# forck.live — 1–7 December 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-W49, 1–7 December 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-W49
- The current issue, and the catalogue of the other weeks: https://forck.live/briefing.md
- Catalogue: https://forck.live/llms.txt

## Announcements

### Lovable — Better planning, better apps: Chat mode & Follow-up questions

- Company: Lovable (lovable.dev)
- Announced: 2025-12-05T15:02:00+00:00
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://lovable.dev/blog/chat-mode-and-questions
- Record: https://forck.live/items/12115-better-planning-better-apps-chat-mode-follow-up-questions
- Subject: Lovable / AI app builder

Lovable introduced Chat Mode, allowing users to discuss and refine their project ideas with the AI before any code is generated, and a follow-up questions feature where the Lovable agent asks clarifying questions about the user's project to reduce assumptions and improve accuracy.

### Vercel — React2Shell Security Bulletin

- Company: Vercel (vercel.com)
- Announced: 2025-12-05T13:00:00+00:00
- Category: safety-policy-update
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://vercel.com/blog/resources-for-protecting-against-react2shell
- Record: https://forck.live/items/20780-react2shell-security-bulletin
- Subject: v0 / AI SDK / AI Gateway

Vercel issued a security bulletin for React2Shell, a critical vulnerability in React Server Components affecting React 19 and Next.js versions 15.0.0 through 16.0.6, and disclosed two additional CVEs (CVE-2025-55184 and CVE-2025-55183) requiring patching. The bulletin provides upgrade instructions, WAF protection details, and automated remediation tools for affected users.

### Supabase — Introducing Supabase for Platforms

- Company: Supabase (supabase.com)
- Announced: 2025-12-05
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://supabase.com/blog/introducing-supabase-for-platforms
- Record: https://forck.live/items/21315-introducing-supabase-for-platforms
- Subject: Supabase

Supabase launched Supabase for Platforms, a white-label offering that lets platforms provision and manage fully managed backends on behalf of their users. The service includes a Management API, remote MCP Server, and Platform Kit UI components. Leading AI builders like Lovable, Bolt.new, and Baidu's MeDo have already used it to create millions of Supabase projects.

### Upstage — Introducing Document Classify: Universal, semantic classification for any document

- Company: Upstage (upstage.ai)
- Announced: 2025-12-05
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://www.upstage.ai/blog/en/introducing-document-classify-universal-semantic-classification-for-any-document
- Record: https://forck.live/items/18458-introducing-document-classify-universal-semantic-classification-for-any-document
- Subject: Solar Pro / Solar Mini

Upstage released Document Classify, a taxonomy-driven document classifier that uses user-defined descriptions instead of fixed categories, allowing zero-shot expansion without model updates. It supports structured, semi-structured, and unstructured documents across formats like PDFs, scanned images, and mobile captures. The model achieved 92.1% accuracy on English and Korean insurance documents in internal evaluations.

### Hugging Face — Introducing swift-huggingface: The Complete Swift Client for Hugging Face

- Company: Hugging Face (huggingface.co)
- Announced: 2025-12-05
- Category: developer-tool-release
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://huggingface.co/blog/swift-huggingface
- Record: https://forck.live/items/1574-introducing-swift-huggingface-the-complete-swift-client-for-hugging-face
- Subject: Platform

Hugging Face released swift-huggingface, a Swift package that provides a full client for the Hugging Face Hub, including network improvements, a shared cache with Python, flexible authentication, and OAuth support.

### Replit — Critical Security Vulnerability in React Server Components

- Company: Replit (replit.com)
- Announced: 2025-12-04T23:13:28+00:00
- Category: safety-policy-update
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://replit.com/blog/critical-security-vulnerability-in-react-server-components
- Record: https://forck.live/items/17931-critical-security-vulnerability-in-react-server-components
- Subject: Replit Agent

Replit announced a critical vulnerability in React Server Components affecting specific versions of reactjs and Next.js. Most Replit applications were not affected upon creation, but those created with the "Mobile app [beta]" option were vulnerable. Replit enabled mitigation in Google Cloud Armor for replit.app deployments and patched the Mobile app option for new apps.

### Alibaba — SAPO: A Stable and Performant Reinforcement Learning Method for Training Large Language Models

- Company: Alibaba (alibaba.com)
- Announced: 2025-12-04T20:00:00+00:00
- Category: research-paper
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://qwen.ai/blog?id=sapo
- Record: https://forck.live/items/4587-sapo-a-stable-and-performant-reinforcement-learning-method-for-training-large
- Subject: Qwen

A new reinforcement learning method called SAPO (Soft Adaptive Policy Optimization) is introduced for training large language models, replacing hard clipping with a smooth, temperature-controlled gating function to improve stability and performance.

### Google Research — Titans + MIRAS: Helping AI have long-term memory

- Company: Google Research (research.google)
- Announced: 2025-12-04T19:26:09+00:00
- Category: research-paper
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://research.google/blog/titans-miras-helping-ai-have-long-term-memory/
- Record: https://forck.live/items/1385-titans-miras-helping-ai-have-long-term-memory
- Subject: Research

Google Research introduces the Titans architecture and MIRAS framework, which enable AI models to update their core memory in real-time, combining the speed of RNNs with the accuracy of transformers. The approach uses a 'surprise metric' to selectively store novel information, allowing efficient handling of massive contexts.

### OpenAI — Introducing OpenAI for Australia

- Company: OpenAI (openai.com)
- Announced: 2025-12-04T19:00:00+00:00
- Category: availability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://openai.com/global-affairs/openai-for-australia
- Record: https://forck.live/items/356-introducing-openai-for-australia
- Subject: GPT / ChatGPT / API

OpenAI is launching an initiative called OpenAI for Australia, which is designed to build sovereign AI infrastructure, provide upskilling for over 1.5 million workers, and accelerate AI innovation in Australia.

### Google DeepMind — Engineering more resilient crops for a warming climate

- Company: Google DeepMind (deepmind.google)
- Announced: 2025-12-04T16:23:24+00:00
- Category: research-paper
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://deepmind.google/blog/engineering-more-resilient-crops-for-a-warming-climate/
- Record: https://forck.live/items/1266-engineering-more-resilient-crops-for-a-warming-climate
- Subject: Gemini

Scientists are using AlphaFold to strengthen a photosynthesis enzyme for resilient, heat-tolerant crops.

### Alibaba — Qwen3-TTS Update! 49 Timbres + 10 Languages + 9 Dialects

- Company: Alibaba (alibaba.com)
- Announced: 2025-12-04T16:00:04+00:00
- Category: model-update
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://qwen.ai/blog?id=qwen3-tts-1128
- Record: https://forck.live/items/4592-qwen3-tts-update-49-timbres-10-languages-9-dialects
- Subject: Qwen

Alibaba updated the Qwen3-TTS model family, including Qwen3-TTS-Flash, with richer timbres (over 49), support for 10 languages and 9 dialects, and improved prosody and speech rate naturalness.

### Harvey — Strengthen Firm and Client Relationships With Shared Spaces

- Company: Harvey (harvey.ai)
- Announced: 2025-12-04T13:10:00+00:00
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://www.harvey.ai/blog/shared-spaces-and-collaboration-in-harvey
- Record: https://forck.live/items/17641-strengthen-firm-and-client-relationships-with-shared-spaces
- Subject: Harvey

Harvey introduced shared Spaces, enabling law firms, clients, and cross-functional teams to collaborate within Harvey with enterprise-grade security and privacy controls. The feature supports sharing Vaults, Workflows, Playbooks, and Artifacts, and will also support guest accounts for clients who are not Harvey customers.

### Harvey — Andreessen Horowitz Leads $160M Investment in Harvey

- Company: Harvey (harvey.ai)
- Announced: 2025-12-04T13:05:35.390000+00:00
- Category: funding-company-news
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://www.harvey.ai/blog/andreessen-horowitz-leads-dollar160m-investment-in-harvey
- Record: https://forck.live/items/17642-andreessen-horowitz-leads-160m-investment-in-harvey
- Subject: Harvey

Harvey announced a $160 million investment at an $8 billion valuation, led by Andreessen Horowitz, with participation from new and returning investors. The company also noted that over 50% of the Am Law 100 law firms now use Harvey, alongside in-house legal teams at Bridgewater, Comcast, and Carvana.

### ElevenLabs — Toyota engages fans with AI-powered Brock Purdy experience

- Company: ElevenLabs (elevenlabs.io)
- Announced: 2025-12-04T12:00:00+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://elevenlabs.io/blog/toyota
- Record: https://forck.live/items/3377-toyota-engages-fans-with-ai-powered-brock-purdy-experience
- Subject: Voice AI

ElevenLabs describes an AI-powered fan engagement experience created for Toyota featuring Brock Purdy.

### Supabase — Build "Sign in with Your App" using Supabase Auth

- Company: Supabase (supabase.com)
- Announced: 2025-12-04
- Category: capability-change
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://supabase.com/blog/oauth2-provider
- Record: https://forck.live/items/21317-build-sign-in-with-your-app-using-supabase-auth
- Subject: Supabase

Supabase added OAuth 2.1 and OpenID Connect server capabilities to its Auth product, allowing Supabase projects to act as identity providers for third-party apps, AI agents via the Model Context Protocol, and enterprise SSO.

### Supabase — Adding Async Streaming to Postgres Foreign Data Wrappers

- Company: Supabase (supabase.com)
- Announced: 2025-12-04
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://supabase.com/blog/adding-async-streaming-to-pg-fdw
- Record: https://forck.live/items/21316-adding-async-streaming-to-postgres-foreign-data-wrappers
- Subject: Supabase

Supabase added asynchronous streaming to its ClickHouse Foreign Data Wrapper using Rust's async runtime and a bounded channel, allowing Postgres to start returning results as soon as the first rows arrive and keeping memory usage predictable even for billion-row scans. The pattern is built into Supabase Wrappers and will be rolled out to connectors for BigQuery, Snowflake, S3, and other backends.

### Hugging Face — We Got Claude to Fine-Tune an Open Source LLM

- Company: Hugging Face (huggingface.co)
- Announced: 2025-12-04
- Category: developer-tool-release
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://huggingface.co/blog/hf-skills-training
- Record: https://forck.live/items/1576-we-got-claude-to-fine-tune-an-open-source-llm
- Subject: Platform

Hugging Face released a new tool called Hugging Face Skills, including the hf-llm-trainer skill that allows coding agents like Claude Code to fine-tune open-source language models on Hugging Face GPUs.

### Hugging Face — DeepMath: A lightweight math reasoning Agent with smolagents

- Company: Hugging Face (huggingface.co)
- Announced: 2025-12-04
- Category: research-paper
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://huggingface.co/blog/intel-deepmath
- Record: https://forck.live/items/1575-deepmath-a-lightweight-math-reasoning-agent-with-smolagents
- Subject: Platform

Intel AI Software Group introduces DeepMath, a math reasoning agent based on Qwen3-4B Thinking fine-tuned with GRPO. The model emits Python snippets for intermediate steps, runs them in a sandbox, and uses the results in its reasoning. It reduces output lengths by up to 66% and improves accuracy on math datasets.

### Google Research — From Waveforms to Wisdom: The New Benchmark for Auditory Intelligence

- Company: Google Research (research.google)
- Announced: 2025-12-03T22:47:00+00:00
- Category: research-paper
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://research.google/blog/from-waveforms-to-wisdom-the-new-benchmark-for-auditory-intelligence/
- Record: https://forck.live/items/1386-from-waveforms-to-wisdom-the-new-benchmark-for-auditory-intelligence
- Subject: Research

Google Research introduces the Massive Sound Embedding Benchmark (MSEB), an open-source platform for evaluating machine sound intelligence across eight core capabilities, presented at NeurIPS 2025.

### Supabase — Introducing Seven New Email Templates for Supabase Auth

- Company: Supabase (supabase.com)
- Announced: 2025-12-03T17:00:00+00:00
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://supabase.com/blog/introducing-seven-new-email-templates-for-auth
- Record: https://forck.live/items/21318-introducing-seven-new-email-templates-for-supabase-auth
- Subject: Supabase

Supabase released seven new security notification email templates for Supabase Auth, covering password changes, email/phone number changes, identity linking/unlinking, and MFA method additions/removals. Each notification includes relevant context such as the old email or provider name. The templates can be enabled, disabled, and customized via the Dashboard, CLI, or Management API, and are also supported through the Auth email send hook.

### OpenAI — OpenAI to acquire Neptune

- Company: OpenAI (openai.com)
- Announced: 2025-12-03T10:00:00+00:00
- Category: partnership-acquisition
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://openai.com/index/openai-to-acquire-neptune
- Record: https://forck.live/items/358-openai-to-acquire-neptune
- Subject: GPT / ChatGPT / API

OpenAI announces acquisition of Neptune to improve model behavior visibility and researcher tools.

### OpenAI — How confessions can keep language models honest

- Company: OpenAI (openai.com)
- Announced: 2025-12-03T10:00:00+00:00
- Category: research-paper
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://openai.com/index/how-confessions-can-keep-language-models-honest
- Record: https://forck.live/items/357-how-confessions-can-keep-language-models-honest
- Subject: GPT / ChatGPT / API

OpenAI researchers are testing a method called 'confessions' that trains models to admit when they make mistakes or act undesirably, aiming to improve AI honesty, transparency, and trust.

### OpenAI — Announcing the initial People-First AI Fund grantees

- Company: OpenAI (openai.com)
- Announced: 2025-12-03T08:00:00+00:00
- Category: funding-company-news
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://openai.com/index/people-first-ai-fund-grantees
- Record: https://forck.live/items/359-announcing-the-initial-people-first-ai-fund-grantees
- Subject: GPT / ChatGPT / API

The OpenAI Foundation announced the first grantees of the People-First AI Fund, distributing $40.5 million in unrestricted grants to 208 nonprofits.

### Supabase — The new Supabase power for Kiro

- Company: Supabase (supabase.com)
- Announced: 2025-12-03
- Category: developer-tool-release
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://supabase.com/blog/supabase-power-for-kiro
- Record: https://forck.live/items/21319-the-new-supabase-power-for-kiro
- Subject: Supabase

Supabase announced new 'powers' for Amazon's Kiro IDE, bundling MCP tools and steering files to give the AI assistant access to database schemas, edge function best practices, RLS security guidelines, and Postgres SQL syntax. Two powers are released: one for hosted Supabase projects and one for local development via the Supabase CLI. The powers enable one-click schema management, edge function review, security and performance checks, and context-aware SQL help.

### Laminar — How we instrumented Claude Agent SDK using a tiny Rust proxy

- Company: Laminar (laminar.sh)
- Announced: 2025-12-03
- Category: developer-tool-release
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://laminar.sh/blog/2025-12-03-claude-agent-sdk-instrumentation
- Record: https://forck.live/items/16853-how-we-instrumented-claude-agent-sdk-using-a-tiny-rust-proxy
- Subject: flow / agent observability

Laminar released instrumentation packages for the Claude Agent SDK in Python and TypeScript. The packages use a tiny Rust proxy shipped inside the package to capture prompts, tool calls, subagent trajectories, and latency metrics from the Claude Code process, with no extra dependencies or infrastructure.

### LG AI Research — [CTAD 2025] MOIRA, a multimodal framework for more effective prediction of Alzheimer's disease with incomplete data

- Company: LG AI Research (lgresearch.ai)
- Announced: 2025-12-03
- Category: research-paper
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://www.lgresearch.ai/blog/view?seq=601
- Record: https://forck.live/items/4770-ctad-2025-moira-a-multimodal-framework-for-more-effective-prediction-of
- Subject: EXAONE

LG AI Research presented MOIRA, a multimodal framework for Alzheimer's disease prediction that handles incomplete data by transforming heterogeneous clinical data into latent representations and projecting them into a shared embedding space.

### Together AI — Introducing AutoJudge: Streamlined inference acceleration via automated dataset curation

- Company: Together AI (together.ai)
- Announced: 2025-12-03
- Category: research-paper
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://www.together.ai/blog/introducing-autojudge-streamlined-inference-acceleration-via-automated-dataset-curation
- Record: https://forck.live/items/2398-introducing-autojudge-streamlined-inference-acceleration-via-automated-dataset
- Subject: Inference platform

Together AI introduces AutoJudge, a research method that accelerates LLM inference through task-specific lossy speculative decoding, which automatically identifies which generated tokens affect downstream quality without manual annotation. It achieves 1.5–2x speedups over standard speculative decoding by accepting up to 40 draft tokens per verification cycle with slight accuracy drop, and will be presented at NeurIPS 2025.

### Together AI — How to run TorchForge reinforcement learning pipelines in the Together AI Native Cloud

- Company: Together AI (together.ai)
- Announced: 2025-12-03
- Category: capability-change
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://www.together.ai/blog/torchforge-reinforcement-learning-pipelines
- Record: https://forck.live/items/2397-how-to-run-torchforge-reinforcement-learning-pipelines-in-the-together-ai
- Subject: Inference platform

Together AI announces that TorchForge reinforcement learning pipelines now run on its Native Cloud platform, supporting distributed training on Instant Clusters with sandboxed environments via CodeSandbox and Code Interpreter. A demo for training a Qwen 1.5B model to play BlackJack using GRPO is released.

### Together AI — Together AI and Meta partner to bring PyTorch Reinforcement Learning to the AI Native Cloud

- Company: Together AI (together.ai)
- Announced: 2025-12-03
- Category: partnership-acquisition
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://www.together.ai/blog/together-ai-and-meta-partner-to-bring-pytorch-reinforcement-learning-to-the-ai-native-cloud
- Record: https://forck.live/items/2396-together-ai-and-meta-partner-to-bring-pytorch-reinforcement-learning-to-the-ai
- Subject: Inference platform

Together AI and the PyTorch team at Meta are partnering to build an open-source reinforcement learning framework for training agentic AI systems, integrated into the Together AI platform.

### Supabase — Introducing Supabase Pipelines

- Company: Supabase (supabase.com)
- Announced: 2025-12-02T18:00:00+00:00
- Category: availability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://supabase.com/blog/introducing-supabase-pipelines
- Record: https://forck.live/items/21320-introducing-supabase-pipelines
- Subject: Supabase

Supabase launched Pipelines, a managed change-data-capture service that replicates Postgres tables to analytical destinations in near real time using logical replication. The first supported destination is Google BigQuery, with ClickHouse, Snowflake, and DuckLake in early access. Pipelines is currently in public alpha and is powered by the open-source Supabase ETL engine.

### Supabase — Introducing Analytics Buckets

- Company: Supabase (supabase.com)
- Announced: 2025-12-02T17:00:00+00:00
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://supabase.com/blog/introducing-analytics-buckets
- Record: https://forck.live/items/21321-introducing-analytics-buckets
- Subject: Supabase

Supabase launched Analytics Buckets, a storage type built on Apache Iceberg and Amazon S3 that stores data in columnar Parquet format for analytical workloads. The feature offers cost savings of 30–90% on storage for large datasets, supports schema evolution and time travel, and can be queried from Postgres via Foreign Data Wrappers. Analytics Buckets are free during the Private Alpha, with standard egress charges for data moved out of region.

### Lovable — Lovable Now Supports Claude Opus 4.5

- Company: Lovable (lovable.dev)
- Announced: 2025-12-02T15:00:00+00:00
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://lovable.dev/blog/claude-opus-4-5-in-chat
- Record: https://forck.live/items/12116-lovable-now-supports-claude-opus-4-5
- Subject: Lovable / AI app builder

Lovable added Claude Opus 4.5 as a core model for chat mode and code generation, reporting 15% fewer mistakes when making edits, 5% higher project success rate, and noticeably better designs, at no additional cost to users.

### Harvey — How our In-House Legal Team Uses Harvey to Work Smarter

- Company: Harvey (harvey.ai)
- Announced: 2025-12-02T14:00:00+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://www.harvey.ai/blog/how-harveys-legal-team-uses-harvey-to-work-smarter
- Record: https://forck.live/items/17643-how-our-in-house-legal-team-uses-harvey-to-work-smarter
- Subject: Harvey

Our in-house lawyers are using Harvey’s latest capabilities to strengthen operational excellence and scale legal support. A lot has changed since we published our blog on How Harvey Uses Harvey , so it felt like a good time to revisit the topic.

### Lovable — Lovable outages on Friday November 28

- Company: Lovable (lovable.dev)
- Announced: 2025-12-02T14:00:00+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://lovable.dev/blog/outage-nov-28
- Record: https://forck.live/items/12117-lovable-outages-on-friday-november-28
- Subject: Lovable / AI app builder

Lovable suffered two availability incidents on November 28, both caused by third-party vendors. From 9:20-10:10 UTC, the site was slow due to a Posthog outage; from 12:00-12:45 UTC, the site was inaccessible due to a Netlify incident. Published user applications hosted by Lovable were not affected.

### Vercel — Investing in the Python ecosystem

- Company: Vercel (vercel.com)
- Announced: 2025-12-02T13:00:00+00:00
- Category: partnership-acquisition
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://vercel.com/blog/investing-in-the-python-ecosystem
- Record: https://forck.live/items/20782-investing-in-the-python-ecosystem
- Subject: v0 / AI SDK / AI Gateway

Vercel is acquiring the Gel Data team (Yury Selivanov and Elvis Pranskevichus) to improve Python support on its AI Cloud platform. The company is also becoming a Maintaining-level sponsor of the Python Software Foundation and will sponsor core maintainer Serhiy Storchaka for one year. Gel Data will wind down on January 31, 2026, but its open source components will remain available.

### Vercel — Billions of requests: Black Friday-Cyber Monday 2025

- Company: Vercel (vercel.com)
- Announced: 2025-12-02T13:00:00+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://vercel.com/blog/bfcm-2025
- Record: https://forck.live/items/20781-billions-of-requests-black-friday-cyber-monday-2025
- Subject: v0 / AI SDK / AI Gateway

Vercel reported handling 115.8 billion total requests during the Black Friday-Cyber Monday weekend, a 33.6% year-over-year increase. The platform processed 518,027 peak requests per second, 24 million AI Gateway requests, and 43.2 billion Fluid compute invocations without requiring configuration changes or manual intervention. Security systems blocked 415 million bot attempts and took 7.5 billion firewall actions.

### Gradium — Gradium: Solving voice

- Company: Gradium (gradium.ai)
- Announced: 2025-12-02
- Category: product-launch
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://gradium.ai/blog/gradium
- Record: https://forck.live/items/18333-gradium-solving-voice
- Subject: Gradium TTS / Phonon

Gradium launched its core engine for real-time, multilingual voice AI, offering streaming transcription and synthesis APIs. The company raised $70M in seed funding from FirstMark Capital, Eurazeo, and others. The platform is already in production use by studios, game developers, and healthcare innovators.

### LG AI Research — The honor of receiving Presidential Citation! Leading software innovation with EXAONE

- Company: LG AI Research (lgresearch.ai)
- Announced: 2025-12-02
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://www.lgresearch.ai/news/view?seq=603
- Record: https://forck.live/items/10997-the-honor-of-receiving-presidential-citation-leading-software-innovation-with
- Subject: EXAONE

LG AI Research announced that Jinsik Lee, Leader of the EXAONE Lab, received the Presidential Citation for Meritorious Service to the Software Industry. The award recognizes his contributions to advancing the software industry and LG AI Research's efforts to secure domestic AI technology independence and expand an open ecosystem. The announcement highlights EXAONE's achievements including being Korea's first open-weight large-scale AI model, reaching 8.6 million downloads on Hugging Face as of December 2025, and the EXAONE 3.5 32B model being named a Notable AI Model by Epoch AI and included in Stanford HAI's AI Index 2025.

### Luma — Luma AI Opens for International Business With Former WPP Executive Jason Day Leading New London Office

- Company: Luma (lumalabs.ai)
- Announced: 2025-12-02
- Category: availability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://lumalabs.ai/news/luma-ai-opens-for-international-business-with-former-wpp-executive-jason-day-leading-new-london-office
- Record: https://forck.live/items/7975-luma-ai-opens-for-international-business-with-former-wpp-executive-jason-day
- Subject: Ray / Dream Machine

Luma AI announced the opening of its first international office in London, UK, and appointed former WPP executive Jason Day as Head of EMEA. The company plans to add 200 jobs in London in 2026.

### Amp — Amp Frontier Corporation

- Company: Amp (ampcode.com)
- Announced: 2025-12-02
- Category: funding-company-news
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://ampcode.com/news/amp-frontier-corporation
- Record: https://forck.live/items/7781-amp-frontier-corporation
- Subject: Amp

Amp is spinning out of Sourcegraph to become an independent research lab called Amp Frontier Corporation. The company is profitable and aims to let software builders harness the full power of artificial intelligence.

### LG AI Research — The honor of receiving Presidential Citation! Leading software innovation with EXAONE

- Company: LG AI Research (lgresearch.ai)
- Announced: 2025-12-02
- Category: funding-company-news
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://www.lgresearch.ai/blog/view?seq=603
- Record: https://forck.live/items/4771-the-honor-of-receiving-presidential-citation-leading-software-innovation-with
- Subject: EXAONE

Jinsik Lee, leader of the EXAONE Lab at LG AI Research, received a Presidential Citation for contributions to the software industry, in part for developing South Korea's first large-scale AI model and releasing open-weight models. The post highlights achievements of the EXAONE model series, including download numbers, benchmark comparisons, and international recognition.

### Mistral AI — Introducing Mistral 3

- Company: Mistral AI (mistral.ai)
- Announced: 2025-12-02
- Category: new-model
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://mistral.ai/news/mistral-3
- Record: https://forck.live/items/3309-introducing-mistral-3
- Subject: Mistral / Le Chat

Mistral AI announces Mistral 3 in a research post.

### OpenAI — Inside Mirakl's agentic commerce vision

- Company: OpenAI (openai.com)
- Announced: 2025-12-01T22:00:00+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://openai.com/index/mirakl
- Record: https://forck.live/items/360-inside-mirakl-s-agentic-commerce-vision
- Subject: GPT / ChatGPT / API

Mirakl is using AI agents and ChatGPT Enterprise to achieve faster documentation and smarter customer support, and is building toward agent-native commerce with Mirakl Nexus.

### Vercel — AWS Databases coming to the Vercel Marketplace

- Company: Vercel (vercel.com)
- Announced: 2025-12-01T13:00:00+00:00
- Category: availability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://vercel.com/blog/aws-databases-coming-to-the-vercel-marketplace
- Record: https://forck.live/items/20783-aws-databases-coming-to-the-vercel-marketplace
- Subject: v0 / AI SDK / AI Gateway

Vercel announced that on December 15th, Aurora PostgreSQL, Amazon DynamoDB, and Aurora DSQL will become available as native integrations in the Vercel Marketplace, allowing developers to provision and connect these databases directly from the Vercel dashboard with automatic credential and environment variable management.

### OpenAI — Funding grants for new research into AI and mental health

- Company: OpenAI (openai.com)
- Announced: 2025-12-01T12:00:00+00:00
- Category: funding-company-news
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://openai.com/index/ai-mental-health-research-grants
- Record: https://forck.live/items/361-funding-grants-for-new-research-into-ai-and-mental-health
- Subject: GPT / ChatGPT / API

OpenAI is awarding up to $2 million in grants for research at the intersection of AI and mental health.

### OpenAI — OpenAI and NORAD team up to bring new magic to “NORAD Tracks Santa”

- Company: OpenAI (openai.com)
- Announced: 2025-12-01T06:00:00+00:00
- Category: partnership-acquisition
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://openai.com/index/norad-holiday-collaboration
- Record: https://forck.live/items/362-openai-and-norad-team-up-to-bring-new-magic-to-norad-tracks-santa
- Subject: GPT / ChatGPT / API

OpenAI and NORAD collaborate to add three ChatGPT holiday tools to NORAD Tracks Santa, enabling families to create festive elves, toy coloring pages, and custom Christmas stories.

### OpenAI — Accenture and OpenAI accelerate enterprise AI success

- Company: OpenAI (openai.com)
- Announced: 2025-12-01T05:00:00+00:00
- Category: partnership-acquisition
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://openai.com/index/accenture-partnership
- Record: https://forck.live/items/364-accenture-and-openai-accelerate-enterprise-ai-success
- Subject: GPT / ChatGPT / API

OpenAI and Accenture announce a collaboration to help enterprises integrate agentic AI into their core business operations.

### OpenAI — OpenAI takes an ownership stake in Thrive Holdings to accelerate enterprise AI adoption

- Company: OpenAI (openai.com)
- Announced: 2025-12-01T05:00:00+00:00
- Category: funding-company-news
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://openai.com/index/thrive-holdings
- Record: https://forck.live/items/363-openai-takes-an-ownership-stake-in-thrive-holdings-to-accelerate-enterprise-ai
- Subject: GPT / ChatGPT / API

OpenAI takes an ownership stake in Thrive Holdings to accelerate enterprise AI adoption, embedding frontier research and engineering into accounting and IT services.

### Supabase — Introducing Vector Buckets

- Company: Supabase (supabase.com)
- Announced: 2025-12-01
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://supabase.com/blog/vector-buckets
- Record: https://forck.live/items/21322-introducing-vector-buckets
- Subject: Supabase

Supabase introduced Vector Buckets, a new storage option that combines Amazon S3 durability with built-in similarity search for large-scale vector datasets. Vector Buckets support up to 50 million vectors per index and allow querying via Supabase SDKs or Postgres through a foreign data wrapper, complementing pgvector for latency-sensitive workloads.

### DeepSeek — DeepSeek-V3.2: Pushing the Frontier of Open Large Language Models

- Company: DeepSeek (deepseek.com)
- Announced: 2025-12-01
- Category: new-model
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://www.deepseek.com/en/news/deepseek-v3-2
- Record: https://forck.live/items/4320-deepseek-v3-2-pushing-the-frontier-of-open-large-language-models
- Subject: DeepSeek models

DeepSeek announces DeepSeek-V3.2 and DeepSeek-V3.2-Speciale, reasoning-first models for agents. V3.2 is the successor to V3.2-Exp and is available on App, Web, and API. V3.2-Speciale is API-only and focuses on maximum reasoning capabilities, achieving gold-level results in IMO, CMO, ICPC World Finals, and IOI 2025. The models introduce thinking in tool-use and a new agent training data synthesis method.

### Together AI — Together AI delivers fastest inference for the top open-source models

- Company: Together AI (together.ai)
- Announced: 2025-12-01
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://www.together.ai/blog/fastest-inference-for-the-top-open-source-models
- Record: https://forck.live/items/2399-together-ai-delivers-fastest-inference-for-the-top-open-source-models
- Subject: Inference platform

Together AI announces that it now delivers up to 2x faster serverless inference for leading open-source LLMs, ranking #1 in output speed benchmarks. The performance improvements come from coordinated enhancements across next-gen GPU hardware, optimized kernels, near-lossless quantization, and production-grade speculative decoding with custom-trained draft models.

### Hugging Face — Transformers v5: Simple model definitions powering the AI ecosystem

- Company: Hugging Face (huggingface.co)
- Announced: 2025-12-01
- Category: developer-tool-release
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://huggingface.co/blog/transformers-v5
- Record: https://forck.live/items/1577-transformers-v5-simple-model-definitions-powering-the-ai-ecosystem
- Subject: Platform

Hugging Face releases Transformers v5, focusing on simplicity, training, inference, and production. Key changes include a modular design, removal of Flax/TensorFlow support in favor of PyTorch, simplification of tokenization (removing Fast/Slow tokenizers), and new tooling for model conversion.