# forck.live — 5–11 January 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-W02, 5–11 January 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-W02 - The current issue, and the catalogue of the other weeks: https://forck.live/briefing.md - Catalogue: https://forck.live/llms.txt ## Announcements ### Sakana AI — Extending the Context of Pretrained LLMs by Dropping Their Positional Embeddings - Company: Sakana AI (sakana.ai) - Announced: 2026-01-11T15:00:00+00:00 - Category: research-paper - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://sakana.ai/drope - Record: https://forck.live/items/4700-extending-the-context-of-pretrained-llms-by-dropping-their-positional-embeddings - Subject: Sakana models Sakana AI introduces DroPE, a method to extend the context length of pretrained LLMs by dropping positional embeddings during inference, requiring less than 1% of the original pretraining budget and outperforming established methods on LongBench and RULER. ### Lovable — How ElevenLabs Builds Mind-Blowing Demos with Lovable - Company: Lovable (lovable.dev) - Announced: 2026-01-09T16:00:00+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://lovable.dev/blog/elevenlabs - Record: https://forck.live/items/12112-how-elevenlabs-builds-mind-blowing-demos-with-lovable - Subject: Lovable / AI app builder ElevenLabs uses Lovable to build interactive demos for enterprise prospects, reducing demo creation time by at least 50% and enabling cross-selling of multiple audio AI products. ### Vercel — How to build agents with filesystems and bash - Company: Vercel (vercel.com) - Announced: 2026-01-09T13:00:00+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://vercel.com/blog/how-to-build-agents-with-filesystems-and-bash - Record: https://forck.live/items/20768-how-to-build-agents-with-filesystems-and-bash - Subject: v0 / AI SDK / AI Gateway Vercel describes an agent architecture that replaces custom tooling with a filesystem tool and a bash tool. The approach reduced the cost of their sales call summarization agent from about $1.00 to about $0.25 per call on Claude Opus 4.5 while improving output quality. The post explains how structuring data as files and using Unix commands allows agents to retrieve precise context on demand. ### ElevenLabs — Introducing Scribe v2 - Company: ElevenLabs (elevenlabs.io) - Announced: 2026-01-09T12:00:00+00:00 - Category: model-update - Coverage: not counted - Announcement: yes - Group: models - Source: https://elevenlabs.io/blog/introducing-scribe-v2 - Record: https://forck.live/items/3372-introducing-scribe-v2 - Subject: Voice AI ElevenLabs introduces Scribe v2. ### OpenAI — OpenAI and SoftBank Group partner with SB Energy - Company: OpenAI (openai.com) - Announced: 2026-01-09T11:00:00+00:00 - Category: partnership-acquisition - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://openai.com/index/stargate-sb-energy-partnership - Record: https://forck.live/items/315-openai-and-softbank-group-partner-with-sb-energy - Subject: GPT / ChatGPT / API OpenAI and SoftBank Group are partnering with SB Energy to develop multi-gigawatt AI data center campuses, including a 1.2 GW facility in Texas that supports the Stargate initiative. ### Amp — Agents Panel - Company: Amp (ampcode.com) - Announced: 2026-01-09 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://ampcode.com/news/agents-panel - Record: https://forck.live/items/7766-agents-panel - Subject: Amp The Amp editor extension added a new panel to view and manage all active agent threads, with keyboard shortcuts for navigation and actions. ### LG AI Research — SciNO: A Functional Diffusion Model with Neural Operators for Causal Ordering - Company: LG AI Research (lgresearch.ai) - Announced: 2026-01-09 - Category: research-paper - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://www.lgresearch.ai/blog/view?seq=615 - Record: https://forck.live/items/4764-scino-a-functional-diffusion-model-with-neural-operators-for-causal-ordering - Subject: EXAONE LG AI Research announces SciNO, a functional diffusion model framework with neural operators for stable and scalable causal ordering in high-dimensional data, presented at NeurIPS 2025. ### OpenAI — Datadog uses Codex for system-level code review - Company: OpenAI (openai.com) - Announced: 2026-01-09 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://openai.com/index/datadog - Record: https://forck.live/items/316-datadog-uses-codex-for-system-level-code-review - Subject: GPT / ChatGPT / API The source text is a description of a brand graphic featuring OpenAI and Datadog logos. It contains no announcement about an AI product or model. ### Harvey — Never Start From Scratch Again With Memory in Harvey - Company: Harvey (harvey.ai) - Announced: 2026-01-08T13:00:00+00:00 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://www.harvey.ai/blog/memory-in-harvey - Record: https://forck.live/items/17629-never-start-from-scratch-again-with-memory-in-harvey - Subject: Harvey Harvey is inviting customers to help shape a new feature called Memory, which will carry forward context such as matter details, precedent, and working preferences across AI outputs. The feature includes governance controls for firm admins to manage access and scope. Harvey plans to host customer listening tours and working sessions to co-develop the feature. ### OpenAI — OpenAI for Healthcare - Company: OpenAI (openai.com) - Announced: 2026-01-08T12:00:00+00:00 - Category: product-launch - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://openai.com/index/openai-for-healthcare - Record: https://forck.live/items/317-openai-for-healthcare - Subject: GPT / ChatGPT / API OpenAI announces OpenAI for Healthcare, a secure, enterprise-grade AI offering that supports HIPAA compliance, aimed at reducing administrative burden and supporting clinical workflows. ### Supabase — Introducing TRAE SOLO integration with Supabase - Company: Supabase (supabase.com) - Announced: 2026-01-08 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://supabase.com/blog/introducing-trae-solo-integration - Record: https://forck.live/items/21309-introducing-trae-solo-integration-with-supabase - Subject: Supabase Supabase announced an integration with TRAE SOLO, ByteDance's AI coding agent, allowing developers to manage their Supabase backend—including database queries, storage, authentication, secrets, and logs—directly within the TRAE IDE without switching to the Supabase dashboard. ### Aleph Alpha — Aleph Alpha Aligns Its Organization With Strategic Priorities - Company: Aleph Alpha (aleph-alpha.com) - Announced: 2026-01-08 - Category: funding-company-news - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://aleph-alpha.com/en/news/organization-aligned-with-strategic-priorities - Record: https://forck.live/items/16859-aleph-alpha-aligns-its-organization-with-strategic-priorities - Subject: Kolibri / PhariaAI Aleph Alpha announced a workforce reduction of approximately 50 employees as part of a strategy process initiated in the second half of 2025. The company stated the reduction is intended to focus on core capabilities and areas with highest growth potential, and that no further measures of this kind are planned. ### Amp — Efficient MCP Tool Loading - Company: Amp (ampcode.com) - Announced: 2026-01-08 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://ampcode.com/news/lazy-load-mcp-with-skills - Record: https://forck.live/items/7768-efficient-mcp-tool-loading - Subject: Amp Amp introduced a feature that lets users combine MCP server configurations with Agent Skills, so that tool definitions are loaded only when the skill is invoked, reducing token consumption. ### Amp — The Frontier Is Now Free - Company: Amp (ampcode.com) - Announced: 2026-01-08 - Category: availability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://ampcode.com/news/amp-free-frontier - Record: https://forck.live/items/7767-the-frontier-is-now-free - Subject: Amp Amp announced that every user now receives daily free credits to access the full Amp experience, including its frontier smart agent powered by Opus 4.5. The free credit grant replenishes hourly, providing $10 worth of credits per day, supported by text ads that users can opt out of by paying for inference. The company also noted that rush mode, offering faster inference at lower cost, is currently driven by Haiku 4.5. ### Baidu — ERNIE-5.0-Preview-1220 Becomes the Sole Chinese Model in LMArena Vision Top 10! - Company: Baidu (baidu.com) - Announced: 2026-01-08 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://ernie.baidu.com/blog/posts/ernie-5.0-preview-1220-release-on-lmarena/ - Record: https://forck.live/items/4646-ernie-5-0-preview-1220-becomes-the-sole-chinese-model-in-lmarena-vision-top-10 - Subject: ERNIE ERNIE-5.0-Preview-1220 achieved a score of 1,226 points, ranking first among Chinese models and eighth globally on LMArena Vision, becoming the only Chinese large model in the global top 10 for visual understanding. ### Together AI — How to choose the right open model for production - Company: Together AI (together.ai) - Announced: 2026-01-08 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://www.together.ai/blog/how-to-choose-the-right-open-model-for-production - Record: https://forck.live/items/2390-how-to-choose-the-right-open-model-for-production - Subject: Inference platform Guidance on selecting open models for production workloads, discussing reasons to choose open models (transparency, adaptability, control), legal considerations, model size guidelines, tradeoffs between cost, speed, and quality, and evaluation methods. ### OpenAI — Netomi’s lessons for scaling agentic systems into the enterprise - Company: OpenAI (openai.com) - Announced: 2026-01-08 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://openai.com/index/netomi - Record: https://forck.live/items/318-netomi-s-lessons-for-scaling-agentic-systems-into-the-enterprise - Subject: GPT / ChatGPT / API Netomi uses OpenAI's GPT-4.1 and GPT-5.2 models to scale enterprise AI agents, combining concurrency, governance, and multi-step reasoning for reliable production workflows. ### Alibaba — Qwen3-VL-Embedding and Qwen3-VL-Reranker: For the Next Generation of Multimodal Retrieval - Company: Alibaba (alibaba.com) - Announced: 2026-01-07T20:00:00+00:00 - Category: new-model - Coverage: not counted - Announcement: yes - Group: models - Source: https://qwen.ai/blog?id=qwen3-vl-embedding - Record: https://forck.live/items/4594-qwen3-vl-embedding-and-qwen3-vl-reranker-for-the-next-generation-of-multimodal - Subject: Qwen Alibaba releases Qwen3-VL-Embedding and Qwen3-VL-Reranker, new multimodal retrieval models built on Qwen3-VL, supporting text, images, screenshots, and video. The embedding model generates semantic vectors, and the reranker provides relevance scores. The 8B embedding model achieves state-of-the-art results on MMEB-v2 benchmark. ### Sakana AI — Digital Red Queen: Adversarial Program Evolution in Core War with LLMs - Company: Sakana AI (sakana.ai) - Announced: 2026-01-07T15:00:00+00:00 - Category: research-paper - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://sakana.ai/drq - Record: https://forck.live/items/4701-digital-red-queen-adversarial-program-evolution-in-core-war-with-llms - Subject: Sakana models Sakana AI, in collaboration with MIT, releases a research paper and technical report on Digital Red Queen (DRQ), an algorithm that uses large language models to drive adversarial evolution of assembly programs (warriors) in the Core War programming game, leading to emergent strategies and convergent evolution. ### Vercel — How we made v0 an effective coding agent - Company: Vercel (vercel.com) - Announced: 2026-01-07T13:00:00+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://vercel.com/blog/how-we-made-v0-an-effective-coding-agent - Record: https://forck.live/items/20769-how-we-made-v0-an-effective-coding-agent - Subject: v0 / AI SDK / AI Gateway Vercel describes how its v0 coding agent uses a dynamic system prompt, a streaming manipulation layer called LLM Suspense, and autofixers to reduce code generation errors. The company states that LLMs in isolation produce errors as often as 10% of the time, and its composite pipeline can achieve a double-digit increase in success rates. ### OpenAI — How Tolan builds voice-first AI with GPT-5.1 - Company: OpenAI (openai.com) - Announced: 2026-01-07T10:00:00+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://openai.com/index/tolan - Record: https://forck.live/items/319-how-tolan-builds-voice-first-ai-with-gpt-5-1 - Subject: GPT / ChatGPT / API Tolan built a voice-first AI companion using GPT-5.1, combining low-latency responses, real-time context reconstruction, and memory-driven personalities for natural conversations. ### Supabase — Supabase Security Retro: 2025 - Company: Supabase (supabase.com) - Announced: 2026-01-07 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://supabase.com/blog/supabase-security-2025-retro - Record: https://forck.live/items/21310-supabase-security-retro-2025 - Subject: Supabase Supabase introduced several security enhancements in 2025, including the ability to disable the Data API, a new API key model with asymmetric JWTs and automatic revocation of leaked keys, RLS enabled by default for new tables, Security Advisors that scan for misconfigurations, and an AI Assistant to help fix security issues. These changes aim to provide safer defaults and better tooling for developers. ### Factory AI — Let Me Droid That For You: What 780K Agent Searches Reveal - Company: Factory AI (factory.com) - Announced: 2026-01-07 - Category: research-paper - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://factory.com/news/what-droid-searches - Record: https://forck.live/items/17875-let-me-droid-that-for-you-what-780k-agent-searches-reveal - Subject: Droid Factory Research analyzed 780,000 web searches performed by its AI coding agent Droid, finding that documentation and learning materials account for 40% of all queries and API references for another 14%. The agent uses WebSearch for exploratory discovery and FetchUrl for targeted retrieval, mirroring human research patterns. GitHub is the most fetched domain, followed by Factory's own documentation. ### Amp — User Invokable Skills - Company: Amp (ampcode.com) - Announced: 2026-01-07 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://ampcode.com/news/user-invokable-skills - Record: https://forck.live/items/7769-user-invokable-skills - Subject: Amp Amp added the ability for users to manually invoke skills, forcing the agent to use a chosen skill on the next message, rather than waiting for the agent to decide. ### Cognition — Infosys partners with Cognition to expand engineering capacity and help scale its enterprise business - Company: Cognition (cognition.com) - Announced: 2026-01-07 - Category: partnership-acquisition - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://cognition.com/blog/infosys-cognition - Record: https://forck.live/items/7684-infosys-partners-with-cognition-to-expand-engineering-capacity-and-help-scale - Subject: Devin / SWE Infosys has partnered with Cognition to deploy Devin, an AI software engineer, across its organization and global client base. The collaboration aims to expand delivery capacity by augmenting Infosys's workforce with autonomous software engineers, initially in its Financial Services practice and later in other verticals. Infosys will use Devin for internal productivity, services delivery, and as a managed service provider within customer environments. ### OpenAI — Introducing ChatGPT Health - Company: OpenAI (openai.com) - Announced: 2026-01-07 - Category: product-launch - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://openai.com/index/introducing-chatgpt-health - Record: https://forck.live/items/320-introducing-chatgpt-health - Subject: GPT / ChatGPT / API OpenAI introduces ChatGPT Health, a dedicated experience that securely connects health data and apps with privacy protections and physician-informed design. ### Harvey — Harvey Expands EMEA Presence With New Dublin Office - Company: Harvey (harvey.ai) - Announced: 2026-01-06T08:00:00+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://www.harvey.ai/blog/harvey-expands-emea-presence-with-new-dublin-office - Record: https://forck.live/items/17630-harvey-expands-emea-presence-with-new-dublin-office - Subject: Harvey Harvey announced it will open a new office in Dublin in late March, with 20 hires planned in the first year across finance and people roles, to strengthen its EMEA presence and support legal and professional services teams in the region. ### xAI — xAI Raises $20B Series E - Company: xAI (x.ai) - Announced: 2026-01-06 - Category: funding-company-news - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://x.ai/news/series-e - Record: https://forck.live/items/3476-xai-raises-20b-series-e - Subject: Grok xAI raised $20 billion in a Series E funding round. ### NVIDIA — NVIDIA Cosmos Reason 2 Brings Advanced Reasoning To Physical AI - Company: NVIDIA (nvidia.com) - Announced: 2026-01-05T22:56:51+00:00 - Category: model-update - Coverage: not counted - Announcement: yes - Group: models - Source: https://huggingface.co/blog/nvidia/nvidia-cosmos-reason-2-brings-advanced-reasoning - Record: https://forck.live/items/1565-nvidia-cosmos-reason-2-brings-advanced-reasoning-to-physical-ai NVIDIA announced the release of Cosmos Reason 2, an open reasoning vision-language model for physical AI, with improvements in spatio-temporal understanding, 256K token context window, and support for 2B and 8B parameter sizes. ### Runway — Runway Advances Video Generation and World Models With NVIDIA Rubin Platform - Company: Runway (runway.com) - Announced: 2026-01-05T10:10:10.110000+00:00 - Category: partnership-acquisition - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://runway.com/news/runway-partners-with-nvidia - Record: https://forck.live/items/7899-runway-advances-video-generation-and-world-models-with-nvidia-rubin-platform - Subject: Gen / Aleph Runway demonstrated its Gen-4.5 video generation model and GWM-1 world model on the NVIDIA Rubin platform, porting Gen-4.5 from Hopper to Vera Rubin NVL72 in a single day. The partnership aims to accelerate video generation and world models for applications like robotics training and interactive avatars. ### TII — Falcon-H1R: Pushing the Reasoning Frontiers with a Hybrid Model for Efficient Test-Time Scaling - Company: TII (tii.ae) - Announced: 2026-01-05T08:00:00+00:00 - Category: new-model - Coverage: not counted - Announcement: yes - Group: models - Source: https://falcon-lm.github.io/blog/falcon-h1r-7b/ - Record: https://forck.live/items/18412-falcon-h1r-pushing-the-reasoning-frontiers-with-a-hybrid-model-for-efficient - Subject: Falcon LLM TII released Falcon H1R 7B, a decoder-only large language model built on the Falcon-H1 Base. The 7B-parameter model matches or outperforms reasoning models 2–7× larger on math, code, and general benchmarks. Its training uses a two-stage pipeline of supervised fine-tuning and GRPO reinforcement learning, and it employs a Deep Think with Confidence (DeepConf) method for test-time scaling. ### TII — Introducing Falcon-H1-Arabic: Pushing the Boundaries of Arabic Language AI with Hybrid Architecture - Company: TII (tii.ae) - Announced: 2026-01-05T01:00:00+00:00 - Category: new-model - Coverage: not counted - Announcement: yes - Group: models - Source: https://falcon-lm.github.io/blog/falcon-h1-arabic/ - Record: https://forck.live/items/18413-introducing-falcon-h1-arabic-pushing-the-boundaries-of-arabic-language-ai-with - Subject: Falcon LLM TII released Falcon-H1-Arabic, a family of Arabic language models (3B, 7B, 34B parameters) built on a hybrid Mamba-Transformer architecture. The models support context windows of up to 256K tokens and were trained on a mix of Arabic, English, and multilingual data totaling around 300 billion tokens. Post-training included supervised fine-tuning and direct preference optimization to improve long-context reasoning and alignment. ### Hugging Face — NVIDIA brings agents to life with DGX Spark and Reachy Mini - Company: Hugging Face (huggingface.co) - Announced: 2026-01-05 - Category: developer-tool-release - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://huggingface.co/blog/nvidia-reachy-mini - Record: https://forck.live/items/1567-nvidia-brings-agents-to-life-with-dgx-spark-and-reachy-mini - Subject: Platform NVIDIA and Hugging Face published a step-by-step guide showing how to build a personal AI agent using the DGX Spark and Reachy Mini robot, combining open models (Nemotron reasoning and vision models) with the NeMo Agent Toolkit for orchestration.