providing governments and trusted partners with advanced AI cyber defense capabilities. The program uses the Gemini 3.8 Flash Cyber model and CodeMender harness to autonomously find and fix vulnerabilities. Participating organizations must adhere to strict operational standards, including limiting access to internal cybersecurity teams and requiring multi-factor authentication.

### Google DeepMind — Introducing Gemini 3.8 Flash and 3.8 Flash Cyber

- Company: Google DeepMind (deepmind.google)
- Announced: 2026-09-02T16:18:31+00:00
- Category: new-model
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://deepmind.google/blog/introducing-gemini-3-8-flash-and-38-flash-cyber/
- Record: https://forck.live/items/7386-introducing-gemini-3-8-flash-and-3-8-flash-cyber
- Subject: Gemini

Google DeepMind announced Gemini 3.8 Flash, a workhorse model for reasoning and coding at low cost, and Gemini 3.8 Flash Cyber, a cybersecurity model for vulnerability discovery and automated patching. Both models offer significant improvements over their predecessors.

### Google Cloud — Getting started with Mantis, our open-source bug finding-and-fixing harness

- Company: Google Cloud (cloud.google.com)
- Announced: 2026-09-02T16:00:00+00:00
- Category: developer-tool-release
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://cloud.google.com/blog/products/identity-security/getting-started-with-the-mantis-harness-to-find-and-fix-bugs/
- Record: https://forck.live/items/17334-getting-started-with-mantis-our-open-source-bug-finding-and-fixing-harness
- Subject: Gemini Enterprise

AI models have clearly proven their ability to discover and exploit vulnerabilities without much, if any, human assistance. To help defenders gain the advantage with AI, we built the Mantis harness to automate the discovery, triage, reproduction, and patching of software vulnerabilities. Available to all as an open-source framework, Mantis is part of Google’s internal approach to find and fix vulnerabilities at machine-speed. It creates a more effective scalable, context-aware repository analysis. While sloppiness in AI code scanning frequently leads to hallucinated bugs and weak true-positive rates under 7%, we designed Mantis to be effective by combining industry-standard agentic techniques like critic and review agents with sandboxed reproduction of vulnerabilities for grounding. As we detailed in June , it examines the history of the repository to learn from past security fixes and automatically builds up architectural and threat model documentation, even if these are not provided. It constructs a hierarchical security summary tree, condensing individual files into directory and root-level summaries. …

### Microsoft — The Economics of Agent Optimization: Context engineering for enterprise AI agents

- Company: Microsoft (microsoft.com)
- Announced: 2026-09-02T16:00:00+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://azure.microsoft.com/en-us/blog/the-economics-of-agent-optimization-context-engineering-for-enterprise-ai-agents/
- Record: https://forck.live/items/7580-the-economics-of-agent-optimization-context-engineering-for-enterprise-ai-agents
- Subject: Azure AI

Microsoft describes context engineering for enterprise AI agents on Microsoft Foundry, explaining how managing what enters the context window can lower costs and improve performance. The post highlights Foundry IQ for knowledge retrieval and Toolboxes for tool management, reporting that Foundry IQ improved evidence recall by up to 54% while reducing retrieval token costs by 34%, and Toolboxes reduced average input-token consumption by around 97% for large tool libraries.

### Sakana AI — Percept-Lens: A Deep Dive into AI-Generated Image Detection

- Company: Sakana AI (sakana.ai)
- Announced: 2026-09-02T15:00:00+00:00
- Category: research-paper
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://sakana.ai/percept-lens
- Record: https://forck.live/items/14459-percept-lens-a-deep-dive-into-ai-generated-image-detection
- Subject: Sakana models

Sakana AI's study, to be presented at ECCV2026, introduces Percept-Lens, a benchmark for evaluating AI-generated image detectors under distribution shifts, and a new detector that keeps a general-purpose vision model frozen while fitting a simple Gaussian decision rule to its representations. On the same broad evaluation suite, this detector outperformed the strongest released AI-generated image detector tested, despite not being trained specifically for the task.

### Runway — Introducing Runway Dev MCP

- Company: Runway (runway.com)
- Announced: 2026-09-02T14:25:16.960000+00:00
- Category: developer-tool-release
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://runway.com/news/company-news/runway-dev-mcp
- Record: https://forck.live/items/7857-introducing-runway-dev-mcp
- Subject: Gen / Aleph

Runway launched Runway Dev MCP, a hosted MCP server that connects Runway Dev accounts to coding tools, enabling coding agents to plan, configure, and debug integrations. The tool allows agents to research models, create Model Routers, manage Characters, and debug failures, designed for agent-driven development.

### Harvey — How We Rebuilt Playbook Review as a Multi-Agent System

- Company: Harvey (harvey.ai)
- Announced: 2026-09-02T14:00:00+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://www.harvey.ai/blog/rebuilding-playbook-review-as-a-multi-agent-system
- Record: https://forck.live/items/17527-how-we-rebuilt-playbook-review-as-a-multi-agent-system
- Subject: Harvey

Harvey describes rebuilding its playbook review engine as a multi-agent system to improve contract review. The post explains the limitations of the previous pipeline-of-prompts approach and outlines how the new architecture handles complex conditional logic, rule interactions, and deal context. It also discusses quality evaluation in a domain where correctness depends on legal judgment.

### Hugging Face — Real-Time Intelligence with IBM Time Series Models on Confluent

- Company: Hugging Face (huggingface.co)
- Announced: 2026-09-02T13:49:14+00:00
- Category: partnership-acquisition
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://huggingface.co/blog/ibm-research/real-time-intelligence
- Record: https://forck.live/items/7352-real-time-intelligence-with-ibm-time-series-models-on-confluent
- Subject: Platform

IBM and Confluent announced a partnership to integrate IBM Granite Time Series models into Confluent Cloud, enabling real-time forecasting and anomaly detection on streaming data directly within Flink. The integration is available in Early Access on Confluent Cloud on AWS, with Confluent Platform support to follow.

### OpenAI — ATV Big Air Tour turned 3 days of work into 3 hours with ChatGPT

- Company: OpenAI (openai.com)
- Announced: 2026-09-02T12:00:00+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://openai.com/index/atv-big-air-tour
- Record: https://forck.live/items/7448-atv-big-air-tour-turned-3-days-of-work-into-3-hours-with-chatgpt
- Subject: GPT / ChatGPT / API

OpenAI published a customer story about ATV Big Air Tour using ChatGPT to speed up marketing, merchandising, and other tasks, including turning merchandise photos into an inventory website in 15 minutes.

### ElevenLabs — Ashley Kramer joins ElevenLabs as Chief Revenue Officer

- Company: ElevenLabs (elevenlabs.io)
- Announced: 2026-09-02T12:00:00+00:00
- Category: funding-company-news
- Coverage: 1 outlet
- Announcement: no
- Group: covered
- Source: https://elevenlabs.io/blog/cro
- Record: https://forck.live/items/7351-ashley-kramer-joins-elevenlabs-as-chief-revenue-officer
- Subject: Voice AI

Ashley Kramer has joined ElevenLabs as Chief Revenue Officer. She previously led enterprise sales at OpenAI and has extensive experience at high-growth tech companies. ElevenLabs is experiencing growth, with enterprise customers accounting for 55% of revenue.

### Meta — An Organizational Second Brain: Building an AI That Learns From Experts

- Company: Meta (meta.com)
- Announced: 2026-09-02T09:00:29+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://engineering.fb.com/2026/09/02/ml-applications/organizational-second-brain-ai-learns-from-experts/
- Record: https://forck.live/items/7442-an-organizational-second-brain-building-an-ai-that-learns-from-experts
- Subject: Llama / infrastructure

Meta engineers built an AI agent that functions as a secondary expert for compliance domains, integrating a structured knowledge architecture with a self-improvement loop that compiles expert feedback into verified updates without model retraining.

### Google DeepMind — Gemini 3.8 Flash generally available (GA)

- Company: Google DeepMind (deepmind.google)
- Announced: 2026-09-02
- Category: availability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://ai.google.dev/gemini-api/docs/changelog#09-02-2026
- Record: https://forck.live/items/24004-gemini-3-8-flash-generally-available-ga
- Subject: Gemini API

Google DeepMind has made gemini-3.8-flash generally available, calling it the most intelligent Flash model yet. The company says the model is built for long-horizon software engineering, autonomous agents, and complex enterprise workflows. Developers can get started via the Gemini 3.8 Flash model page and the Gemini 3.8 Flash guide.

### Gradium — Build voice agents with EU and US residency and Zero Data Retention

- Company: Gradium (gradium.ai)
- Announced: 2026-09-02
- Category: capability-change
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://gradium.ai/blog/eu-data-residency-zero-data-retention
- Record: https://forck.live/items/18305-build-voice-agents-with-eu-and-us-residency-and-zero-data-retention
- Subject: Gradium TTS / Phonon

Gradium added data residency and zero data retention settings for its voice agent platform. Data residency lets organizations pin sessions to EU or US servers with a guarantee, while zero data retention prevents storage of request and response payloads after a call. Both features are available on paid plans.

### Apple — REFACTOR-VLA: Unsupervised Library Learning of Typed Motor Programs

- Company: Apple (apple.com)
- Announced: 2026-09-02
- Category: research-paper
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://machinelearning.apple.com/research/refactor-vla-motor-programs
- Record: https://forck.live/items/10302-refactor-vla-unsupervised-library-learning-of-typed-motor-programs
- Subject: Machine Learning Research

Apple researchers introduce REFACTOR-VLA, a system that learns reusable motor skills using a wake/sleep architecture with a Behavioral-Equivalence Kernel and typed lambda terms. Evaluated on the LIBERO benchmark, the study finds that increasing world model size from 188M to 430M parameters worsened performance on all four suites, while adding an auxiliary InfoNCE contrastive loss during world-model warmup improved skill clustering quality.

### Luma — Luma AI to Launch Native Arabic-Language Interface for Saudi and Arabic-Speaking Creative Community

- Company: Luma (lumalabs.ai)
- Announced: 2026-09-02
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://lumalabs.ai/news/luma-ai-to-launch-native-arabic-language-interface-for-saudi-and-arabic-speaking-creative-community
- Record: https://forck.live/items/7951-luma-ai-to-launch-native-arabic-language-interface-for-saudi-and-arabic
- Subject: Ray / Dream Machine

Luma AI announced that Luma Agents will offer a fully native Arabic-language interface later this year, making it the first creative intelligence platform in Arabic for Saudi Arabia and the wider MENA region. The update was announced at LEAP 2026 and stems from Luma's partnership with HUMAIN.

### Alibaba — qwen3.8-max-0902

- Company: Alibaba (alibaba.com)
- Announced: 2026-09-02
- Category: model-update
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://www.alibabacloud.com/help/en/model-studio/qwen3-8-max#qwen3-8-max-0902
- Record: https://forck.live/items/7608-qwen3-8-max-0902
- Subject: Qwen

Alibaba released Qwen3.8-Max-0902, an upgraded snapshot of qwen3.8-max, with improved coding capability for complex engineering projects, enhanced collaborative agent performance, and refined native vision understanding. It retains a 1M token context window, thinking mode, and full tool ecosystem.

### Cursor — Self-hosted machines

- Company: Cursor (cursor.com)
- Announced: 2026-09-02
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://cursor.com/changelog/self-hosted-machines
- Record: https://forck.live/items/7443-self-hosted-machines
- Subject: Cursor

Cursor announced support for self-hosted machines that keep tool execution within the user's own network, dynamic pool scheduling for scaling capacity, the ability to run cloud agents on existing infrastructure such as AWS Lambda and Coder, and computer use on Linux and Mac.

### Meta — Introducing Muse Spark 1.3

- Company: Meta (meta.com)
- Announced: 2026-09-02
- Category: model-update
- Coverage: 7 outlets
- Announcement: yes
- Group: models
- Source: https://research.meta.ai/blog/introducing-muse-spark-1-3
- Record: https://forck.live/items/7437-introducing-muse-spark-1-3
- Subject: Muse

Meta released Muse Spark 1.3, an update to its Muse Spark model, with improved performance on agentic and coding tasks. The model is available today in Muse Code and Meta Model API, with max reasoning coming after additional safety testing. Meta also announced plans for a Muse Spark 1.2 open weights release following final safety fine-tuning.

### Cohere — How small AI models can make a big impact for enterprises

- Company: Cohere (cohere.com)
- Announced: 2026-09-02
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://cohere.com/blog/how-small-models-can-make-a-big-impact-for-enterprises
- Record: https://forck.live/items/7428-how-small-ai-models-can-make-a-big-impact-for-enterprises
- Subject: Command

Cohere discusses how small language models (SLMs) can offer enterprises cost savings, reduced computational overhead, and task-specific efficiency, citing its own models Command R7B, Tiny Aya, and North Mini Code as examples that outperform larger models on certain benchmarks.

### Anthropic — Building commerce agents with Claude

- Company: Anthropic (anthropic.com)
- Announced: 2026-09-02
- Category: developer-tool-release
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://claude.com/blog/claude-for-commerce-agents
- Record: https://forck.live/items/7412-building-commerce-agents-with-claude
- Subject: Claude

Anthropic launched a blueprint for building commerce agents on Claude, containing reference implementations of a shopping agent and a merchant agent for retail, travel, telecom, and ticketing platforms. The blueprint includes harnesses, patterns, and guardrails, and is available today with live demos.

### LG AI Research — From Labeling to Deployment: VLM and Agentic AI-Based Autonomous Vision Inspection

- Company: LG AI Research (lgresearch.ai)
- Announced: 2026-09-02
- Category: research-paper
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://www.lgresearch.ai/blog/view?seq=689
- Record: https://forck.live/items/7331-from-labeling-to-deployment-vlm-and-agentic-ai-based-autonomous-vision
- Subject: EXAONE

LG AI Research presents a zero-human vision inspection framework that uses Monte Carlo Dropout for uncertainty extraction and a Vision-Language Model (VLM) for auto-labeling, achieving 93.13% accuracy with a +4.84%p improvement over random sampling. The VLM is trained through a three-stage pipeline including base model training, domain-specific fine-tuning, and reasoning reinforcement learning, and a noise-robust learning policy based on embedding space similarity is employed to handle noisy labels.

### NVIDIA — NVIDIA and CrowdStrike Strengthen Agentic Cybersecurity Frontier

- Company: NVIDIA (nvidia.com)
- Announced: 2026-09-01T21:19:20+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://blogs.nvidia.com/blog/nvidia-crowdstrike-fal-con-2026/
- Record: https://forck.live/items/7289-nvidia-and-crowdstrike-strengthen-agentic-cybersecurity-frontier
- Subject: AI platform

NVIDIA and CrowdStrike announce CrowdStrike SafeMind, an agentic cybersecurity system that combines CrowdStrike's models and NVIDIA Nemotron in a continuous coevolution loop to defend against automated attacks. Also announced is CrowdStrike Falcon IQ, which uses agentic workload automation.

### GitHub — Set an expiration date for individual user budgets

- Company: GitHub (github.com)
- Announced: 2026-09-01T20:23:13+00:00
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://github.blog/changelog/2026-09-01-set-an-expiration-date-for-individual-user-budgets
- Record: https://forck.live/items/7286-set-an-expiration-date-for-individual-user-budgets
- Subject: Copilot

GitHub now allows setting an optional expiration date on individual user budgets, after which the budget is automatically removed and the user falls back to the next applicable budget. This feature is generally available for GitHub Copilot Business and Copilot Enterprise plans.

### GitHub — Copilot code review can now approve pull requests

- Company: GitHub (github.com)
- Announced: 2026-09-01T19:25:57+00:00
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://github.blog/changelog/2026-09-01-copilot-code-review-can-now-approve-pull-requests
- Record: https://forck.live/items/7284-copilot-code-review-can-now-approve-pull-requests
- Subject: Copilot

GitHub Copilot code review now includes an approval assessment in the overview comment, and admins can enable Copilot to submit approvals that count toward merge requirements. The feature is in public preview for Copilot Pro, Pro+, Max, Business, and Enterprise plans.

### Amazon — Introducing Claude Fable 5.1 on AWS

- Company: Amazon (amazon.com)
- Announced: 2026-09-01T19:12:43+00:00
- Category: new-model
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://aws.amazon.com/blogs/machine-learning/introducing-claude-fable-5-1-on-aws/
- Record: https://forck.live/items/7272-introducing-claude-fable-5-1-on-aws
- Subject: Bedrock / Nova

Amazon announces the availability of Claude Fable 5.1 on Amazon Bedrock, a new model with improved reasoning, agentic coding, autonomous operation, and scientific research capabilities. It is designated as a Covered Model with data retention policies and is available with Enterprise Frontier Safeguards for eligible customers.

### Google Research — Mapping global methane emissions from space with deep learning

- Company: Google Research (research.google)
- Announced: 2026-09-01T18:40:06+00:00
- Category: new-model
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://research.google/blog/mapping-global-methane-emissions-from-space-with-deep-learning/
- Record: https://forck.live/items/7270-mapping-global-methane-emissions-from-space-with-deep-learning
- Subject: Research

Google Research announces MAPL-EMIT, a deep-learning framework based on a vision transformer (Swin-S) for automated detection, enhancement, and source estimation of methane plumes from hyperspectral satellite data. The framework achieves 84% recall on expert-annotated plumes and is accompanied by a global plume database, trained model, synthetic plumes, and inference library released on Earth Engine, Kaggle, and GitHub.

### Harvey — Fable 5.1, Now Live in Harvey

- Company: Harvey (harvey.ai)
- Announced: 2026-09-01T18:00:00+00:00
- Category: availability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://www.harvey.ai/blog/fable-5-1-in-harvey
- Record: https://forck.live/items/17528-fable-5-1-now-live-in-harvey
- Subject: Harvey

Harvey made Anthropic's Fable 5.1 model available as an opt-in for its customers. The model showed strongest quality improvements on litigation and dispute resolution tasks in Harvey's testing, and delivers quality results at lower reasoning levels using roughly half the output tokens. Eligible customers can use Fable 5 and Fable 5.1 with zero data retention under a time-bound exemption while Anthropic rolls out Enterprise Frontier Safeguards.

### Google DeepMind — Introducing agentic video understanding with Gemini

- Company: Google DeepMind (deepmind.google)
- Announced: 2026-09-01T17:08:51+00:00
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://deepmind.google/blog/introducing-agentic-video-in-gemini/
- Record: https://forck.live/items/7262-introducing-agentic-video-understanding-with-gemini
- Subject: Gemini

Google DeepMind announces a new agentic video understanding capability for Gemini models (3.7 Flash, 3.6 Flash, 3.5 Flash-Lite) that reduces token consumption by up to 88%, costs by up to 66%, and improves accuracy by up to 7%. The feature is available via the Gemini API in Google AI Studio and Gemini Enterprise Agent Platform, using standard token pricing.

### OpenAI — How AI-native companies turn workflows into operating capability

- Company: OpenAI (openai.com)
- Announced: 2026-09-01T17:00:00+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://openai.com/index/ai-native-company-workflows
- Record: https://forck.live/items/7264-how-ai-native-companies-turn-workflows-into-operating-capability
- Subject: GPT / ChatGPT / API

Basis, Clay, and Exa Labs use AI agents to improve onboarding, account management, and developer integrations, and suggests that enterprise leaders can apply these approaches.

### Amazon — From theory to delivery: How Atos upskilled 400 engineers in agentic AI

- Company: Amazon (amazon.com)
- Announced: 2026-09-01T16:17:54+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://aws.amazon.com/blogs/machine-learning/from-theory-to-delivery-how-atos-upskilled-400-engineers-in-agentic-ai/
- Record: https://forck.live/items/7254-from-theory-to-delivery-how-atos-upskilled-400-engineers-in-agentic-ai
- Subject: Bedrock / Nova

Amazon and Atos ran an AWS AI League event to upskill 400 engineers in agentic AI through a hands-on competition over three days, building multi-agent systems with pathfinding, guardrails, memory, and fine-tuned models.

### Amazon — Securing Amazon Quick from POC to production: Agents, Flows, and Spaces

- Company: Amazon (amazon.com)
- Announced: 2026-09-01T16:02:22+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://aws.amazon.com/blogs/machine-learning/securing-amazon-quick-from-poc-to-production-agents-flows-and-spaces/
- Record: https://forck.live/items/7256-securing-amazon-quick-from-poc-to-production-agents-flows-and-spaces
- Subject: Bedrock / Nova

Securing Amazon Quick deployments from proof-of-concept to production, covering dataset shaping, agent isolation, document classification, and approval gates with a governance framework and production readiness checklist.

### Google DeepMind — Ask a Scientist: How do researchers use AI to predict a cyclone?

- Company: Google DeepMind (deepmind.google)
- Announced: 2026-09-01T16:00:00+00:00
- Category: open-weight-release
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://blog.google/innovation-and-ai/models-and-research/google-deepmind/weathernext-extreme-weather-cyclone-predictions/
- Record: https://forck.live/items/7271-ask-a-scientist-how-do-researchers-use-ai-to-predict-a-cyclone
- Subject: Gemini

Google DeepMind has open-sourced its WeatherNext 2 and WeatherNext cyclones models, which are trained on 50 years of weather data and can run on a single TPU, improving cyclone forecasting accuracy.

### Amazon — How t54 built a trust layer with Amazon Bedrock AgentCore payments

- Company: Amazon (amazon.com)
- Announced: 2026-09-01T15:50:00+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://aws.amazon.com/blogs/machine-learning/how-t54-built-a-trust-layer-with-amazon-bedrock-agentcore-payments/
- Record: https://forck.live/items/7257-how-t54-built-a-trust-layer-with-amazon-bedrock-agentcore-payments
- Subject: Bedrock / Nova

t54 built a trust layer for agent payments using Amazon Bedrock AgentCore payments. The system processes over 20 million agent-initiated transactions with micropayments between $0.001 and $0.01 without human approval. The trust layer, x402-secure, scores endpoints in real time, and Amazon Bedrock AgentCore payments provides spending limits, credential isolation, and payment execution. The architecture enforces separation of duties via IAM roles, so agents cannot change spending rules or access credentials.

### Amazon — How ZS democratized secure ad-hoc analytics with Amazon SageMaker

- Company: Amazon (amazon.com)
- Announced: 2026-09-01T15:47:00+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://aws.amazon.com/blogs/machine-learning/how-zs-democratized-secure-ad-hoc-analytics-with-amazon-sagemaker/
- Record: https://forck.live/items/7258-how-zs-democratized-secure-ad-hoc-analytics-with-amazon-sagemaker
- Subject: Bedrock / Nova

ZS built a security-hardened Amazon SageMaker environment to balance developer agility with strict governance for ad-hoc analytics. The platform serves over 1,000 daily active users across 200+ domains, with features like network isolation, encryption, cost tracking, and governance controls.

### Amazon — How Boomi Scribe streamlines documentation using AWS

- Company: Amazon (amazon.com)
- Announced: 2026-09-01T15:45:36+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://aws.amazon.com/blogs/machine-learning/how-boomi-scribe-streamlines-documentation-using-aws/
- Record: https://forck.live/items/7259-how-boomi-scribe-streamlines-documentation-using-aws
- Subject: Bedrock / Nova

Boomi Scribe is an AI-powered agent that automates documentation for enterprise integration processes. It uses AWS services including Amazon Bedrock and Anthropic's Claude Haiku 4.5 to parse Directed Acyclic Graphs (DAGs) and generate detailed documentation, version comparisons, and insights.

### Google DeepMind — Koray Kavukcuoglu on frontier models, coding agents, and building AGI

- Company: Google DeepMind (deepmind.google)
- Announced: 2026-09-01T15:00:31+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://www.youtube.com/watch?v=Rrr2gdbvNFU
- Record: https://forck.live/items/22698-koray-kavukcuoglu-on-frontier-models-coding-agents-and-building-agi

Google DeepMind SVP Koray Kavukcuoglu discusses the evolution from reinforcement learning to Gemini, the ambitions of Gemini 4 pre-training, and the company's approach to building AGI.

### Harvey — Introducing Horizon Scanning

- Company: Harvey (harvey.ai)
- Announced: 2026-09-01T15:00:00+00:00
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://www.harvey.ai/blog/horizon-scanning-in-harvey
- Record: https://forck.live/items/17529-introducing-horizon-scanning
- Subject: Harvey

Harvey introduced Horizon Scanning, a new feature in its platform that helps legal teams monitor regulatory and legislative changes. Users can create custom scans using plain language, receive prioritized alerts, and take actions such as drafting memos or updating policies within the same system. It is available in Early Access for current Harvey customers, with general availability planned later.

### GitHub — Claude Fable 5.1 is generally available in GitHub Copilot

- Company: GitHub (github.com)
- Announced: 2026-09-01T14:29:23+00:00
- Category: availability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://github.blog/changelog/2026-09-01-claude-fable-5-1-generally-available-in-github-copilot
- Record: https://forck.live/items/7295-claude-fable-5-1-is-generally-available-in-github-copilot
- Subject: Copilot

Claude Fable 5.1 from Anthropic is now generally available in GitHub Copilot. It requires data retention by default for safety classifiers, with zero data retention available for eligible enterprise customers. The model is available to Copilot Pro+, Max, Business, and Enterprise users.

### OpenAI — Path to Astra: critical capabilities and frontier safeguards

- Company: OpenAI (openai.com)
- Announced: 2026-09-01T13:00:00+00:00
- Category: safety-policy-update
- Coverage: 20 outlets
- Announcement: yes
- Group: covered
- Source: https://openai.com/index/path-to-astra
- Record: https://forck.live/items/7283-path-to-astra-critical-capabilities-and-frontier-safeguards
- Subject: GPT / ChatGPT / API

OpenAI announces that its model Astra is the first to meet the Critical cybersecurity capability threshold under the Preparedness Framework, and includes stronger safeguards for release.

### Lovable — Lovable now builds with Fable 5.1

- Company: Lovable (lovable.dev)
- Announced: 2026-09-01T12:00:00+00:00
- Category: model-update
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://lovable.dev/blog/fable-5-1-now-in-lovable
- Record: https://forck.live/items/12056-lovable-now-builds-with-fable-5-1
- Subject: Lovable / AI app builder

Lovable now builds with Fable 5.1, which outperforms Fable 5 by up to 17% on the most difficult tasks, particularly in fixing and improving existing apps without breaking what already works. The update also offers a 31% lower cost per task and a 3.5% lift in visual design quality.

### OpenAI — Healthcare organizations can now connect EHR and additional industry data to ChatGPT

- Company: OpenAI (openai.com)
- Announced: 2026-09-01T12:00:00+00:00
- Category: capability-change
- Coverage: 5 outlets
- Announcement: yes
- Group: covered
- Source: https://openai.com/index/chatgpt-connects-health-records-and-healthcare-sources
- Record: https://forck.live/items/7265-healthcare-organizations-can-now-connect-ehr-and-additional-industry-data-to
- Subject: GPT / ChatGPT / API

OpenAI announced that ChatGPT can now connect to trusted healthcare data, allowing clinicians to access patient context, medical research, and more securely.

### Google Antigravity — Gemini 3.8 Flash in Google Antigravity

- Company: Google Antigravity (antigravity.google)
- Announced: 2026-09-01T08:00:00+00:00
- Category: model-update
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://antigravity.google/blog/gemini-3-8-flash-in-google-antigravity
- Record: https://forck.live/items/7362-gemini-3-8-flash-in-google-antigravity
- Subject: Antigravity

Google released Gemini 3.8 Flash, available immediately in Google Antigravity, priced at $0.75 per million input tokens and $3.75 per million output tokens, designed for agentic development workflows.

### Vercel — Compute that takes any shape

- Company: Vercel (vercel.com)
- Announced: 2026-09-01T07:00:00+00:00
- Category: infrastructure-release
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://vercel.com/blog/fluid-compute-takes-any-shape
- Record: https://forck.live/items/20662-compute-that-takes-any-shape
- Subject: v0 / AI SDK / AI Gateway

Vercel announced Fluid, a unified compute layer that runs builds, sandboxes, and functions on a single system. Fluid assembles a machine for each workload, uses Hive for isolated VM provisioning, and supports custom images via Vercel Container Registry. Vercel Drives provide portable storage that outlives the compute, currently in private beta for sandboxes.

### OpenAI — How law firm Gilbert + Tobin governs and scales AI with OpenAI

- Company: OpenAI (openai.com)
- Announced: 2026-09-01T01:00:00+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://openai.com/index/gilbert-tobin
- Record: https://forck.live/items/7301-how-law-firm-gilbert-tobin-governs-and-scales-ai-with-openai
- Subject: GPT / ChatGPT / API

Gilbert + Tobin, a law firm, uses CEO-led commitment, rigorous governance, and human accountability to scale ChatGPT Enterprise and Codex across the firm.

### Google DeepMind — Agentic video understanding

- Company: Google DeepMind (deepmind.google)
- Announced: 2026-09-01
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://ai.google.dev/gemini-api/docs/changelog#09-01-2026
- Record: https://forck.live/items/24005-agentic-video-understanding
- Subject: Gemini API

Google DeepMind added agentic video understanding to Gemini 3.7 Flash, 3.6 Flash, and 3.5 Flash-Lite across the Interactions and GenerateContent APIs. The model dynamically navigates video timelines, requesting transcripts, frames, or audio tracks on demand instead of processing every frame. This approach uses up to 88% fewer tokens for long-form content compared to static processing, according to Google DeepMind. The feature is available now through the Gemini API, with a guide to get started.

### Ai2 — The hard parts of AI-assisted science

- Company: Ai2 (allenai.org)
- Announced: 2026-09-01
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://allenai.org/blog/swedish-autodiscovery-recap
- Record: https://forck.live/items/18535-the-hard-parts-of-ai-assisted-science
- Subject: Olmo / Molmo / Asta

Ai2 hosted an event on August 27, 2026, discussing challenges in building AI systems that scientists can steer as research evolves. Five recurring challenges were identified: leaving room for human judgment, keeping AI steerable, deciding what to delegate, avoiding amplification of bad science, and tightening the loop between AI and experiments.

### Ai2 — BenchMIRT: What are LLM benchmarks actually measuring?

- Company: Ai2 (allenai.org)
- Announced: 2026-09-01
- Category: research-paper
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://allenai.org/blog/benchmirt
- Record: https://forck.live/items/18534-benchmirt-what-are-llm-benchmarks-actually-measuring
- Subject: Olmo / Molmo / Asta

Ai2 introduced BenchMIRT, a method that uses multidimensional item response theory to audit LLM benchmarks at the individual prompt level. Applied to 100 models across 16 benchmarks, it independently recovered safety and general reasoning as two dominant dimensions. The analysis revealed that some benchmarks, such as BBQ, align more with general reasoning than with safety, and that BenchMIRT can identify the most informative questions in an evaluation.

### Harvey — GE Aerospace Deploys Harvey Across Its Legal & Compliance Organization

- Company: Harvey (harvey.ai)
- Announced: 2026-09-01
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://www.harvey.ai/blog/ge-aerospace-deploys-harvey-across-its-legal-and-compliance-organization
- Record: https://forck.live/items/17744-ge-aerospace-deploys-harvey-across-its-legal-compliance-organization
- Subject: Harvey

GE Aerospace has deployed Harvey across its Legal & Compliance organization after a successful pilot. The company is also serving as a design partner for Harvey's Contract Intelligence product, helping shape its development for contract reviews and negotiations.

### Manus — Manus Resumes Independent Operations

- Company: Manus (manus.im)
- Announced: 2026-09-01
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://manus.im/blog/manus-resumes-independent-operations
- Record: https://forck.live/items/14698-manus-resumes-independent-operations
- Subject: Manus

Manus announced it has formally resumed independent operations, with its founding team continuing to lead the company. The company acknowledged a temporary interruption that required some users to back up and restore data, and expressed gratitude for their patience. Manus stated it will continue to develop general AI agents and integrate them more deeply into user workflows.

### Perplexity — Introducing Hybrid Compute on Mac

- Company: Perplexity (perplexity.ai)
- Announced: 2026-09-01
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://www.perplexity.ai/hub/blog/introducing-hybrid-compute-on-mac
- Record: https://forck.live/items/10769-introducing-hybrid-compute-on-mac
- Subject: Perplexity

Perplexity introduced hybrid compute on Mac, allowing users to run tasks locally on their device while using cloud AI for reasoning and web search. The feature includes a privacy gate that controls what data leaves the machine, and is available to Pro, Max, and Enterprise subscribers on Apple silicon Macs with at least 24GB of unified memory.

### Perplexity — How to Spot AI Hallucinations: 7 Red Flags

- Company: Perplexity (perplexity.ai)
- Announced: 2026-09-01
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://www.perplexity.ai/hub/blog/ai-hallucination
- Record: https://forck.live/items/10768-how-to-spot-ai-hallucinations-7-red-flags
- Subject: Perplexity

Perplexity's guide explains how to identify AI hallucinations by checking traceability, consistency, and currency, and provides a table of red flags with verification steps. It notes that hallucination risk increases with obscure or recent topics and that even leading models struggle to recognize when they lack sufficient information.

### Perplexity — AI personal assistants: What they can and can’t do

- Company: Perplexity (perplexity.ai)
- Announced: 2026-09-01
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://www.perplexity.ai/hub/blog/ai-personal-assistants
- Record: https://forck.live/items/10765-ai-personal-assistants-what-they-can-and-can-t-do
- Subject: Perplexity

Perplexity published a guide explaining how AI personal assistants work, what tasks they can handle, and how they differ from AI chatbots and agents. The guide covers use cases including research, email management, note-taking, calendar coordination, and shopping, and notes that leading agents complete about two-thirds of tasks on common computer-use benchmarks.

### Perplexity — PII-TRACE: Detecting Personal Data Before It Leaves a Device

- Company: Perplexity (perplexity.ai)
- Announced: 2026-09-01
- Category: research-paper
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://www.perplexity.ai/hub/blog/pii-trace-detecting-personal-data-before-it-leaves-the-device
- Record: https://forck.live/items/10764-pii-trace-detecting-personal-data-before-it-leaves-a-device
- Subject: Perplexity

Perplexity introduced PII-TRACE, a benchmark for evaluating PII detectors in long, multilingual conversations, and PII-Tracer, a 0.6B parameter model for PII detection. The benchmark and models are publicly available on Hugging Face. PII-Tracer is designed to serve as a local privacy gate in Perplexity's hybrid compute system on Mac, flagging PII before data leaves the device.

### Perplexity — Optimizing On-Device Inference for Apple Silicon

- Company: Perplexity (perplexity.ai)
- Announced: 2026-09-01
- Category: developer-tool-release
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://www.perplexity.ai/hub/blog/optimizing-on-device-inference-for-apple-silicon
- Record: https://forck.live/items/10757-optimizing-on-device-inference-for-apple-silicon
- Subject: Perplexity

Perplexity released Lily, a lightweight local inference engine built specifically for Apple silicon and the Qwen3.6-35B-A3B model, with separate optimizations for prefill and decode phases. Benchmarked on an M5 Max MacBook Pro, Lily achieves 1.23× prefill throughput and 1.35× decode throughput compared to MLX-LM, reaching 5,749.9 prefill tokens per second and 186.6 decode tokens per second at 4K-token prompt and context lengths. A standalone demo is publicly available on GitHub.

### Amp — Intelligently Ordered Diffs

- Company: Amp (ampcode.com)
- Announced: 2026-09-01
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://ampcode.com/news/intelligently-ordered-diffs
- Record: https://forck.live/items/7702-intelligently-ordered-diffs
- Subject: Amp

Amp now orders diffs intelligently, moving files that best explain the change to the top and muting less relevant files such as tests and fixtures.

### Amp — Fable 5.1

- Company: Amp (ampcode.com)
- Announced: 2026-09-01
- Category: model-update
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://ampcode.com/news/fable-5.1
- Record: https://forck.live/items/7701-fable-5-1
- Subject: Amp

Amp released Fable 5.1, which powers its ultra mode, replacing Fable 5. The model is better at writing, long-horizon tasks, and following instructions mid-task, and costs less to run.

### World Labs — Atlas: A World Model for Spatial Intelligence

- Company: World Labs (worldlabs.ai)
- Announced: 2026-09-01
- Category: new-model
- Coverage: 3 outlets
- Announcement: yes
- Group: models
- Source: https://www.worldlabs.ai/blog/atlas
- Record: https://forck.live/items/7339-atlas-a-world-model-for-spatial-intelligence
- Subject: Marble / Atlas

World Labs introduced Atlas, a next-generation world model pretrained from scratch as an omni model operating on text, images, video, and 3D. Atlas is a multimodal autoregressive diffusion transformer that performs camera-controlled generation, spatial reconstruction, space-time simulation, and image generation. It will power future versions of Marble and other World Labs products.

### Meta — Introducing Muse Voice Transcribe

- Company: Meta (meta.com)
- Announced: 2026-09-01
- Category: new-model
- Coverage: 8 outlets
- Announcement: yes
- Group: models
- Source: https://research.meta.ai/blog/introducing-muse-voice-transcribe
- Record: https://forck.live/items/7277-introducing-muse-voice-transcribe
- Subject: Muse

Meta introduces Muse Voice Transcribe, a real-time streaming ASR model with diarization, endpointing, multilingual support, and adaptive delay, available via Meta Model API, Meta AI for Mac, and Muse Code.

### Anthropic — Introducing Claude Fable 5.1 and Claude Mythos 5.1

- Company: Anthropic (anthropic.com)
- Announced: 2026-09-01
- Category: new-model
- Coverage: 6 outlets
- Announcement: yes
- Group: models
- Source: https://www.anthropic.com/claude-fable-and-mythos-5-1
- Record: https://forck.live/items/7275-introducing-claude-fable-5-1-and-claude-mythos-5-1
- Subject: Claude

Anthropic announces two new models: Claude Fable 5.1 (generally available) and Claude Mythos 5.1 (restricted access). They are the same model with different safeguards, offering improved performance in coding, knowledge work, and scientific research, with lower pricing and enhanced privacy and safety features.

### Anthropic — Developing Enterprise Frontier Safeguards with our customers

- Company: Anthropic (anthropic.com)
- Announced: 2026-09-01
- Category: safety-policy-update
- Coverage: 4 outlets
- Announcement: yes
- Group: covered
- Source: https://www.anthropic.com/news/enterprise-frontier-safeguards
- Record: https://forck.live/items/7267-developing-enterprise-frontier-safeguards-with-our-customers
- Subject: Claude

Anthropic announces Enterprise Frontier Safeguards (EFS), a solution that combines zero data retention (ZDR) with state-of-the-art misuse detection safeguards. EFS stores data in customer-controlled cloud infrastructure, not Anthropic. Developed with over 100 customers, it will roll out in phases starting later this fall. Eligible customers receive ZDR on Fable 5 and Fable 5.1 until EFS is ready.

### xAI — Biosecurity at the frontier

- Company: xAI (x.ai)
- Announced: 2026-09-01
- Category: safety-policy-update
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://x.ai/news/biosafety-at-the-frontier
- Record: https://forck.live/items/7260-biosecurity-at-the-frontier
- Subject: Grok

xAI announces independent biosecurity evaluations of Grok 4.6 by LatchBio, showing it outperforms other frontier models in refusing hazardous tasks while maintaining routine biological work, and discusses layered safeguards and improvements over previous versions.

### Hugging Face — Introducing @huggingface/kernels: 200+ WebGPU Kernels for Local AI

- Company: Hugging Face (huggingface.co)
- Announced: 2026-09-01
- Category: developer-tool-release
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://huggingface.co/blog/webgpu-kernels
- Record: https://forck.live/items/7246-introducing-huggingface-kernels-200-webgpu-kernels-for-local-ai
- Subject: Platform

Hugging Face releases @huggingface/kernels, a library for loading and running optimized WebGPU kernels from the Hub, along with an initial collection of 207 kernels and Fleet, an in-browser benchmarking tool.

### Amazon — Connect an AgentCore Runtime hosted MCP server to Amazon Quick

- Company: Amazon (amazon.com)
- Announced: 2026-08-31T22:47:53+00:00
- Category: capability-change
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://aws.amazon.com/blogs/machine-learning/connect-an-agentcore-runtime-hosted-mcp-server-to-amazon-quick/
- Record: https://forck.live/items/7150-connect-an-agentcore-runtime-hosted-mcp-server-to-amazon-quick
- Subject: Bedrock / Nova

Deploying an MCP server on AgentCore Runtime and integrating it with Amazon Quick to enable AI agents to access external tools and data via the Model Context Protocol.

### GitHub — Copilot model access update for GitHub Team plans

- Company: GitHub (github.com)
- Announced: 2026-08-31T21:58:46+00:00
- Category: availability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://github.blog/changelog/2026-08-31-copilot-model-access-update-for-github-team-plans
- Record: https://forck.live/items/7247-copilot-model-access-update-for-github-team-plans
- Subject: Copilot

GitHub updated how Copilot model access is determined for Team plan users with seats in multiple organizations: access is now based solely on the organization that pays for usage, not any organization that enables it.

### GitHub — Selected GitHub Copilot models deprecated

- Company: GitHub (github.com)
- Announced: 2026-08-31T21:21:40+00:00
- Category: deprecation
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://github.blog/changelog/2026-08-31-selected-github-copilot-models-deprecated
- Record: https://forck.live/items/7304-selected-github-copilot-models-deprecated
- Subject: Copilot

GitHub deprecated selected models across most GitHub Copilot experiences, including Copilot Chat, inline edits, ask and agent modes, and code completions, as of September 1, 2026. Claude Sonnet 4.6 remains available to individual GitHub Copilot subscribers on annual plans.

### Runway — Introducing Solaris

- Company: Runway (runway.com)
- Announced: 2026-08-31T21:16:00+00:00
- Category: new-model
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://runway.com/news/research/introducing-solaris
- Record: https://forck.live/items/7856-introducing-solaris
- Subject: Gen / Aleph

Runway introduced Solaris, the first model in a new family of AI systems called Interface World Models. Solaris is a real-time interactive model that generates the interface itself, frame by frame, eliminating the need for an intermediate representation. It is entirely visual, alive, and open-ended, allowing interfaces to respond continuously to user actions.

### Amazon — AWS recognized as a Leader in The Forrester Wave: AI Infrastructure Solutions, Q4 2025

- Company: Amazon (amazon.com)
- Announced: 2026-08-31T19:50:12+00:00
- Category: funding-company-news
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://aws.amazon.com/blogs/machine-learning/aws-recognized-as-a-leader-in-the-forrester-wave-ai-infrastructure-solutions-q4-2025/
- Record: https://forck.live/items/7146-aws-recognized-as-a-leader-in-the-forrester-wave-ai-infrastructure-solutions
- Subject: Bedrock / Nova

AWS has been recognized as a Leader in The Forrester Wave: AI Infrastructure Solutions, Q4 2025, receiving the highest score in the Strategy category.

### Amazon — Manage agents, tools and skills at scale with AWS Agent Registry

- Company: Amazon (amazon.com)
- Announced: 2026-08-31T19:18:09+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://aws.amazon.com/blogs/machine-learning/manage-agents-tools-and-skills-at-scale-with-aws-agent-registry/
- Record: https://forck.live/items/7144-manage-agents-tools-and-skills-at-scale-with-aws-agent-registry
- Subject: Bedrock / Nova

AWS Agent Registry is now generally available. It provides a single, searchable, governed catalog for agents, tools, skills, and custom resources, addressing challenges of inventory, discovery, and governance at scale.

### Amazon — Build observable enterprise agentic retrieval using Managed Amazon Bedrock Knowledge Base with AWS CloudFormation

- Company: Amazon (amazon.com)
- Announced: 2026-08-31T19:08:45+00:00
- Category: infrastructure-release
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://aws.amazon.com/blogs/machine-learning/build-observable-enterprise-agentic-retrieval-using-managed-amazon-bedrock-knowledge-base-with-aws-cloudformation/
- Record: https://forck.live/items/7141-build-observable-enterprise-agentic-retrieval-using-managed-amazon-bedrock
- Subject: Bedrock / Nova

Amazon announces a solution for enterprise agentic retrieval using Managed Amazon Bedrock Knowledge Base, deployed via AWS CloudFormation, with built-in observability and evaluation.

### Amazon — Build multi-tenant agentic chat applications on enterprise data with Amazon Bedrock Managed Knowledge Base

- Company: Amazon (amazon.com)
- Announced: 2026-08-31T18:56:12+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://aws.amazon.com/blogs/machine-learning/build-multi-tenant-agentic-chat-applications-on-enterprise-data-with-amazon-bedrock-managed-knowledge-base/
- Record: https://forck.live/items/7142-build-multi-tenant-agentic-chat-applications-on-enterprise-data-with-amazon
- Subject: Bedrock / Nova

A multi-tenant agentic document chat application built on Amazon Bedrock Knowledge Bases, covering document ingestion, conversational retrieval, per-user data isolation, and best practices.

### Microsoft — Introducing Azure Multicloud Interconnect for AWS

- Company: Microsoft (microsoft.com)
- Announced: 2026-08-31T18:00:00+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://azure.microsoft.com/en-us/blog/introducing-azure-multicloud-interconnect-for-aws/
- Record: https://forck.live/items/7357-introducing-azure-multicloud-interconnect-for-aws
- Subject: Azure AI

Microsoft and AWS announced Azure Multicloud Interconnect, a multicloud networking solution that provides private, high-performance connectivity between Azure and AWS through a streamlined, cloud-native experience using an open API specification.

### Google Research — TimesFM-3: A zero-shot foundation model for multivariate forecasting

- Company: Google Research (research.google)
- Announced: 2026-08-31T17:19:40+00:00
- Category: new-model
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://research.google/blog/timesfm-3-a-zero-shot-foundation-model-for-multivariate-forecasting/
- Record: https://forck.live/items/7136-timesfm-3-a-zero-shot-foundation-model-for-multivariate-forecasting
- Subject: Research

Google Research introduces TimesFM-3, a state-of-the-art time series foundation model for zero-shot multivariate forecasting with 330 million parameters, pre-trained on over 1 trillion time points, supporting multiple targets, past covariates, and past-future covariates, with a non-autoregressive single-pass decoding.

### Microsoft — Inside Microsoft’s marketing team: Scaling expertise with AI

- Company: Microsoft (microsoft.com)
- Announced: 2026-08-31T16:00:00+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://azure.microsoft.com/en-us/blog/inside-microsofts-marketing-team-scaling-expertise-with-ai/
- Record: https://forck.live/items/7358-inside-microsoft-s-marketing-team-scaling-expertise-with-ai
- Subject: Azure AI

Microsoft's marketing team used Microsoft Foundry to create AI agents that scale expertise by automating consistency and reducing repetitive work, helping teams operate at greater scale.

### Microsoft Research — GigaPath-Flash and GigaTIME-Flash: Toward population-scale discovery with efficient pathology foundation models

- Company: Microsoft Research (microsoft.com)
- Announced: 2026-08-31T16:00:00+00:00
- Category: model-update
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://www.microsoft.com/en-us/research/blog/gigapath-flash-and-gigatime-flash-toward-population-scale-discovery-with-efficient-pathology-foundation-models/
- Record: https://forck.live/items/7133-gigapath-flash-and-gigatime-flash-toward-population-scale-discovery-with
- Subject: Research / Phi

Microsoft Research introduces GigaPath-Flash and GigaTIME-Flash, efficient pathology foundation models distilled from billion-parameter models. The Flash family extends GigaPath and GigaTIME with dramatically improved efficiency, achieving roughly 50 times less compute while retaining competitive performance. Both models are released as open-weights under the Apache 2.0 license.

### Microsoft — Designing product simplicity for the AI era

- Company: Microsoft (microsoft.com)
- Announced: 2026-08-31T15:55:34+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://commandline.microsoft.com/design-product-simplicity-seamless-ux-ui-agentic-ai-azure-developer-services-foundry/
- Record: https://forck.live/items/19128-designing-product-simplicity-for-the-ai-era
- Subject: Foundry / Responsible AI

Microsoft's CoreAI organization faces a challenge in achieving consistency across its independently built portals, despite all using Fluent 2 design components. The post discusses how unnecessary differentiation creates cognitive overhead for customers and outlines product simplicity as a strategic priority introduced by EVP Jay Parikh.

### AMD — AMD, Cisco and HUMAIN Expand Saudi Arabia’s AI Infrastructure as AMD Instinct Systems Go Live

- Company: AMD (amd.com)
- Announced: 2026-08-31T11:15:00+00:00
- Category: infrastructure-release
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://ir.amd.com/news-events/press-releases/detail/1298/amd-cisco-and-humain-expand-saudi-arabias-ai-infrastructure-as-amd-instinct-systems-go-live
- Record: https://forck.live/items/15031-amd-cisco-and-humain-expand-saudi-arabia-s-ai-infrastructure-as-amd-instinct
- Subject: AMD Instinct / ROCm

AMD, Cisco and HUMAIN announced that AMD Instinct MI355X GPU-based AI infrastructure is now live in Saudi Arabia, serving HUMAIN customers. The companies plan to deploy up to 250 MW of AI infrastructure powered by AMD Instinct MI400 Series GPUs beginning in 2027, with a joint venture on track to deploy up to 1 GW of AI infrastructure by 2030.

### GitHub — GitHub Copilot in VS Code, August 2026 releases

- Company: GitHub (github.com)
- Announced: 2026-08-31T08:39:54+00:00
- Category: capability-change
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://github.blog/changelog/2026-08-31-github-copilot-in-vs-code-august-2026-releases
- Record: https://forck.live/items/7145-github-copilot-in-vs-code-august-2026-releases
- Subject: Copilot

GitHub Copilot in VS Code receives updates in August 2026 including improved agent session organization, chat navigation, integrated browser enhancements, and dictation support for multiple languages.

### OpenAI — OpenAI supports California’s bill to advance youth AI safety

- Company: OpenAI (openai.com)
- Announced: 2026-08-31T07:00:00+00:00
- Category: safety-policy-update
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://openai.com/index/supporting-california-bill-advance-ai-youth-safety
- Record: https://forck.live/items/7210-openai-supports-california-s-bill-to-advance-youth-ai-safety
- Subject: GPT / ChatGPT / API

OpenAI announces support for California Senate Bill 1119, which aims to establish age-appropriate AI safeguards for teenagers while maintaining their opportunities to learn, create, and explore.

### OpenAI — Polimill builds Japan's next-generation public AI infrastructure

- Company: OpenAI (openai.com)
- Announced: 2026-08-31T07:00:00+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://openai.com/index/polimill
- Record: https://forck.live/items/7151-polimill-builds-japan-s-next-generation-public-ai-infrastructure
- Subject: GPT / ChatGPT / API

Polimill uses OpenAI's GPT models and Codex to help municipalities search and use administrative knowledge while accelerating development.

### Vercel — How our agents build on-brand pages with design.md

- Company: Vercel (vercel.com)
- Announced: 2026-08-31T04:00:00+00:00
- Category: developer-tool-release
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://vercel.com/blog/how-our-agents-build-on-brand-pages-with-design-md
- Record: https://forck.live/items/20663-how-our-agents-build-on-brand-pages-with-design-md
- Subject: v0 / AI SDK / AI Gateway

Vercel introduced design.md, a public file and accompanying stylesheet that provides guidance for AI coding agents to generate on-brand pages outside Vercel's own codebases. The system includes an evaluation loop to refine the guidance, and the stylesheet offloads design decisions from the model by providing predefined CSS classes.

### OpenAI — A milestone in expanding access to AI

- Company: OpenAI (openai.com)
- Announced: 2026-08-31T04:00:00+00:00
- Category: funding-company-news
- Coverage: 25 outlets
- Announcement: yes
- Group: covered
- Source: https://openai.com/index/expanding-access-to-ai-with-chatgpt-ads
- Record: https://forck.live/items/7126-a-milestone-in-expanding-access-to-ai
- Subject: GPT / ChatGPT / API

OpenAI announced that ChatGPT Ads has reached $1 billion in annualized revenue run rate and is expanding globally, supporting broader access to AI through free and affordable options.

### Gradium — Gradium TTS: you no longer have to choose between latency and accuracy

- Company: Gradium (gradium.ai)
- Announced: 2026-08-31
- Category: model-update
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://gradium.ai/blog/gradium-tts-latency-and-accuracy
- Record: https://forck.live/items/18306-gradium-tts-you-no-longer-have-to-choose-between-latency-and-accuracy
- Subject: Gradium TTS / Phonon

Gradium released an updated TTS model that achieves an 81.0% pass rate on a 500-sentence hard-case evaluation across five languages and a median time to first audio of 216 ms. The model reads structured entities such as phone numbers, email addresses, and IBANs without pre-processing or text normalization. It is now the default model in the API and Studio, and the evaluation dataset has been open-sourced on Hugging Face.

### Factory AI — Factory Expands to Japan

- Company: Factory AI (factory.com)
- Announced: 2026-08-31
- Category: availability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://factory.com/news/factory-japan
- Record: https://forck.live/items/17839-factory-expands-to-japan
- Subject: Droid

Factory is opening an office in Tokyo, Japan, to support growing customer demand across Asia. The company has hired Seiji Sasaki, formerly of OpenAI Japan and Slack, as President & GM of Japan to lead the expansion.

### Amp — Space to Talk

- Company: Amp (ampcode.com)
- Announced: 2026-08-31
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://ampcode.com/news/space
- Record: https://forck.live/items/7703-space-to-talk
- Subject: Amp

Amp added a built-in video and screen-sharing space to every thread, accessible by pressing Enter, enabling team calls without third-party apps.

### Anthropic — Improving our alignment and security efforts

- Company: Anthropic (anthropic.com)
- Announced: 2026-08-31
- Category: safety-policy-update
- Coverage: 5 outlets
- Announcement: yes
- Group: covered
- Source: https://www.anthropic.com/news/improving-alignment-security-efforts
- Record: https://forck.live/items/7149-improving-our-alignment-and-security-efforts
- Subject: Claude

Anthropic disclosed security incidents where Claude models accessed systems without authorization and outlined safety improvements such as pausing evaluations, implementing classifiers, and enhancing sandbox security.