# forck.live — 27 October – 2 November 2025

> 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 2025-W44, 27 October – 2 November 2025, 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/2025-W44
- The current issue, and the catalogue of the other weeks: https://forck.live/briefing.md
- Catalogue: https://forck.live/llms.txt

## Announcements

### Runway — Runway Gen-4.5: State-of-the-Art AI Video Generation

- Company: Runway (runway.com)
- Announced: 2025-11-01T16:22:35.195000+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://runway.com/research/introducing-runway-gen-4.5
- Record: https://forck.live/items/7901-runway-gen-4-5-state-of-the-art-ai-video-generation
- Subject: Gen / Aleph

### Vercel — Vercel Agent can now run AI investigations

- Company: Vercel (vercel.com)
- Announced: 2025-10-31T13:00:00+00:00
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://vercel.com/blog/vercel-agent-can-now-run-ai-investigations
- Record: https://forck.live/items/20796-vercel-agent-can-now-run-ai-investigations
- Subject: v0 / AI SDK / AI Gateway

Vercel launched Vercel Agent Investigations in Public Beta, a new skill that automatically detects issues, conducts root cause analysis, and provides remediation plans for incident response. It builds on anomaly alerts available to Observability Plus subscribers, with no configuration needed. New Agent users receive a $100 free credit.

### Vercel — BotID Deep Analysis catches a sophisticated bot network in real-time

- Company: Vercel (vercel.com)
- Announced: 2025-10-31T13:00:00+00:00
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://vercel.com/blog/botid-deep-analysis-catches-a-sophisticated-bot-network-in-real-time
- Record: https://forck.live/items/20795-botid-deep-analysis-catches-a-sophisticated-bot-network-in-real-time
- Subject: v0 / AI SDK / AI Gateway

Vercel's BotID Deep Analysis automatically detected and blocked a novel browser bot network in under 10 minutes without manual intervention. The system identified coordinated activity by correlating identical browser fingerprints across proxy IP addresses, a pattern that evades traditional defenses.

### Harvey — Harvey’s Principles for AI Adoption and Rollout in Law Schools

- Company: Harvey (harvey.ai)
- Announced: 2025-10-31T12:00:00+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://www.harvey.ai/blog/harveys-principles-for-ai-adoption-and-rollout-in-law-schools
- Record: https://forck.live/items/17658-harvey-s-principles-for-ai-adoption-and-rollout-in-law-schools
- Subject: Harvey

Harvey published a set of five principles and a four-stage rollout framework for law schools to integrate AI into legal education responsibly, based on its work with over 20 schools through its Law School Program. The principles state that AI should augment rather than replace thinking, require human verification, promote proficiency through practice, mandate transparency, and ensure equitable access. The rollout stages cover policy development, structured assignments, and encouraging experimentation.

### Google Research — Accelerating the magic cycle of research breakthroughs and real-world applications

- Company: Google Research (research.google)
- Announced: 2025-10-31T07:40:00+00:00
- Category: research-paper
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://research.google/blog/accelerating-the-magic-cycle-of-research-breakthroughs-and-real-world-applications/
- Record: https://forck.live/items/1397-accelerating-the-magic-cycle-of-research-breakthroughs-and-real-world
- Subject: Research

Google Research announces new scientific breakthroughs in earth science, genomics, and quantum computing, including Earth AI, DeepSomatic, and Quantum Echoes, and highlights how AI tools accelerate innovation.

### OpenAI — Expanding Stargate to Michigan

- Company: OpenAI (openai.com)
- Announced: 2025-10-30T13:30:00+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://openai.com/index/expanding-stargate-to-michigan
- Record: https://forck.live/items/400-expanding-stargate-to-michigan
- Subject: GPT / ChatGPT / API

OpenAI is expanding its Stargate project to Michigan with a new one-gigawatt campus to strengthen AI infrastructure.

### Stability AI — Universal Music Group and Stability AI Announce Strategic Alliance to Co-Develop Professional AI Music Creation Tools

- Company: Stability AI (stability.ai)
- Announced: 2025-10-30T12:06:00+00:00
- Category: partnership-acquisition
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://stability.ai/news-updates/universal-music-group-and-stability-ai-announce-strategic-alliance
- Record: https://forck.live/items/17479-universal-music-group-and-stability-ai-announce-strategic-alliance-to-co
- Subject: Stable Diffusion

Universal Music Group and Stability AI announced a strategic alliance to co-develop professional AI music creation tools, centering artists in the development process and prioritizing feedback from the creative community to guide the creation of fully licensed, commercially safe AI music tools.

### OpenAI — Introducing Aardvark: OpenAI’s agentic security researcher

- Company: OpenAI (openai.com)
- Announced: 2025-10-30T11:00:00+00:00
- Category: product-launch
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://openai.com/index/introducing-aardvark
- Record: https://forck.live/items/401-introducing-aardvark-openai-s-agentic-security-researcher
- Subject: GPT / ChatGPT / API

OpenAI announces Aardvark, an AI-powered security researcher that autonomously finds, validates, and helps fix software vulnerabilities at scale. The system is currently in private beta with sign-ups for early testing.

### Google Research — Toward provably private insights into AI use

- Company: Google Research (research.google)
- Announced: 2025-10-30T10:56:00+00:00
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://research.google/blog/toward-provably-private-insights-into-ai-use/
- Record: https://forck.live/items/1398-toward-provably-private-insights-into-ai-use
- Subject: Research

Google Research announced a new provably private insights (PPI) system that combines large language models, differential privacy, and trusted execution environments to analyze unstructured generative AI data, deployed in the Recorder app on Pixel and open-sourced as part of confidential federated analytics.

### MiniMax — Aligning to What? Rethinking Agent Generalization in MiniMax M2

- Company: MiniMax (minimax.io)
- Announced: 2025-10-30T10:03:45+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://huggingface.co/blog/MiniMax-AI/aligning-to-what
- Record: https://forck.live/items/1589-aligning-to-what-rethinking-agent-generalization-in-minimax-m2

The challenges of agent generalization for the MiniMax M2 model, emphasizing the need for interleaved thinking and the insight that true generalization requires adapting to perturbations across the entire operational space, not just tool scaling.

### Amp — Workspace Entitlements

- Company: Amp (ampcode.com)
- Announced: 2025-10-30
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://ampcode.com/news/workspace-entitlements
- Record: https://forck.live/items/7788-workspace-entitlements
- Subject: Amp

Amp added per-user spending limits called entitlements to its Enterprise Premium workspaces, with configurable defaults such as $50/week for regular users and $200/week for senior engineers.

### OpenAI — How we built OWL, the new architecture behind our ChatGPT-based browser, Atlas

- Company: OpenAI (openai.com)
- Announced: 2025-10-30
- Category: capability-change
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://openai.com/index/building-chatgpt-atlas
- Record: https://forck.live/items/402-how-we-built-owl-the-new-architecture-behind-our-chatgpt-based-browser-atlas
- Subject: GPT / ChatGPT / API

OpenAI describes OWL, a new architecture for ChatGPT Atlas that decouples Chromium to enable fast startup, rich UI, and agentic browsing.

### Google Research — StreetReaderAI: Towards making street view accessible via context-aware multimodal AI

- Company: Google Research (research.google)
- Announced: 2025-10-29T16:38:00+00:00
- Category: research-paper
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://research.google/blog/streetreaderai-towards-making-street-view-accessible-via-context-aware-multimodal-ai/
- Record: https://forck.live/items/1399-streetreaderai-towards-making-street-view-accessible-via-context-aware
- Subject: Research

Google Research introduced StreetReaderAI, a proof-of-concept accessible street view prototype that uses context-aware, real-time multimodal AI and accessible navigation controls, presented at UIST'25.

### Sakana AI — Applied Team メンバー紹介

- Company: Sakana AI (sakana.ai)
- Announced: 2025-10-29T15:00:00+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://sakana.ai/applied-team-profiles
- Record: https://forck.live/items/4705-applied-team
- Title in English (translated by forck.live from ja): Applied Team Member Introduction
- Subject: Sakana models

Sakana AI announced the launch of its Business Development Division to begin providing services to customers, focusing on solving business and social challenges for Japanese companies using cutting-edge AI. The post introduces several members of the Applied Team, including their backgrounds, missions, and technical interests.

### Google DeepMind — Accelerating discovery with the AI for Math Initiative

- Company: Google DeepMind (deepmind.google)
- Announced: 2025-10-29T14:31:13+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://deepmind.google/blog/accelerating-discovery-with-the-ai-for-math-initiative/
- Record: https://forck.live/items/1282-accelerating-discovery-with-the-ai-for-math-initiative
- Title in English (translated by forck.live from en): Accelerating discovery with the AI for Math Initiative
- Subject: Gemini

Google DeepMind announced the AI for Math Initiative, which brings together prestigious research institutions to pioneer the use of AI in mathematical research.

### Hugging Face — On the Shifting Global Compute Landscape

- Company: Hugging Face (huggingface.co)
- Announced: 2025-10-29T13:56:45+00:00
- Category: research-paper
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://huggingface.co/blog/huggingface/shifting-compute-landscape
- Record: https://forck.live/items/1590-on-the-shifting-global-compute-landscape
- Subject: Platform

U.S. export controls on AI chips have accelerated China's domestic chip production and algorithmic innovations, leading to a shift in the global compute landscape where Chinese open-weight models and chips are gaining prominence.

### Vercel — Vercel achieves TISAX AL2 compliance to serve automotive partners

- Company: Vercel (vercel.com)
- Announced: 2025-10-29T13:00:00+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://vercel.com/blog/vercel-achieves-tisax-al2-compliance-to-serve-automotive-partners
- Record: https://forck.live/items/20797-vercel-achieves-tisax-al2-compliance-to-serve-automotive-partners
- Subject: v0 / AI SDK / AI Gateway

Vercel has completed its assessment for TISAX Level 2 (AL2), an internationally-recognized security standard for the automotive industry. The assessment was independently reviewed by an ENX-accredited auditor, and results are accessible to registered ENX portal participants.

### Harvey — Building Harvey’s iManage Integration

- Company: Harvey (harvey.ai)
- Announced: 2025-10-29T13:00:00+00:00
- Category: capability-change
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://www.harvey.ai/blog/building-harveys-imanage-integration
- Record: https://forck.live/items/17659-building-harvey-s-imanage-integration
- Subject: Harvey

Harvey announced a direct integration with iManage, allowing users to pull documents into Harvey, work with them, and push artifacts back without exports or duplicates. The integration includes end-to-end encryption, OAuth 2.0 authentication, and a distributed rate limiter to manage API quota across product surfaces. Harvey also built support for iManage on-premises environments, addressing connectivity challenges for enterprise customers.

### Thinking Machines Lab — Tinker: Announcing Research and Teaching Grants

- Company: Thinking Machines Lab (thinkingmachines.ai)
- Announced: 2025-10-29
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://thinkingmachines.ai/news/tinker-research-and-teaching-grants
- Record: https://forck.live/items/18563-tinker-announcing-research-and-teaching-grants
- Subject: Thinking Machines / Inkling

Thinking Machines Lab is offering research and teaching grants for access to its Tinker platform, providing $250 in free credits per student for academic classes and grants starting at $5,000 for research projects and open-source software.

### Genspark — Genspark Hub: Dedicated Space and Memory for Every Project

- Company: Genspark (genspark.ai)
- Announced: 2025-10-29
- Category: not stated
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://www.genspark.ai/blog/genspark-hub
- Record: https://forck.live/items/18265-genspark-hub-dedicated-space-and-memory-for-every-project
- Subject: Genspark

### Amp — Amp Now Reads Threads

- Company: Amp (ampcode.com)
- Announced: 2025-10-29
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://ampcode.com/news/read-threads
- Record: https://forck.live/items/7789-amp-now-reads-threads
- Subject: Amp

Amp now allows users to reference threads in messages, and the system fetches and extracts relevant context from them. The feature uses a new read_thread tool that retrieves the thread as Markdown and then uses another model to extract context based on user instructions.

### Cognition — Introducing SWE-1.5: Our Fast Agent Model

- Company: Cognition (cognition.com)
- Announced: 2025-10-29
- Category: new-model
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://cognition.com/blog/swe-1-5
- Record: https://forck.live/items/7687-introducing-swe-1-5-our-fast-agent-model
- Subject: Devin / SWE

Cognition released SWE-1.5, a frontier-size model with hundreds of billions of parameters optimized for software engineering, achieving near-SOTA coding performance. The model is served at up to 950 tok/s through a partnership with Cerebras, making it 6x faster than Haiku 4.5 and 13x faster than Sonnet 4.5. SWE-1.5 is now available in Windsurf.

### Hugging Face — Building a Healthcare Robot from Simulation to Deployment with NVIDIA Isaac

- Company: Hugging Face (huggingface.co)
- Announced: 2025-10-29
- Category: developer-tool-release
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://huggingface.co/blog/lerobotxnvidia-healthcare
- Record: https://forck.live/items/1591-building-a-healthcare-robot-from-simulation-to-deployment-with-nvidia-isaac
- Subject: Platform

A workflow for building a surgical assistant robot using NVIDIA Isaac for Healthcare v0.4, including data collection, training with GR00T N1.5, and deployment. It highlights that over 93% of training data was synthetic.

### OpenAI — gpt-oss-safeguard technical report

- Company: OpenAI (openai.com)
- Announced: 2025-10-29
- Category: new-model
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://openai.com/index/gpt-oss-safeguard-technical-report
- Record: https://forck.live/items/404-gpt-oss-safeguard-technical-report
- Subject: GPT / ChatGPT / API

OpenAI released two open-weight reasoning models, gpt-oss-safeguard-120b and gpt-oss-safeguard-20b, post-trained from the gpt-oss models to label content under a provided policy.

### OpenAI — Introducing gpt-oss-safeguard

- Company: OpenAI (openai.com)
- Announced: 2025-10-29
- Category: open-weight-release
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://openai.com/index/introducing-gpt-oss-safeguard
- Record: https://forck.live/items/403-introducing-gpt-oss-safeguard
- Subject: GPT / ChatGPT / API

OpenAI introduces gpt-oss-safeguard, open-weight reasoning models for safety classification that allow developers to apply and iterate on custom policies.

### NVIDIA — How to Build a Healthcare Robot from Simulation to Deployment with NVIDIA Isaac for Healthcare

- Company: NVIDIA (nvidia.com)
- Announced: 2025-10-28T20:42:35+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://huggingface.co/blog/nvidia/nvidia-isaac-for-healthcare
- Record: https://forck.live/items/1592-how-to-build-a-healthcare-robot-from-simulation-to-deployment-with-nvidia

This is a hands-on guide to building a healthcare robot using NVIDIA Isaac for Healthcare, covering data collection, post-training of GR00T N1.5, and deployment to real hardware.

### OpenAI — Advancing organizational transformation for business innovation

- Company: OpenAI (openai.com)
- Announced: 2025-10-28T17:00:00+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://openai.com/index/dai-nippon-printing
- Record: https://forck.live/items/405-advancing-organizational-transformation-for-business-innovation
- Subject: GPT / ChatGPT / API

Dai Nippon Printing (DNP) deployed ChatGPT Enterprise across ten core departments, achieving significant improvements in patent research speed, processing volume, automation, and knowledge reuse within three months.

### Hugging Face — Granite 4.0 Nano: Just how small can you go?

- Company: Hugging Face (huggingface.co)
- Announced: 2025-10-28T14:59:38+00:00
- Category: new-model
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://huggingface.co/blog/ibm-granite/granite-4-nano
- Record: https://forck.live/items/1593-granite-4-0-nano-just-how-small-can-you-go
- Subject: Platform

IBM released Granite 4.0 Nano, a family of small language models designed for edge and on-device applications.

### Vercel — Bun runtime on Vercel Functions

- Company: Vercel (vercel.com)
- Announced: 2025-10-28T13:00:00+00:00
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://vercel.com/blog/bun-runtime-on-vercel-functions
- Record: https://forck.live/items/20798-bun-runtime-on-vercel-functions
- Subject: v0 / AI SDK / AI Gateway

Vercel added Bun as a runtime option for Vercel Functions, available in Public Beta. Internal testing showed Bun reduced average latency by 28% in CPU-bound Next.js rendering workloads compared to Node.js. The Bun runtime supports Next.js, Express, Hono, and Nitro, and runs on Fluid compute with active CPU pricing.

### OpenAI — Doppel’s AI defense system stops attacks before they spread

- Company: OpenAI (openai.com)
- Announced: 2025-10-28T10:00:00+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://openai.com/index/doppel
- Record: https://forck.live/items/406-doppel-s-ai-defense-system-stops-attacks-before-they-spread
- Subject: GPT / ChatGPT / API

Doppel uses GPT-5 and reinforcement fine-tuning to stop deepfake and impersonation attacks, reducing analyst workloads by 80% and response times from hours to minutes.

### OpenAI — Built to benefit everyone

- Company: OpenAI (openai.com)
- Announced: 2025-10-28T06:00:00+00:00
- Category: funding-company-news
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://openai.com/index/built-to-benefit-everyone
- Record: https://forck.live/items/408-built-to-benefit-everyone
- Subject: GPT / ChatGPT / API

OpenAI announces a recapitalization that strengthens mission-focused governance and expands resources to ensure AI benefits everyone while advancing innovation responsibly.

### OpenAI — The next chapter of the Microsoft–OpenAI partnership

- Company: OpenAI (openai.com)
- Announced: 2025-10-28T06:00:00+00:00
- Category: partnership-acquisition
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://openai.com/index/next-chapter-of-microsoft-openai-partnership
- Record: https://forck.live/items/407-the-next-chapter-of-the-microsoft-openai-partnership
- Subject: GPT / ChatGPT / API

OpenAI and Microsoft have signed a new agreement that strengthens their long-term partnership, expands innovation, and ensures responsible AI progress.

### Ant Ling — Ming-flash-omni-Preview: A Sparse, Unified Architecture for Multimodal Perception and Generation

- Company: Ant Ling (ant-ling.com)
- Announced: 2025-10-28
- Category: new-model
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://developer.ant-ling.com/en/blogs/ming-flash-omni-preview
- Record: https://forck.live/items/22335-ming-flash-omni-preview-a-sparse-unified-architecture-for-multimodal
- Subject: Ling / Ring / Ming

Ant Ling released Ming-flash-omni-Preview, an open-source omnimodal large model with 103B total parameters and 9B activated, based on Ling 2.0's sparse MoE architecture. It improves over Ming-lite-omni-1.5 in omnimodal understanding and generation, achieving a GenEval score of 0.90 and SOTA on all 12 ContextASR subtasks. The model supports controllable image generation, streaming video understanding, speech and dialect recognition for 15 Chinese dialects, and voice cloning with continuous tokenizers.

### Amp — Command Palette, Not Slash Commands

- Company: Amp (ampcode.com)
- Announced: 2025-10-28
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://ampcode.com/news/command-palette
- Record: https://forck.live/items/7790-command-palette-not-slash-commands
- Subject: Amp

Amp introduced a Command Palette for its CLI and editor extensions, replacing the old slash menu and allowing users to run commands without editing their prompt, accessible at all times.

### Cognition — Automating .NET Framework → .NET Core with Devin

- Company: Cognition (cognition.com)
- Announced: 2025-10-28
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://cognition.com/blog/dotnet-migration-with-devin
- Record: https://forck.live/items/7688-automating-net-framework-net-core-with-devin
- Subject: Devin / SWE

Cognition describes how its autonomous coding agent Devin can automate .NET Framework to .NET Core migration, reducing timelines from months to as little as two weeks. The post details how Devin uses its capabilities to plan, execute, and validate each stage of a Strangler Fig migration.

### LG AI Research — EXAONE, transcending language barriers into the world

- Company: LG AI Research (lgresearch.ai)
- Announced: 2025-10-28
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://www.lgresearch.ai/blog/view?seq=592
- Record: https://forck.live/items/4775-exaone-transcending-language-barriers-into-the-world
- Subject: EXAONE

An interview with Sunkyoung Kim, an AI researcher at LG AI Research's EXAONE Lab, discussing her background, work on data for EXAONE's pre-training, and challenges in multilingual data.

### Together AI — Dynamic AI agent testing for the real world with Collinear Simulations and Together Evals

- Company: Together AI (together.ai)
- Announced: 2025-10-28
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://www.together.ai/blog/collinear-simulations-together-evals
- Record: https://forck.live/items/2403-dynamic-ai-agent-testing-for-the-real-world-with-collinear-simulations-and
- Subject: Inference platform

Together AI announces an integration with Collinear Simulations' TraitMix product, enabling dynamic, persona-driven AI agent testing via the Together Evals API. The combined pipeline allows teams to generate realistic multi-turn conversations with controllable user traits and automatically evaluate agent performance using standardized rubrics.

### Hugging Face — Voice Cloning with Consent

- Company: Hugging Face (huggingface.co)
- Announced: 2025-10-28
- Category: research-paper
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://huggingface.co/blog/voice-consent-gate
- Record: https://forck.live/items/1594-voice-cloning-with-consent
- Subject: Platform

The concept of a 'voice consent gate' to support voice cloning with consent, where a voice can be cloned only when the speaker explicitly says they consent. It provides an example Space and accompanying code to demonstrate the idea.

### Google Research — How we are building the personal health coach

- Company: Google Research (research.google)
- Announced: 2025-10-27T22:36:00+00:00
- Category: availability-change
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://research.google/blog/how-we-are-building-the-personal-health-coach/
- Record: https://forck.live/items/1400-how-we-are-building-the-personal-health-coach
- Subject: Research

Google Research announces a public preview of a personal health coach powered by Gemini models, rolling out to eligible US-based Fitbit Premium Android users starting tomorrow.

### Vercel — Vercel Ship AI 2025 recap

- Company: Vercel (vercel.com)
- Announced: 2025-10-27T13:00:00+00:00
- Category: capability-change
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://vercel.com/blog/ship-ai-2025-recap
- Record: https://forck.live/items/20800-vercel-ship-ai-2025-recap
- Subject: v0 / AI SDK / AI Gateway

Vercel announced several updates at Ship AI 2025, including AI SDK 6 with an agent-first architecture and human-in-the-loop tool approval, a Marketplace for agents and AI tools, a Workflow Development Kit for durable TypeScript functions, the Vercel Agent for automated code reviews and anomaly investigations, zero-configuration Python support, and open-source lead and data analyst agents.

### Vercel — David Totten Joins Vercel to Lead Global Field Engineering

- Company: Vercel (vercel.com)
- Announced: 2025-10-27T13:00:00+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://vercel.com/blog/david-totten-joins-vercel-to-lead-global-field-engineering
- Record: https://forck.live/items/20799-david-totten-joins-vercel-to-lead-global-field-engineering
- Subject: v0 / AI SDK / AI Gateway

Vercel announced that David Totten has joined the company as VP of Global Field Engineering, leading a unified organization that includes Sales Engineering, Developer Success, Professional Services, and Customer Support Engineering. Totten previously worked at Databricks and spent nearly two decades at a Fortune 5 company.

### Harvey — A Framework for Domain-Specific Evaluations

- Company: Harvey (harvey.ai)
- Announced: 2025-10-27T13:00:00+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://www.harvey.ai/blog/a-framework-for-domain-specific-evaluations
- Record: https://forck.live/items/17660-a-framework-for-domain-specific-evaluations
- Subject: Harvey

Harvey is expanding its public evaluation work into four areas: Insights, Research, Approaches, and Context. The company argues that the most important missing piece in LLM evaluation is a shared understanding of what evaluation does and does not tell us, and that misaligned benchmarks risk pushing models toward unhelpful behaviors. Harvey plans to publish at least two novel benchmarks: Contract Intelligence and a BLB Challenge benchmark focused on tasks where leading models struggle.

### OpenAI — Seizing the AI opportunity

- Company: OpenAI (openai.com)
- Announced: 2025-10-27T12:00:00+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://openai.com/global-affairs/seizing-the-ai-opportunity
- Record: https://forck.live/items/409-seizing-the-ai-opportunity
- Subject: GPT / ChatGPT / API

OpenAI submits a document to the White House on strategic investment in energy and infrastructure to meet AI demands and sustain U.S. leadership.

### OpenAI — Addendum to GPT-5 System Card: Sensitive conversations

- Company: OpenAI (openai.com)
- Announced: 2025-10-27T10:00:00+00:00
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://openai.com/index/gpt-5-system-card-sensitive-conversations
- Record: https://forck.live/items/411-addendum-to-gpt-5-system-card-sensitive-conversations
- Subject: GPT / ChatGPT / API

OpenAI published a system card addendum detailing GPT-5's improvements in handling sensitive conversations, including new benchmarks for emotional reliance, mental health, and jailbreak resistance.

### OpenAI — Strengthening ChatGPT’s responses in sensitive conversations

- Company: OpenAI (openai.com)
- Announced: 2025-10-27T10:00:00+00:00
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://openai.com/index/strengthening-chatgpt-responses-in-sensitive-conversations
- Record: https://forck.live/items/410-strengthening-chatgpt-s-responses-in-sensitive-conversations
- Subject: GPT / ChatGPT / API

OpenAI collaborated with over 170 mental health experts to improve ChatGPT's ability to recognize distress, respond empathetically, and guide users toward real-world support, reducing unsafe responses by up to 80%.

### Odyssey — Introducing Odyssey-2: A General-Purpose World Model

- Company: Odyssey (odyssey.systems)
- Announced: 2025-10-27
- Category: new-model
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://odyssey.systems/introducing-odyssey-2
- Record: https://forck.live/items/18604-introducing-odyssey-2-a-general-purpose-world-model
- Subject: Odyssey world models

Odyssey released Odyssey-2, a general-purpose world model that generates and streams interactive AI video in real time. The model is causal and autoregressive, producing a new frame every 50 milliseconds (20 fps) and responding to natural text prompts as the video plays. Odyssey-2 was developed through a novel multi-stage training pipeline that transitions the model from bidirectional to causal behavior.

### Thinking Machines Lab — On-Policy Distillation

- Company: Thinking Machines Lab (thinkingmachines.ai)
- Announced: 2025-10-27
- Category: research-paper
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://thinkingmachines.ai/blog/on-policy-distillation/
- Record: https://forck.live/items/18568-on-policy-distillation
- Subject: Thinking Machines / Inkling

Thinking Machines Lab describes on-policy distillation, a post-training approach that samples trajectories from the student model and uses a high-performing teacher to grade each token. The method aims to combine the on-policy relevance of reinforcement learning with the dense reward signal of distillation, addressing the sparse feedback of RL and the compounding error of off-policy distillation.

### Factory AI — AWS and Factory Announce Strategic Partnership

- Company: Factory AI (factory.com)
- Announced: 2025-10-27
- Category: availability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://factory.com/news/blog-post-2025-10-27
- Record: https://forck.live/items/17819-aws-and-factory-announce-strategic-partnership
- Subject: Droid

Factory announced a strategic partnership with AWS and the availability of its agent-native software development lifecycle platform on AWS Marketplace. Enterprise engineering teams can use existing AWS Enterprise Discount Program (EDP) commitments to procure Factory. The platform integrates with AWS security and compliance standards and supports multi-model access via Amazon Bedrock.

### MiniMax — MiniMax M2 & Agent: Ingenious in Simplicity

- Company: MiniMax (minimax.io)
- Announced: 2025-10-27
- Category: open-weight-release
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://www.minimax.io/blog/minimax-m2-en-1748600000
- Record: https://forck.live/items/3404-minimax-m2-agent-ingenious-in-simplicity
- Subject: MiniMax models

MiniMax launched and open-sourced the M2 model, an agent-first model that delivers top-tier coding, tool use, and deep search at 8% of Claude Sonnet's price with 2x speed. A free API and Agent product are available.

### Hugging Face — huggingface_hub v1.0: Five Years of Building the Foundation of Open Machine Learning

- Company: Hugging Face (huggingface.co)
- Announced: 2025-10-27
- Category: developer-tool-release
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://huggingface.co/blog/huggingface-hub-v1
- Record: https://forck.live/items/1596-huggingface-hub-v1-0-five-years-of-building-the-foundation-of-open-machine
- Subject: Platform

huggingface_hub v1.0 released, marking five years of development. Major changes: migration to httpx, redesigned CLI (hf), and adoption of hf_xet for file transfers.

### Hugging Face — Streaming datasets: 100x More Efficient

- Company: Hugging Face (huggingface.co)
- Announced: 2025-10-27
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://huggingface.co/blog/streaming-datasets
- Record: https://forck.live/items/1595-streaming-datasets-100x-more-efficient
- Subject: Platform

Hugging Face announced significant performance improvements to its datasets library's streaming feature, achieving up to 100x fewer startup requests, 10x faster data resolution, and 2x faster streaming speed, enabling efficient training on multi-TB datasets without downloading.

### OpenAI — Steuerrecht.com delivers client-ready legal analysis with ChatGPT

- Company: OpenAI (openai.com)
- Announced: 2025-10-27
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://openai.com/index/steuerrecht
- Record: https://forck.live/items/412-steuerrecht-com-delivers-client-ready-legal-analysis-with-chatgpt
- Subject: GPT / ChatGPT / API

Steuerrecht.com uses ChatGPT Business to streamline legal workflows, automate tax research, and deliver faster, client-ready analysis for law firms.