# forck.live — 21–27 July 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-W30, 21–27 July 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-W30 - The current issue, and the catalogue of the other weeks: https://forck.live/briefing.md - Catalogue: https://forck.live/llms.txt ## Announcements ### Alibaba — GSPO: Towards Scalable Reinforcement Learning for Language Models - Company: Alibaba (alibaba.com) - Announced: 2025-07-27T07:00:00+00:00 - Category: research-paper - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://qwenlm.github.io/blog/gspo/ - Record: https://forck.live/items/2284-gspo-towards-scalable-reinforcement-learning-for-language-models - Subject: Qwen Alibaba proposes the Group Sequence Policy Optimization (GSPO) algorithm to address instability and model collapse issues in existing reinforcement learning algorithms for language models, aiming to enable scalable RL training. ### Runway — Introducing Runway Aleph - Company: Runway (runway.com) - Announced: 2025-07-25T16:22:35.195000+00:00 - Category: new-model - Coverage: not counted - Announcement: yes - Group: models - Source: https://runway.com/research/introducing-runway-aleph - Record: https://forck.live/items/7905-introducing-runway-aleph - Subject: Gen / Aleph Runway introduced Runway Aleph, a state-of-the-art in-context video model that can perform multiple visual generation and editing tasks on input videos. The model is now available for all paid users. ### Vercel — Vercel and Solara6 partner to build better ecommerce experiences - Company: Vercel (vercel.com) - Announced: 2025-07-25T13:00:00+00:00 - Category: partnership-acquisition - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://vercel.com/blog/vercel-and-solara6-partner-to-build-better-ecommerce-experiences - Record: https://forck.live/items/20852-vercel-and-solara6-partner-to-build-better-ecommerce-experiences - Subject: v0 / AI SDK / AI Gateway Vercel announced a partnership with digital agency Solara6 to improve ecommerce experiences. The collaboration aims to enhance SEO, site speed, scalability, and reliability for ecommerce teams, particularly during peak traffic events. Solara6 will leverage Vercel's infrastructure and features like Previews, Functions, and Web Analytics. ### Vercel — Model Context Protocol (MCP) explained: An FAQ - Company: Vercel (vercel.com) - Announced: 2025-07-25T13:00:00+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://vercel.com/blog/model-context-protocol-mcp-explained - Record: https://forck.live/items/20851-model-context-protocol-mcp-explained-an-faq - Subject: v0 / AI SDK / AI Gateway Vercel's FAQ explains the Model Context Protocol (MCP), a specification that standardizes how large language models access external data sources, tools, and systems, enabling safe and predictable interactions. It describes MCP as a universal connector between AI and applications, contrasting it with custom integrations and APIs. ### Together AI — Qwen3-Coder: The Most Capable Agentic Coding Model Now Available on Together AI - Company: Together AI (together.ai) - Announced: 2025-07-25 - Category: new-model - Coverage: not counted - Announcement: yes - Group: models - Source: https://www.together.ai/blog/qwen-3-coder - Record: https://forck.live/items/2420-qwen3-coder-the-most-capable-agentic-coding-model-now-available-on-together-ai - Subject: Inference platform Together AI announces the availability of Qwen3-Coder-480B-A35B-Instruct, a 480-billion-parameter agentic coding model with 256K context and frontier performance on SWE-bench Verified, comparable to Claude Sonnet 4. ### Hugging Face — Parquet Content-Defined Chunking - Company: Hugging Face (huggingface.co) - Announced: 2025-07-25 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://huggingface.co/blog/parquet-cdc - Record: https://forck.live/items/1649-parquet-content-defined-chunking - Subject: Platform Parquet Content-Defined Chunking (CDC) is now available in PyArrow and Pandas, enabling efficient deduplication of Parquet files on Hugging Face's Xet storage layer, reducing data transfer and storage costs. ### Hugging Face — Say hello to `hf`: a faster, friendlier Hugging Face CLI ✨ - Company: Hugging Face (huggingface.co) - Announced: 2025-07-25 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://huggingface.co/blog/hf-cli - Record: https://forck.live/items/1648-say-hello-to-hf-a-faster-friendlier-hugging-face-cli - Subject: Platform Hugging Face announced a major update to its CLI, renaming it from huggingface-cli to hf and reorganizing commands into a "hf " pattern. Additionally, a new "hf jobs" command was introduced, allowing users to run scripts or Docker images on Hugging Face infrastructure with pay-as-you-go billing, available to Pro and Team/Enterprise users. ### Alibaba — Qwen-MT: Where Speed Meets Smart Translation - Company: Alibaba (alibaba.com) - Announced: 2025-07-24T14:00:00+00:00 - Category: model-update - Coverage: not counted - Announcement: yes - Group: models - Source: https://qwenlm.github.io/blog/qwen-mt/ - Record: https://forck.live/items/2285-qwen-mt-where-speed-meets-smart-translation - Subject: Qwen Alibaba introduces an update to Qwen-MT (qwen-mt-turbo) via Qwen API, built on Qwen3 with trillions of multilingual and translation tokens, and reinforcement learning to improve translation accuracy and fluency, supporting 92 languages. ### Factory AI — Microsoft Azure Marketplace now features Factory - Company: Factory AI (factory.com) - Announced: 2025-07-24 - Category: availability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://factory.com/news/microsoft-azure-marketplace - Record: https://forck.live/items/17858-microsoft-azure-marketplace-now-features-factory - Subject: Droid Factory is now available on the Microsoft Azure Marketplace, allowing enterprise engineering teams to acquire it using existing Microsoft Azure Consumption Commitment (MACC) to accelerate procurement. ### Aleph Alpha — Aleph Alpha Expands Leadership Team - Company: Aleph Alpha (aleph-alpha.com) - Announced: 2025-07-24 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://aleph-alpha.com/en/news/leadership-team-expanded - Record: https://forck.live/items/16860-aleph-alpha-expands-leadership-team - Subject: Kolibri / PhariaAI Aleph Alpha appointed Reto Spörri as Co-CEO and Ilhan Scheer as Chief Growth Officer, and promoted Christopher Kränzler to Chief Product Officer. The changes follow the development of its PhariaAI operating system and aim to strengthen go-to-market execution and commercialization. ### Amp — amp dash x - Company: Amp (ampcode.com) - Announced: 2025-07-24 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://ampcode.com/news/amp-x - Record: https://forck.live/items/7826-amp-dash-x - Subject: Amp The Amp CLI added a new --execute flag (short -x) that enables programmatic use, sending a prompt to the agent and printing the last message. It also supports piping input and automatic activation when stdout is redirected. ### OpenAI — Resolving digital threats 100x faster with OpenAI - Company: OpenAI (openai.com) - Announced: 2025-07-24 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://openai.com/index/outtake - Record: https://forck.live/items/516-resolving-digital-threats-100x-faster-with-openai - Subject: GPT / ChatGPT / API Outtake uses OpenAI models GPT-4.1 and o3 to build AI agents that detect and resolve digital threats 100 times faster. ### Lovable — $100M ARR & Lovable Agent - Company: Lovable (lovable.dev) - Announced: 2025-07-23T18:00:00+00:00 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://lovable.dev/blog/agent - Record: https://forck.live/items/12137-100m-arr-lovable-agent - Subject: Lovable / AI app builder Lovable announced it has surpassed $100 million in annual recurring revenue (ARR) within eight months of its first $1 million, making it the fastest-growing startup globally. The company also introduced Lovable Agent as the default mode, which handles complex, multi-step tasks end-to-end and reduces errors by 91%. Pricing is credit-based, with costs varying by request complexity. ### Vercel — Build your own AI app builder with the v0 Platform API - Company: Vercel (vercel.com) - Announced: 2025-07-23T13:00:00+00:00 - Category: api-release - Coverage: not counted - Announcement: no - Group: routine - Source: https://vercel.com/blog/build-your-own-ai-app-builder-with-the-v0-platform-api - Record: https://forck.live/items/20853-build-your-own-ai-app-builder-with-the-v0-platform-api - Subject: v0 / AI SDK / AI Gateway Vercel announced the v0 Platform API, a REST interface in beta that provides programmatic access to v0's code generation pipeline, allowing developers to automate building web apps, integrate code generation into existing features, and build new products on top of LLM-generated UIs. The API includes endpoints for natural language app generation, bringing custom context, and project management with deployments. A TypeScript SDK (v0-sdk) is available to simplify interaction with the API. ### Genspark — Seeing AGI (5): Multi-Model and Multi-Agent is the Future - Company: Genspark (genspark.ai) - Announced: 2025-07-23 - Category: not stated - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://www.genspark.ai/blog/seeing-agi-impact5 - Record: https://forck.live/items/18275-seeing-agi-5-multi-model-and-multi-agent-is-the-future - Subject: Genspark ### Amp — Towards a New CLI - Company: Amp (ampcode.com) - Announced: 2025-07-23 - Category: developer-tool-release - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://ampcode.com/news/towards-a-new-cli - Record: https://forck.live/items/7827-towards-a-new-cli - Subject: Amp Amp released a new version of its CLI, designed to support agents in the terminal. ### Hugging Face — Fast LoRA inference for Flux with Diffusers and PEFT - Company: Hugging Face (huggingface.co) - Announced: 2025-07-23 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://huggingface.co/blog/lora-fast - Record: https://forck.live/items/1651-fast-lora-inference-for-flux-with-diffusers-and-peft - Subject: Platform An optimization recipe for LoRA inference on the Flux.1-Dev model using Diffusers and PEFT, achieving approximately 2.3x speedup through Flash Attention 3, torch.compile, FP8 quantization, and hotswapping without recompilation. ### Hugging Face — TimeScope: How Long Can Your Video Large Multimodal Model Go? - Company: Hugging Face (huggingface.co) - Announced: 2025-07-23 - Category: research-paper - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://huggingface.co/blog/timescope-video-lmm-benchmark - Record: https://forck.live/items/1650-timescope-how-long-can-your-video-large-multimodal-model-go - Subject: Platform TimeScope is an open-source benchmark designed to measure how well vision-language models understand long videos, evaluating localized retrieval, information synthesis, and fine-grained temporal perception across videos from 1 minute to 8 hours. ### OpenAI — Announcing OpenAI DevDay 2025 - Company: OpenAI (openai.com) - Announced: 2025-07-23 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://openai.com/index/announcing-devday-2025 - Record: https://forck.live/items/518-announcing-openai-devday-2025 - Subject: GPT / ChatGPT / API OpenAI announced the date and location of its third annual DevDay event, scheduled for October 6, 2025 in San Francisco. ### OpenAI — Model ML is helping financial firms rebuild with AI from the ground up - Company: OpenAI (openai.com) - Announced: 2025-07-23 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://openai.com/index/model-ml-chaz-englander - Record: https://forck.live/items/517-model-ml-is-helping-financial-firms-rebuild-with-ai-from-the-ground-up - Subject: GPT / ChatGPT / API OpenAI published an article in its Executive Function series featuring an interview with Model ML CEO Chaz Englander about AI-native infrastructure and autonomous agents in financial services. ### Replit — Introducing Queue: A smarter way to work with Agent - Company: Replit (replit.com) - Announced: 2025-07-22T22:08:00+00:00 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://replit.com/blog/introducing-queue-a-smarter-way-to-work-with-agent - Record: https://forck.live/items/17948-introducing-queue-a-smarter-way-to-work-with-agent - Subject: Replit Agent Replit introduced Queue, a feature that lets users submit multiple requests to the Replit Agent while it is actively working, allowing tasks to be processed in order without waiting. Users can edit, reorder, or reprioritize messages and attach files or adjust settings per task. Queue is available to all Replit Core and Replit Teams users. ### Harvey — Preparing for Legal Super Intelligence - Company: Harvey (harvey.ai) - Announced: 2025-07-22T20:00:00+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://www.harvey.ai/blog/preparing-for-legal-super-intelligence - Record: https://forck.live/items/17695-preparing-for-legal-super-intelligence - Subject: Harvey As AGI starts to come into focus, we explore what legal super intelligence means and how the legal industry can start preparing now. Clients have been increasingly asking how we view artificial general intelligence (AGI) and artificial super-intelligence (ASI). ### Harvey — Harvey Expands Global Data Coverage - Company: Harvey (harvey.ai) - Announced: 2025-07-22T15:29:00+00:00 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://www.harvey.ai/blog/harvey-expands-global-data-coverage - Record: https://forck.live/items/17696-harvey-expands-global-data-coverage - Subject: Harvey Harvey has integrated eight new national legal knowledge sources into its platform, covering Austria, Finland, Germany, India, the Netherlands, Norway, Spain, and Switzerland. These additions expand the existing coverage of Australia, EDGAR, Eur-Lex, Singapore, and Sweden, enabling cross-jurisdictional research in a unified interface. The company also lists 17 additional countries planned for coverage by October. ### Alibaba — Qwen3-Coder: Agentic Coding in the World - Company: Alibaba (alibaba.com) - Announced: 2025-07-22T13:00:00+00:00 - Category: new-model - Coverage: not counted - Announcement: yes - Group: models - Source: https://qwenlm.github.io/blog/qwen3-coder/ - Record: https://forck.live/items/2286-qwen3-coder-agentic-coding-in-the-world - Subject: Qwen Alibaba's Qwen team announces Qwen3-Coder, a new agentic code model, with its most powerful variant Qwen3-Coder-480B-A35B-Instruct (480B parameters, 35B active, MoE) supporting 256K native context and 1M with extrapolation, achieving state-of-the-art results among open models on agentic coding, browser-use, and tool-use tasks, comparable to Claude Sonnet 4. ### OpenAI — Pioneering an AI clinical copilot with Penda Health - Company: OpenAI (openai.com) - Announced: 2025-07-22T10:00:00+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://openai.com/index/ai-clinical-copilot-penda-health - Record: https://forck.live/items/519-pioneering-an-ai-clinical-copilot-with-penda-health - Subject: GPT / ChatGPT / API OpenAI and Penda Health debut an AI clinical copilot that reduces diagnostic errors by 16% in real-world use, offering a new path for safe, effective AI in healthcare. ### Suno — Songs of Love Foundation - Company: Suno (suno.com) - Announced: 2025-07-22 - Category: partnership-acquisition - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://suno.com/blog/songs-of-love-suno - Record: https://forck.live/items/10331-songs-of-love-foundation - Subject: Music / Studio Suno has partnered with Songs of Love Foundation, a nonprofit that creates personalized songs for children and older adults facing illness or hardship. The collaboration will use Suno's AI music tools to help volunteers produce custom songs at scale, and Suno is launching a Creator Program to recruit additional volunteers. ### Cognition — Devin's MCP Marketplace - Company: Cognition (cognition.com) - Announced: 2025-07-22 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://cognition.com/blog/mcp-marketplace - Record: https://forck.live/items/7696-devin-s-mcp-marketplace - Subject: Devin / SWE Cognition announced that Devin, its AI software engineering tool, is now compatible with the Model Context Protocol (MCP) and has introduced an MCP Marketplace. This allows users to connect Devin to various MCP servers for tasks such as querying databases, creating documents, and interacting with tools like Datadog, Sentry, and Figma. ### Mistral AI — Our contribution to a global environmental standard for AI - Company: Mistral AI (mistral.ai) - Announced: 2025-07-22 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://mistral.ai/news/our-contribution-to-a-global-environmental-standard-for-ai - Record: https://forck.live/items/3317-our-contribution-to-a-global-environmental-standard-for-ai - Subject: Mistral / Le Chat Mistral AI announces its contribution to a global environmental standard for AI. ### OpenAI — OpenAI’s new economic analysis - Company: OpenAI (openai.com) - Announced: 2025-07-22 - Category: research-paper - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://openai.com/global-affairs/new-economic-analysis - Record: https://forck.live/items/521-openai-s-new-economic-analysis - Subject: GPT / ChatGPT / API OpenAI publishes an economic analysis on ChatGPT's impact and announces a new research collaboration on AI's effects on labor and productivity. ### OpenAI — Stargate advances with 4.5 GW partnership with Oracle - Company: OpenAI (openai.com) - Announced: 2025-07-22 - Category: partnership-acquisition - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://openai.com/index/stargate-advances-with-partnership-with-oracle - Record: https://forck.live/items/520-stargate-advances-with-4-5-gw-partnership-with-oracle - Subject: GPT / ChatGPT / API Oracle and OpenAI partner to develop 4.5 GW of additional Stargate data center capacity in the U.S., creating jobs and advancing AI leadership. ### NVIDIA — Accelerate a World of LLMs on Hugging Face with NVIDIA NIM - Company: NVIDIA (nvidia.com) - Announced: 2025-07-21T18:01:30+00:00 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://huggingface.co/blog/nvidia/multi-llm-nim - Record: https://forck.live/items/1652-accelerate-a-world-of-llms-on-hugging-face-with-nvidia-nim NVIDIA NIM now supports deploying a broad range of LLMs from Hugging Face using a single Docker container that automatically selects and optimizes an inference backend (TensorRT-LLM, vLLM, or SGLang). ### Replit — Introducing a safer way to Vibe Code with Replit Databases - Company: Replit (replit.com) - Announced: 2025-07-21T16:06:00+00:00 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://replit.com/blog/introducing-a-safer-way-to-vibe-code-with-replit-databases - Record: https://forck.live/items/17949-introducing-a-safer-way-to-vibe-code-with-replit-databases - Subject: Replit Agent Replit introduced separate development and production databases for its apps, rolling out in Beta for new apps and gradually reaching all users. This separation allows users to safely test and iterate without affecting live customer data, with automatic migration of existing apps planned in the coming weeks. ### OpenAI — OpenAI and UK Government announce strategic partnership to deliver AI-driven growth - Company: OpenAI (openai.com) - Announced: 2025-07-21T10:00:00+00:00 - Category: partnership-acquisition - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://openai.com/global-affairs/openai-and-uk-government-partnership - Record: https://forck.live/items/522-openai-and-uk-government-announce-strategic-partnership-to-deliver-ai-driven - Subject: GPT / ChatGPT / API OpenAI partners with the UK Government to boost AI adoption, drive economic growth, and enhance public services. ### Ant Ling — Introducing Ming-Lite-Omni V1.5 - Company: Ant Ling (ant-ling.com) - Announced: 2025-07-21 - Category: model-update - Coverage: not counted - Announcement: yes - Group: models - Source: https://developer.ant-ling.com/en/blogs/ming-lite-omni-1_5 - Record: https://forck.live/items/22341-introducing-ming-lite-omni-v1-5 - Subject: Ling / Ring / Ming Ant Ling released Ming-Lite-Omni V1.5, an open-weight upgrade to its omni-modal model. Built on Ling-lite-1.5, it has 20.3 billion total parameters with 3 billion active in its MoE section, and improves performance in image-text, document, video, and speech understanding, as well as image generation and editing. The model achieves a GenEval score of 0.87 and a win rate of 87.07% against the previous version on internal evaluations. ### Upstage — Solar Pro 2 Breaks Into Global Frontier AI - Company: Upstage (upstage.ai) - Announced: 2025-07-21 - Category: funding-company-news - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://www.upstage.ai/news/solar-pro-2-frontier - Record: https://forck.live/items/18429-solar-pro-2-breaks-into-global-frontier-ai - Subject: Solar Pro / Solar Mini Upstage's proprietary LLM Solar Pro 2 was named one of the top 10 frontier models by Artificial Analysis, scoring 58 on the Intelligence Index and outperforming GPT-4.1 (53), GPT-4o (41), and LLaMA 4 Maverick (51). It is the only Korean-developed model on the index and achieved this with 31 billion parameters, surpassing many larger models. Upstage is developing a 100B+ parameter model to push further into the upper tier. ### Factory AI — Compressing Context - Company: Factory AI (factory.com) - Announced: 2025-07-21 - Category: research-paper - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://factory.com/news/compressing-context - Record: https://forck.live/items/17824-compressing-context - Subject: Droid Factory describes a method for compressing conversation context in LLM-based agents by maintaining a persistent, incrementally updated summary rather than re-summarizing the entire prefix each turn. The approach uses two thresholds to trigger compression and control the tradeoff between context preservation and cost. The post also discusses the risks of over-compression and the importance of retaining session intent, high-level play-by-play, and artifact trails. ### Replicate — Generate consistent characters - Company: Replicate (replicate.com) - Announced: 2025-07-21 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://replicate.com/blog/generate-consistent-characters - Record: https://forck.live/items/2483-generate-consistent-characters - Subject: Platform A blog post comparing several image generation models for creating consistent characters from a single reference image, including pricing and speed comparisons. ### OpenAI — AI as the greatest source of empowerment for all - Company: OpenAI (openai.com) - Announced: 2025-07-21 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://openai.com/index/ai-as-the-greatest-source-of-empowerment-for-all - Record: https://forck.live/items/523-ai-as-the-greatest-source-of-empowerment-for-all - Subject: GPT / ChatGPT / API This is a personal blog post where the author expresses a pragmatic view of AI technology, emphasizing its potential to empower people and create opportunities. The post does not announce any product, model, or specific features, but rather shares the author's positive outlook on AI's impact.