# forck.live — 2–8 February 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-W06, 2–8 February 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-W06
- The current issue, and the catalogue of the other weeks: https://forck.live/briefing.md
- Catalogue: https://forck.live/llms.txt

## Announcements

### Sarvam AI — Introducing Bulbul V3: Natural. Expressive. Production-ready.

- Company: Sarvam AI (sarvam.ai)
- Announced: 2026-02-07T12:00:00+00:00
- Category: new-model
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://www.sarvam.ai/blogs/bulbul-v3/
- Record: https://forck.live/items/18396-introducing-bulbul-v3-natural-expressive-production-ready
- Subject: Sarvam / Bulbul / Saaras

Sarvam AI released Bulbul V3, a text-to-speech model for Indian languages. An independent third-party blind A/B human listening study across 11 languages found it achieved high listener preference at 48 kHz and was the most-preferred model in 8 kHz telephony, with low character error rates on code-mixing and numerics, and minimal word skips and mispronunciations. The model includes a low-latency streaming output mode for conversational applications.

### OpenClaw — OpenClaw Partners with VirusTotal for Skill Security

- Company: OpenClaw (openclaw.ai)
- Announced: 2026-02-07
- Category: safety-policy-update
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://openclaw.ai/blog/virustotal-partnership/
- Record: https://forck.live/items/17770-openclaw-partners-with-virustotal-for-skill-security
- Subject: OpenClaw

OpenClaw partnered with VirusTotal to bring security scanning to ClawHub, its skill marketplace. All skills published to ClawHub are now scanned using VirusTotal's threat intelligence and Code Insight capability, with automatic approval for benign skills, warnings for suspicious ones, and blocking for malicious ones. Daily re-scans are performed to detect previously clean skills that become malicious.

### Vercel — Introducing Geist Pixel

- Company: Vercel (vercel.com)
- Announced: 2026-02-06T13:00:00+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://vercel.com/blog/introducing-geist-pixel
- Record: https://forck.live/items/20752-introducing-geist-pixel
- Subject: v0 / AI SDK / AI Gateway

Vercel introduced Geist Pixel, a bitmap-inspired typeface that extends the Geist font family. It includes five variants (Square, Grid, Circle, Triangle, Line) and is designed for functional use in modern web products, sharing the same system principles as Geist Sans and Geist Mono. The font is available via the `geist` npm package.

### OpenAI — Making AI work for everyone, everywhere: our approach to localization

- Company: OpenAI (openai.com)
- Announced: 2026-02-06T10:00:00+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://openai.com/index/our-approach-to-localization
- Record: https://forck.live/items/277-making-ai-work-for-everyone-everywhere-our-approach-to-localization
- Subject: GPT / ChatGPT / API

OpenAI discusses its approach to localizing AI models for different languages, laws, and cultures while maintaining safety.

### Supabase — X / Twitter OAuth 2.0 is now available for Supabase Auth

- Company: Supabase (supabase.com)
- Announced: 2026-02-06
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://supabase.com/blog/x-twitter-oauth-2-provider
- Record: https://forck.live/items/21303-x-twitter-oauth-2-0-is-now-available-for-supabase-auth
- Subject: Supabase

Supabase Auth now supports X / Twitter OAuth 2.0, replacing the legacy OAuth 1.0a flow. The new provider includes PKCE support and requires creating an OAuth 2.0 app in the X Developer Portal. The legacy provider remains available while X continues to support OAuth 1.0a.

### Cua — Announcing the First Multi-Player Computer-Use — Live from ClawCon

- Company: Cua (cua.ai)
- Announced: 2026-02-06
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://cua.ai/blog/clawcon-multiplayer
- Record: https://forck.live/items/16823-announcing-the-first-multi-player-computer-use-live-from-clawcon
- Subject: Cua Driver / Lume / Cua Bench / Cua Fleets

Cua announced CuaBot, a CLI that gives coding agents a sandboxed Linux desktop with computer-use capabilities, at its ClawCon community event. The tool streams an agent's application windows individually to the host desktop through Xpra, so the human keeps full use of their machine while the agent works. CuaBot supports any coding agent, including Claude Code, Gemini CLI, Codex, OpenClaw, and Aider.

### Ideogram — Skechers enhances design innovation with Ideogram

- Company: Ideogram (ideogram.ai)
- Announced: 2026-02-06
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://ideogram.ai/blog/skechers-case-study
- Record: https://forck.live/items/15465-skechers-enhances-design-innovation-with-ideogram
- Subject: Ideogram image models

Ideogram announced that Skechers is adopting its Custom Branded Models to accelerate design workflows, maintain brand consistency, and reduce sample costs and production timelines. The partnership enables Skechers' creative teams to use Material Swap and Product Swap features for rapid ideation and marketing asset iteration.

### Cartesia — Cartesia Is Hiring a SWE to Raise an Army of Claudes

- Company: Cartesia (cartesia.ai)
- Announced: 2026-02-06
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://www.cartesia.ai/blog/claude-army
- Record: https://forck.live/items/8013-cartesia-is-hiring-a-swe-to-raise-an-army-of-claudes
- Subject: Sonic / Ink

Cartesia describes its adoption of multiple coding agents and identifies the remaining bottleneck as agents' inability to close feedback loops by checking production metrics, dashboards, or experiment results. The company is hiring an AI & Developer Acceleration engineer to build infrastructure that lets agents iterate end-to-end.

### Baidu — ERNIE 5.0: A 2.4 Trillion-Parameter Unified Multimodal Foundation Model

- Company: Baidu (baidu.com)
- Announced: 2026-02-06
- Category: new-model
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://ernie.baidu.com/blog/posts/ernie5.0/
- Record: https://forck.live/items/4643-ernie-5-0-a-2-4-trillion-parameter-unified-multimodal-foundation-model
- Subject: ERNIE

Baidu announced ERNIE 5.0, a 2.4 trillion parameter multimodal foundation model with a unified autoregressive framework integrating text, images, video, and audio, achieving state-of-the-art results across modalities.

### Together AI — What do LLMs think when you don't tell them what to think about?

- Company: Together AI (together.ai)
- Announced: 2026-02-06
- Category: research-paper
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://www.together.ai/blog/what-llms-think
- Record: https://forck.live/items/2381-what-do-llms-think-when-you-don-t-tell-them-what-to-think-about
- Subject: Inference platform

A study of near-unconstrained LLM generation reveals that different model families exhibit distinct topical preferences: GPT-OSS favors programming and math, Llama leans literary, DeepSeek often produces religious content, and Qwen tends toward multiple-choice questions.

### Lovable — Claude Opus 4.6 now in Lovable

- Company: Lovable (lovable.dev)
- Announced: 2026-02-05T18:00:00+00:00
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://lovable.dev/blog/opus-4-6-now-in-lovable
- Record: https://forck.live/items/12109-claude-opus-4-6-now-in-lovable
- Subject: Lovable / AI app builder

Lovable has integrated Claude Opus 4.6 as a core model, resulting in a 21% improvement on its app-building benchmark and doubling the runtime on complex tasks. The upgrade is available at no additional cost to Lovable users and is particularly suited for design-heavy applications such as creative landing pages and polished marketing sites.

### Harvey — Opus 4.6, Now Live in Harvey

- Company: Harvey (harvey.ai)
- Announced: 2026-02-05T17:55:00+00:00
- Category: availability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://www.harvey.ai/blog/opus-4-6-now-live-in-harvey
- Record: https://forck.live/items/17617-opus-4-6-now-live-in-harvey
- Subject: Harvey

Harvey made Anthropic's Claude Opus 4.6 available in its platform. The model scored 90.2% on Harvey's BigLaw Bench evaluation, the highest for the Claude family, with 40% of tasks receiving perfect scores. Harvey will roll out Opus 4.6 to eligible US and EU clients in the coming days, with Australian availability to be announced later.

### Hugging Face — Introducing SyGra Studio

- Company: Hugging Face (huggingface.co)
- Announced: 2026-02-05T16:52:28+00:00
- Category: developer-tool-release
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://huggingface.co/blog/ServiceNow-AI/sygra-studio
- Record: https://forck.live/items/1550-introducing-sygra-studio
- Subject: Platform

ServiceNow-AI announces SyGra Studio, an interactive visual environment for synthetic data generation workflows that lets users compose flows on a canvas, preview datasets, tune prompts with inline variable hints, and monitor execution in real time.

### Harvey — The Brief: February 2026

- Company: Harvey (harvey.ai)
- Announced: 2026-02-05T16:00:00+00:00
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://www.harvey.ai/blog/the-brief-february-2026
- Record: https://forck.live/items/17618-the-brief-february-2026
- Subject: Harvey

Harvey introduced a homepage refresh with suggested prompts, 70+ new Library prompts, an improved Word interface, and simplified Vault file uploads. It added independent review tables, support for new file types (JSON, Python, XML), review tables as knowledge sources, and a NetDocuments integration. Harvey also launched Agentic Word for deeper document editing, workflow optimization suggestions, and review table nudges. The update includes 200 new legal research sources across multiple jurisdictions and centralized knowledge source admin toggles.

### Vercel — The Vercel AI Accelerator is back with $6m in credits

- Company: Vercel (vercel.com)
- Announced: 2026-02-05T13:00:00+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://vercel.com/blog/the-vercel-ai-accelerator-is-back-with-6-million-in-credits
- Record: https://forck.live/items/20753-the-vercel-ai-accelerator-is-back-with-6m-in-credits
- Subject: v0 / AI SDK / AI Gateway

Vercel announced the return of its AI Accelerator program, offering over $6 million in credits from Vercel, AWS, and other AI platforms to 40 selected early-stage teams. The six-week program runs from March 2 to April 16, 2025, and includes mentorship, weekly sessions, and a demo day in San Francisco. Applications are open until February 16, 2025.

### Sarvam AI — Sarvam Vision

- Company: Sarvam AI (sarvam.ai)
- Announced: 2026-02-05T12:00:00+00:00
- Category: new-model
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://www.sarvam.ai/blogs/sarvam-vision/
- Record: https://forck.live/items/18397-sarvam-vision
- Subject: Sarvam / Bulbul / Saaras

Sarvam AI introduced Sarvam Vision, a 3B-parameter state-space vision-language model capable of image captioning, scene text recognition, chart interpretation, and complex table parsing. The model is designed for high-accuracy document intelligence across English and all 22 scheduled Indian languages. It was trained via continual pretraining on the base Sarvam sovereign 3B model, followed by supervised fine-tuning and reinforcement learning using verifiable rewards.

### OpenAI — GPT-5 lowers the cost of cell-free protein synthesis

- Company: OpenAI (openai.com)
- Announced: 2026-02-05T11:00:00+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://openai.com/index/gpt-5-lowers-protein-synthesis-cost
- Record: https://forck.live/items/278-gpt-5-lowers-the-cost-of-cell-free-protein-synthesis
- Subject: GPT / ChatGPT / API

OpenAI's GPT-5, combined with Ginkgo Bioworks' cloud automation, reduced cell-free protein synthesis costs by 40% through closed-loop experimentation.

### OpenAI — Introducing Trusted Access for Cyber

- Company: OpenAI (openai.com)
- Announced: 2026-02-05T10:00:00+00:00
- Category: safety-policy-update
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://openai.com/index/trusted-access-for-cyber
- Record: https://forck.live/items/279-introducing-trusted-access-for-cyber
- Subject: GPT / ChatGPT / API

OpenAI introduced Trusted Access for Cyber, a trust-based framework that expands access to frontier cyber capabilities while strengthening safeguards against misuse.

### Google Research — How AI tools can redefine universal design to increase accessibility

- Company: Google Research (research.google)
- Announced: 2026-02-05T08:28:00+00:00
- Category: research-paper
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://research.google/blog/how-ai-agents-can-redefine-universal-design-to-increase-accessibility/
- Record: https://forck.live/items/1369-how-ai-tools-can-redefine-universal-design-to-increase-accessibility
- Subject: Research

Google Research introduces Natively Adaptive Interfaces (NAI), a framework for creating more accessible applications through multimodal AI tools that adapt to user needs, co-developed with the accessibility community.

### OpenAI — Introducing OpenAI Frontier

- Company: OpenAI (openai.com)
- Announced: 2026-02-05T06:00:00+00:00
- Category: product-launch
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://openai.com/index/introducing-openai-frontier
- Record: https://forck.live/items/280-introducing-openai-frontier
- Subject: GPT / ChatGPT / API

OpenAI announced OpenAI Frontier, an enterprise platform for building, deploying, and managing AI agents with features like shared context, onboarding, permissions, and governance.

### Gradium — Acolad and Gradium Partner to Advance Enterprise-Ready AI Interpreting

- Company: Gradium (gradium.ai)
- Announced: 2026-02-05
- Category: partnership-acquisition
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://gradium.ai/blog/acolad-gradium-partnership
- Record: https://forck.live/items/18329-acolad-and-gradium-partner-to-advance-enterprise-ready-ai-interpreting
- Subject: Gradium TTS / Phonon

Acolad and Gradium announced a strategic partnership to develop AI-powered real-time interpreting solutions for enterprise and public-sector use, combining Acolad's language workflows with Gradium's voice AI models including real-time streaming Text-To-Speech and Speech-To-Text technologies.

### OpenAI — GPT-5.3-Codex System Card

- Company: OpenAI (openai.com)
- Announced: 2026-02-05
- Category: new-model
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://openai.com/index/gpt-5-3-codex-system-card
- Record: https://forck.live/items/282-gpt-5-3-codex-system-card
- Subject: GPT / ChatGPT / API

OpenAI announces GPT-5.3-Codex, described as the most capable agentic coding model to date, combining the coding performance of GPT-5.2-Codex with the reasoning and knowledge capabilities of GPT-5.2.

### OpenAI — Introducing GPT-5.3-Codex

- Company: OpenAI (openai.com)
- Announced: 2026-02-05
- Category: new-model
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://openai.com/index/introducing-gpt-5-3-codex
- Record: https://forck.live/items/281-introducing-gpt-5-3-codex
- Subject: GPT / ChatGPT / API

OpenAI introduced GPT-5.3-Codex, a Codex-native agent that combines frontier coding performance with general reasoning for long-horizon, real-world technical work.

### Google Research — ​Sequential Attention: Making AI models leaner and faster without sacrificing accuracy

- Company: Google Research (research.google)
- Announced: 2026-02-04T15:14:00+00:00
- Category: research-paper
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://research.google/blog/sequential-attention-making-ai-models-leaner-and-faster-without-sacrificing-accuracy/
- Record: https://forck.live/items/1370-sequential-attention-making-ai-models-leaner-and-faster-without-sacrificing
- Subject: Research

Google Research introduces Sequential Attention, a subset selection algorithm for making large-scale ML models more efficient by greedily selecting features, layers or blocks during training without sacrificing accuracy.

### OpenAI — Unlocking the Codex harness: how we built the App Server

- Company: OpenAI (openai.com)
- Announced: 2026-02-04T13:00:00+00:00
- Category: developer-tool-release
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://openai.com/index/unlocking-the-codex-harness
- Record: https://forck.live/items/283-unlocking-the-codex-harness-how-we-built-the-app-server
- Subject: GPT / ChatGPT / API

OpenAI describes how to embed the Codex agent using the Codex App Server, a bidirectional JSON-RPC API for streaming progress, tool use, approvals, and diffs.

### ElevenLabs — ElevenLabs raises $500M Series D at $11B valuation

- Company: ElevenLabs (elevenlabs.io)
- Announced: 2026-02-04T12:00:00+00:00
- Category: funding-company-news
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://elevenlabs.io/blog/series-d
- Record: https://forck.live/items/3364-elevenlabs-raises-500m-series-d-at-11b-valuation
- Subject: Voice AI

ElevenLabs announced a $500 million Series D funding round at an $11 billion valuation.

### LG AI Research — AX Innovation: LG Graduate School of AI's Vision for the Future

- Company: LG AI Research (lgresearch.ai)
- Announced: 2026-02-04
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://www.lgresearch.ai/news/view?seq=623
- Record: https://forck.live/items/10996-ax-innovation-lg-graduate-school-of-ai-s-vision-for-the-future
- Subject: EXAONE

LG AI Research held the 2025 graduation ceremony for the LG Graduate School of AI's doctoral program, awarding degrees to Yunho Shin of LG Uplus and Youngtak Na of LG Electronics. The program also received official accreditation from South Korea's Ministry of Education for its doctoral program, the first such accreditation in the country.

### Amp — Liberating Code Review

- Company: Amp (ampcode.com)
- Announced: 2026-02-04
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://ampcode.com/news/liberating-code-review
- Record: https://forck.live/items/7758-liberating-code-review
- Subject: Amp

Amp decoupled its code review agent from any user interface, making it a composable and extensible subroutine invocable from the command line or via natural language requests in threads. It also introduced customizable Checks defined in .agents/checks/ directories, which allow teams to define review criteria scoped to specific parts of the codebase.

### LG AI Research — AX Innovation: LG Graduate School of AI's Vision for the Future

- Company: LG AI Research (lgresearch.ai)
- Announced: 2026-02-04
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://www.lgresearch.ai/blog/view?seq=623
- Record: https://forck.live/items/4761-ax-innovation-lg-graduate-school-of-ai-s-vision-for-the-future
- Subject: EXAONE

LG AI Research announced the graduation of two PhD students from its LG Graduate School of AI, highlighting their research on video action recognition and monocular depth estimation, and the school's accreditation by South Korea's Ministry of Education.

### Mistral AI — Voxtral transcribes at the speed of sound.

- Company: Mistral AI (mistral.ai)
- Announced: 2026-02-04
- Category: research-paper
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://mistral.ai/news/voxtral-transcribe-2
- Record: https://forck.live/items/3304-voxtral-transcribes-at-the-speed-of-sound
- Subject: Mistral / Le Chat

Mistral AI announces Voxtral, a model capable of transcribing at the speed of sound.

### Together AI — Rime Arcana V3 Turbo and Rime Arcana V3 now available on Together AI

- Company: Together AI (together.ai)
- Announced: 2026-02-04
- Category: new-model
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://www.together.ai/blog/rime-arcana-v3-turbo-and-rime-arcana-v3-now-available-on-together-ai
- Record: https://forck.live/items/2382-rime-arcana-v3-turbo-and-rime-arcana-v3-now-available-on-together-ai
- Subject: Inference platform

Together AI announces the availability of two new Rime models: Rime Arcana V3 Turbo (English–Spanish, performance) and Rime Arcana V3 (11-language switching), both supporting native code-switching with low latency on Together AI dedicated endpoints.

### Hugging Face — Community Evals: Because we're done trusting black-box leaderboards over the community

- Company: Hugging Face (huggingface.co)
- Announced: 2026-02-04
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://huggingface.co/blog/community-evals
- Record: https://forck.live/items/1551-community-evals-because-we-re-done-trusting-black-box-leaderboards-over-the
- Subject: Platform

Hugging Face announces Community Evals, a new feature on the Hugging Face Hub that allows benchmark datasets to host leaderboards, models to store their own eval scores, and the community to submit evaluation results via pull requests, with verified badges for reproducible results.

### Google Research — Collaborating on a nationwide randomized study of AI in real-world virtual care

- Company: Google Research (research.google)
- Announced: 2026-02-03T18:15:01+00:00
- Category: research-paper
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://research.google/blog/collaborating-on-a-nationwide-randomized-study-of-ai-in-real-world-virtual-care/
- Record: https://forck.live/items/1371-collaborating-on-a-nationwide-randomized-study-of-ai-in-real-world-virtual-care
- Subject: Research

Google Research announces a partnership with Included Health to launch a nationwide randomized study evaluating conversational AI in real-world virtual care, pending IRB approval. The study builds on prior research including AMIE and Personal Health Agent.

### H Company — H Company's new Holo2 model takes the lead in UI Localization

- Company: H Company (hcompany.ai)
- Announced: 2026-02-03T17:40:14+00:00
- Category: new-model
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://huggingface.co/blog/Hcompany/introducing-holo2-235b-a22b
- Record: https://forck.live/items/1552-h-company-s-new-holo2-model-takes-the-lead-in-ui-localization

H Company announces Holo2-235B-A22B Preview, a new UI localization model achieving 78.5% on Screenspot-Pro and 79.0% on OSWorld G, available on Hugging Face as a research release.

### Supabase — BKND joins Supabase

- Company: Supabase (supabase.com)
- Announced: 2026-02-03T17:00:00+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://supabase.com/blog/bknd-joins-supabase
- Record: https://forck.live/items/21304-bknd-joins-supabase
- Subject: Supabase

Supabase has hired Dennis Senn, creator of BKND, to work on a Supabase Lite offering for agentic workloads. The company states that the focus is not to immediately integrate BKND into Supabase, and that the offering is still a work-in-progress. BKND will remain open source.

### Hugging Face — The Future of the Global Open-Source AI Ecosystem: From DeepSeek to AI+

- Company: Hugging Face (huggingface.co)
- Announced: 2026-02-03T15:03:19+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://huggingface.co/blog/huggingface/one-year-since-the-deepseek-moment-blog-3
- Record: https://forck.live/items/1553-the-future-of-the-global-open-source-ai-ecosystem-from-deepseek-to-ai
- Subject: Platform

This blog post analyzes the development of the open-source AI ecosystem in China since the DeepSeek Moment, examining the strategies of major Chinese AI organizations such as Alibaba, Tencent, ByteDance, and Baidu, as well as startups like Moonshot, Z.ai, and MiniMax. It highlights the organic growth of open-source collaboration and the increasing influence of Chinese models and papers.

### Harvey — Introducing Harvey’s Responsible Business Program

- Company: Harvey (harvey.ai)
- Announced: 2026-02-03T14:00:00+00:00
- Category: safety-policy-update
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://www.harvey.ai/blog/introducing-harveys-responsible-business-program
- Record: https://forck.live/items/17619-introducing-harvey-s-responsible-business-program
- Subject: Harvey

Harvey announced the launch of its Responsible Business Program, which includes hiring a full-time Head of Responsible Business, continuing the Harvey Law Schools Program, and committing to security, trust, privacy governance, and sustainability efforts.

### Vercel — Introducing the new v0

- Company: Vercel (vercel.com)
- Announced: 2026-02-03T13:00:00+00:00
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://vercel.com/blog/introducing-the-new-v0
- Record: https://forck.live/items/20756-introducing-the-new-v0
- Subject: v0 / AI SDK / AI Gateway

Vercel rebuilt v0 to support production app development with enterprise-grade security, Git integration, and direct database connections. The new version can import GitHub repos, create branches and PRs, and deploy on merge, enabling non-engineers to ship production code through proper workflows.

### Vercel — The Vercel OSS Bug Bounty program is now available

- Company: Vercel (vercel.com)
- Announced: 2026-02-03T13:00:00+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://vercel.com/blog/the-vercel-oss-bug-bounty-program-is-now-available
- Record: https://forck.live/items/20755-the-vercel-oss-bug-bounty-program-is-now-available
- Subject: v0 / AI SDK / AI Gateway

Vercel opened its Open Source Software bug bounty program to the public on HackerOne, inviting security researchers to find vulnerabilities in its open source projects. The program builds on a private bounty run since August 2025 and a separate WAF-focused bounty that paid out over $1M.

### Vercel — Making agent-friendly pages with content negotiation

- Company: Vercel (vercel.com)
- Announced: 2026-02-03T13:00:00+00:00
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://vercel.com/blog/making-agent-friendly-pages-with-content-negotiation
- Record: https://forck.live/items/20754-making-agent-friendly-pages-with-content-negotiation
- Subject: v0 / AI SDK / AI Gateway

Vercel updated its blog and changelog pages to support content negotiation via the HTTP Accept header, allowing agents to request clean markdown instead of full HTML. The markdown version of a page is 3KB versus 500KB for HTML, a 99.37% reduction in payload size. Vercel also added markdown sitemaps for agent discovery.

### Sarvam AI — Sarvam Audio: Speech Recognition beyond Transcription

- Company: Sarvam AI (sarvam.ai)
- Announced: 2026-02-03T12:00:00+00:00
- Category: not stated
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://www.sarvam.ai/blogs/sarvam-audio/
- Record: https://forck.live/items/18398-sarvam-audio-speech-recognition-beyond-transcription
- Subject: Sarvam / Bulbul / Saaras

India is a voice-first country. The rigid structure of a keyboard struggles to capture the fluidity of Indian languages, and for most people, speaking is simply more natural than typing.

### Hugging Face — Training Design for Text-to-Image Models: Lessons from Ablations

- Company: Hugging Face (huggingface.co)
- Announced: 2026-02-03T11:25:53+00:00
- Category: research-paper
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://huggingface.co/blog/Photoroom/prx-part2
- Record: https://forck.live/items/1554-training-design-for-text-to-image-models-lessons-from-ablations
- Subject: Platform

This is the second part of a series on training efficient text-to-image models from scratch. It documents training techniques that improved convergence and representation learning for the PRX model, including representation alignment, training objectives, token routing, and data strategies.

### Tencent — Learning from context is harder than we thought

- Company: Tencent (tencent.com)
- Announced: 2026-02-03T03:54:58+00:00
- Category: research-paper
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://hunyuan.tencent.com/research/100025
- Record: https://forck.live/items/4582-learning-from-context-is-harder-than-we-thought
- Subject: Hunyuan

Tencent Hunyuan introduces CL-bench, a benchmark to evaluate whether language models can learn new knowledge from context rather than relying solely on parametric memory, covering four types of real-world context learning scenarios with a contamination-free design.

### Supabase — Supabase is now an official Claude connector

- Company: Supabase (supabase.com)
- Announced: 2026-02-03
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://supabase.com/blog/supabase-is-now-an-official-claude-connector
- Record: https://forck.live/items/21305-supabase-is-now-an-official-claude-connector
- Subject: Supabase

Supabase's MCP server has been listed as an official connector for Claude, enabling users to manage database infrastructure, execute SQL queries, modify schemas, deploy edge functions, and troubleshoot projects through natural language prompts. The connector includes 32 tools covering database management, project operations, branching, storage, edge functions, and documentation search. It works on all Supabase plans and all paid Claude plans (Pro, Team, Enterprise, Max).

### Moonshot AI — WorldVQA

- Company: Moonshot AI (moonshot.ai)
- Announced: 2026-02-03
- Category: research-paper
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://www.kimi.ai/blog/worldvqa
- Record: https://forck.live/items/4625-worldvqa
- Subject: Kimi

Moonshot AI releases WorldVQA, a benchmark of 3,500 image-question pairs to evaluate factual visual world knowledge in multimodal LLMs, with a focus on head vs. tail distribution. Experiments show frontier models often fall below 50% accuracy, and the dataset and evaluation scripts are open-sourced.

### Z.ai — GLM-OCR

- Company: Z.ai (z.ai)
- Announced: 2026-02-03
- Category: new-model
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://docs.z.ai/release-notes/new-released#2026-02-03
- Record: https://forck.live/items/4452-glm-ocr
- Subject: GLM

Z.ai launched GLM-OCR, a compact and high-performance optical character recognition model using a self-developed CogViT and GLM-0.5B encoder-decoder architecture with CLIP pre-training for robust visual semantic understanding.

### Together AI — Together AI welcomes Alon Gavrielov as VP of Infrastructure Strategy

- Company: Together AI (together.ai)
- Announced: 2026-02-03
- Category: funding-company-news
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://www.together.ai/blog/alon-gavrielov-as-vp-of-infrastructure-strategy
- Record: https://forck.live/items/2383-together-ai-welcomes-alon-gavrielov-as-vp-of-infrastructure-strategy
- Subject: Inference platform

Together AI announces the appointment of Alon Gavrielov as VP of Infrastructure Strategy, who will lead the expansion of AI factories to support growing AI-native applications.

### OpenAI — The Sora feed philosophy

- Company: OpenAI (openai.com)
- Announced: 2026-02-03
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://openai.com/index/sora-feed-philosophy
- Record: https://forck.live/items/284-the-sora-feed-philosophy
- Subject: GPT / ChatGPT / API

OpenAI describes the philosophy behind the Sora feed, emphasizing creativity, connection, safety, personalized recommendations, parental controls, and guardrails.

### Harvey — Using Agents to Scale Harvey’s Knowledge Sources

- Company: Harvey (harvey.ai)
- Announced: 2026-02-02T20:00:00+00:00
- Category: capability-change
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://www.harvey.ai/blog/using-agents-to-scale-harveys-knowledge-sources
- Record: https://forck.live/items/17620-using-agents-to-scale-harvey-s-knowledge-sources
- Subject: Harvey

Harvey built an automated pipeline called The Data Factory that discovers legal sources and turns them into tools for its agents. Since August 2025, the pipeline has scaled Harvey's knowledge source coverage from six jurisdictions to over 60 and from 20 unique legal data sources to more than 400. The system uses a declarative configuration layer and a four-step evaluation pipeline to validate agent performance before launch.

### Alibaba — Qwen3-Coder-Next: Pushing Small Hybrid Models on Agentic Coding

- Company: Alibaba (alibaba.com)
- Announced: 2026-02-02T20:00:00+00:00
- Category: new-model
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://qwen.ai/blog?id=qwen3-coder-next
- Record: https://forck.live/items/4599-qwen3-coder-next-pushing-small-hybrid-models-on-agentic-coding
- Subject: Qwen

Qwen3-Coder-Next is an open-weight language model designed for coding agents and local development, built on Qwen3-Next-80B-A3B-Base with hybrid attention and MoE. It achieves over 70% on SWE-Bench Verified and comparable performance to much larger models on SWE-Bench-Pro with only 3B active parameters.

### ElevenLabs — Eleven v3 is Now Generally Available

- Company: ElevenLabs (elevenlabs.io)
- Announced: 2026-02-02T12:00:00+00:00
- Category: availability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://elevenlabs.io/blog/eleven-v3-is-now-generally-available
- Record: https://forck.live/items/3369-eleven-v3-is-now-generally-available
- Subject: Voice AI

Eleven v3 has been released as generally available.

### OpenAI — Snowflake and OpenAI partner to bring frontier intelligence to enterprise data

- Company: OpenAI (openai.com)
- Announced: 2026-02-02T06:00:00+00:00
- Category: partnership-acquisition
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://openai.com/index/snowflake-partnership
- Record: https://forck.live/items/285-snowflake-and-openai-partner-to-bring-frontier-intelligence-to-enterprise-data
- Subject: GPT / ChatGPT / API

OpenAI and Snowflake have entered a $200M partnership to integrate frontier AI capabilities into Snowflake's enterprise data platform, enabling AI agents and insights within Snowflake.

### Aleph Alpha — Aleph Alpha Appoints Ilhan Scheer as Co-CEO

- Company: Aleph Alpha (aleph-alpha.com)
- Announced: 2026-02-02
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://aleph-alpha.com/en/news/ilhan-scheer-appointed-co-ceo
- Record: https://forck.live/items/16858-aleph-alpha-appoints-ilhan-scheer-as-co-ceo
- Subject: Kolibri / PhariaAI

Aleph Alpha appointed Ilhan Scheer as Co-CEO, effective after the company's strategic and operational realignment in late 2025. Scheer, previously Chief Growth Officer since August 2025, now shares leadership with Co-CEO Reto Spörri, overseeing commercial execution, product direction, and technical strategy. The company said it will focus on building specialized language models for regulated and system-relevant domains.

### Luma — Luma AI Offers $1 Million For Cannes Lions Gold Winner

- Company: Luma (lumalabs.ai)
- Announced: 2026-02-02
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://lumalabs.ai/news/luma-dream-brief
- Record: https://forck.live/items/7972-luma-ai-offers-1-million-for-cannes-lions-gold-winner
- Subject: Ray / Dream Machine

Luma AI announced the Luma Dream Brief, a global creative competition that offers a $1 million grand prize to creative work using Luma AI that wins a 2026 Cannes Lions Gold Lion. The competition invites advertising creatives to submit commercials created with Luma AI, with selected finalists receiving paid media support.

### xAI — xAI joins SpaceX

- Company: xAI (x.ai)
- Announced: 2026-02-02
- Category: funding-company-news
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://x.ai/news/xai-joins-spacex
- Record: https://forck.live/items/3474-xai-joins-spacex
- Subject: Grok

SpaceX has acquired xAI, as announced by SpaceX on February 2, 2026.

### Together AI — Together Evaluations now supports comparing top commercial APIs vs. open source models

- Company: Together AI (together.ai)
- Announced: 2026-02-02
- Category: capability-change
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://www.together.ai/blog/together-evaluations-v2
- Record: https://forck.live/items/2385-together-evaluations-now-supports-comparing-top-commercial-apis-vs-open-source
- Subject: Inference platform

Together Evaluations now supports proprietary models from OpenAI, Anthropic, and Google as both judges and targets, allows evaluation of Together fine-tuned models, and provides new recipes demonstrating fine-tuned open judges that outperform GPT-5.2 at lower cost and higher speed, plus prompt optimization with GEPA.

### Together AI — Fine-tuning open LLM judges to outperform GPT-5.2

- Company: Together AI (together.ai)
- Announced: 2026-02-02
- Category: research-paper
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://www.together.ai/blog/fine-tuning-open-llm-judges-to-outperform-gpt-5-2
- Record: https://forck.live/items/2384-fine-tuning-open-llm-judges-to-outperform-gpt-5-2
- Subject: Inference platform

Open-source LLM judges fine-tuned with DPO can outperform GPT-5.2 at evaluating model outputs. Together AI trained GPT-OSS 120B on 5,400 preference pairs to beat GPT-5.2's accuracy, delivering superior performance at 15x lower cost and 14x faster speeds.

### OpenAI — Introducing the Codex app

- Company: OpenAI (openai.com)
- Announced: 2026-02-02
- Category: product-launch
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://openai.com/index/introducing-the-codex-app
- Record: https://forck.live/items/286-introducing-the-codex-app
- Subject: GPT / ChatGPT / API

OpenAI introduced the Codex app for macOS, a command center for AI coding and software development featuring multiple agents, parallel workflows, and long-running tasks.