# forck.live — 1–7 September 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-W36, 1–7 September 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-W36 - The current issue, and the catalogue of the other weeks: https://forck.live/briefing.md - Catalogue: https://forck.live/llms.txt ## Announcements ### Harvey — 5 Questions With Kelly Kriegshauser - Company: Harvey (harvey.ai) - Announced: 2025-09-05T14:00:00+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://www.harvey.ai/blog/5-questions-with-kelly-kriegshauser - Record: https://forck.live/items/17679-5-questions-with-kelly-kriegshauser - Subject: Harvey Harvey published an interview with its Enterprise Sales Manager, Kelly Kriegshauser, who discusses the company's sales process for in-house legal teams, common objections such as the perception that Harvey is built for law firms, and the value drivers of time and cost savings through workflow automation and tool consolidation. ### OpenAI — Why language models hallucinate - Company: OpenAI (openai.com) - Announced: 2025-09-05T10:00:00+00:00 - Category: research-paper - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://openai.com/index/why-language-models-hallucinate - Record: https://forck.live/items/478-why-language-models-hallucinate - Subject: GPT / ChatGPT / API OpenAI published new research on why language models hallucinate and how better evaluations can improve AI reliability, honesty, and safety. ### OpenAI — OpenAI and Greek Government launch ‘OpenAI for Greece’ - Company: OpenAI (openai.com) - Announced: 2025-09-05T08:00:00+00:00 - Category: partnership-acquisition - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://openai.com/global-affairs/openai-for-greece - Record: https://forck.live/items/479-openai-and-greek-government-launch-openai-for-greece - Subject: GPT / ChatGPT / API OpenAI and the Greek Government launched a partnership called 'OpenAI for Greece' to introduce ChatGPT Edu into secondary schools and promote responsible AI learning. ### Factory AI — Using Linters to Direct Agents - Company: Factory AI (factory.com) - Announced: 2025-09-05 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://factory.com/news/using-linters-to-direct-agents - Record: https://forck.live/items/17874-using-linters-to-direct-agents - Subject: Droid Factory AI's guide explains how linters can encode architecture, boundaries, and ergonomics into the code generation loop for LLM agents, enabling agents to self-heal by obeying lint rules and reducing human iterations. It covers categories of agent-focused lint rules, how linters power agent-native development, and how they can drive mass refactor migrations. ### Replit — Replit expands Google Cloud Collaboration with new Google Cloud Marketplace Listing - Company: Replit (replit.com) - Announced: 2025-09-04T18:51:50+00:00 - Category: availability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://replit.com/blog/replit-expands-google-cloud-collaboration-with-new-google-cloud-marketplace-listing - Record: https://forck.live/items/17941-replit-expands-google-cloud-collaboration-with-new-google-cloud-marketplace - Subject: Replit Agent Replit announced that its platform is now available for purchase on Google Cloud Marketplace, allowing Google Cloud customers to procure Replit directly and streamline adoption. The company also noted that over 500,000 business users already build solutions using Replit, and that Replit's enterprise customers run their apps on Google Cloud infrastructure. ### Harvey — The Brief: September - Company: Harvey (harvey.ai) - Announced: 2025-09-04T13:30:00+00:00 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://www.harvey.ai/blog/the-brief-september-2025 - Record: https://forck.live/items/17680-the-brief-september - Subject: Harvey Harvey added GPT-5, deep analysis over the web, magic prompt knowledge source recommendations, full document editing in its Word Add-In, custom export formats, playbook permission levels, export verified/flagged colored cells, SharePoint one-way sync, preset files in Workflow Builder, and UX/AI enhancements to Workflow Builder. ### Vercel — Stress testing Biome's noFloatingPromises lint rule - Company: Vercel (vercel.com) - Announced: 2025-09-04T13:00:00+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://vercel.com/blog/stress-testing-biomes-nofloatingpromises-lint-rule - Record: https://forck.live/items/20828-stress-testing-biome-s-nofloatingpromises-lint-rule - Subject: v0 / AI SDK / AI Gateway Vercel partnered with the Biome team to stress-test the noFloatingPromises lint rule, which catches unhandled Promises. Engineers created edge cases including array of Promises, Promise-like objects, structural typing tricks, conditional type aliases, and Proxy-based Promises to test the rule's coverage. ### OpenAI — Expanding economic opportunity with AI - Company: OpenAI (openai.com) - Announced: 2025-09-04T11:30:00+00:00 - Category: product-launch - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://openai.com/index/expanding-economic-opportunity-with-ai - Record: https://forck.live/items/480-expanding-economic-opportunity-with-ai - Subject: GPT / ChatGPT / API OpenAI is launching a Jobs Platform and new Certifications to connect workers with jobs, training, and certifications. ### Cognition — How Eight Sleep Uses Devin as a Data Analyst - Company: Cognition (cognition.com) - Announced: 2025-09-04 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://cognition.com/blog/how-eight-sleep-uses-devin-as-a-data-analyst - Record: https://forck.live/items/7693-how-eight-sleep-uses-devin-as-a-data-analyst - Subject: Devin / SWE Eight Sleep integrated Cognition's Devin AI into its analytics stack as a data analyst. Devin fields ad-hoc data requests via Slack, queries Snowflake and Looker, and has helped resolve suspicious data issues. The company reports shipping three times as many data features and investigations per week since adopting Devin. ### Hugging Face — Welcome EmbeddingGemma, Google's new efficient embedding model - Company: Hugging Face (huggingface.co) - Announced: 2025-09-04 - Category: new-model - Coverage: not counted - Announcement: yes - Group: models - Source: https://huggingface.co/blog/embeddinggemma - Record: https://forck.live/items/1628-welcome-embeddinggemma-google-s-new-efficient-embedding-model - Subject: Platform Google releases EmbeddingGemma, a multilingual embedding model with 308M parameters, a 2k token context window, and support for over 100 languages, designed for on-device use and achieving high performance on MTEB. ### Replit — Introducing Plan Mode: A safer way to vibe code - Company: Replit (replit.com) - Announced: 2025-09-03T19:43:18+00:00 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://replit.com/blog/introducing-plan-mode-a-safer-way-to-vibe-code - Record: https://forck.live/items/17943-introducing-plan-mode-a-safer-way-to-vibe-code - Subject: Replit Agent Replit introduced Plan Mode, a new feature for its Replit Agent that provides a dedicated space for brainstorming, task planning, and strategic discussions without making any code changes. The default mode is now called Build mode. Plan Mode prohibits the Agent from making code edits but allows it to inspect the repository, search documentation, and answer questions, with the option to transition to Build mode for implementation. ### Lovable — From command line to community impact: How Shalini Ananda built Fire Fairness with Lovable - Company: Lovable (lovable.dev) - Announced: 2025-09-03T17:35:00+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://lovable.dev/blog/from-command-line-to-community-impact - Record: https://forck.live/items/12133-from-command-line-to-community-impact-how-shalini-ananda-built-fire-fairness - Subject: Lovable / AI app builder Shalini Ananda, an AI researcher with no frontend experience, built Fire Fairness using Lovable—an insurance equity engine that helps non-profit legal aid groups process wildfire insurance claims 360x faster while detecting and eliminating demographic bias. The tool, already in use by attorneys supporting victims of the 2025 Los Angeles fires, has processed 2,847 claims to date and achieves a 99.8% cost reduction compared to manual review. ### Vercel — Open SDK strategy - Company: Vercel (vercel.com) - Announced: 2025-09-03T13:00:00+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://vercel.com/blog/open-sdk-strategy - Record: https://forck.live/items/20829-open-sdk-strategy - Subject: v0 / AI SDK / AI Gateway Vercel announced an 'Open SDK' strategy, committing to build its frameworks, SDKs, and tools in the open with permissive licenses, ensure they are portable and usable on any platform, and first iterate on Vercel before generalizing to other platforms. The strategy applies to libraries and frameworks including Nuxt, Svelte, Flags, Nitro, Turborepo, and SWR, but not necessarily to Vercel's commercial products. ### Replit — Agent is now available to build apps with any framework - Company: Replit (replit.com) - Announced: 2025-09-02T17:30:00+00:00 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://replit.com/blog/agent-on-any-framework - Record: https://forck.live/items/17942-agent-is-now-available-to-build-apps-with-any-framework - Subject: Replit Agent Replit Agent now supports building apps with any language or framework, including Python, Java, Rust, Go, C#, Angular, Vue, and more. Users can import existing projects from GitHub or start from Replit's language templates and receive Agent assistance. The update removes the previous limitation of curated app types or fully custom builds without Agent support. ### Hugging Face — SAIR: Accelerating Pharma R&D with AI-Powered Structural Intelligence - Company: Hugging Face (huggingface.co) - Announced: 2025-09-02T16:54:29+00:00 - Category: open-weight-release - Coverage: not counted - Announcement: yes - Group: models - Source: https://huggingface.co/blog/SandboxAQ/sair-data-accelerating-drug-discovery-with-ai - Record: https://forck.live/items/1629-sair-accelerating-pharma-r-d-with-ai-powered-structural-intelligence - Subject: Platform SandboxAQ released the Structurally Augmented IC50 Repository (SAIR), a large open dataset of 5.24 million AI-generated 3D protein-ligand structures paired with experimental IC50 measurements, available under a permissive CC BY 4.0 license on Hugging Face. The dataset is intended to accelerate AI-powered drug discovery. ### OpenAI — Vijaye Raji to become CTO of Applications with acquisition of Statsig - Company: OpenAI (openai.com) - Announced: 2025-09-02T11:00:00+00:00 - Category: partnership-acquisition - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://openai.com/index/vijaye-raji-to-become-cto-of-applications-with-acquisition-of-statsig - Record: https://forck.live/items/481-vijaye-raji-to-become-cto-of-applications-with-acquisition-of-statsig - Subject: GPT / ChatGPT / API OpenAI acquires Statsig, and Vijaye Raji becomes CTO of Applications. ### Swiss AI Initiative — Apertus 1.0 - Company: Swiss AI Initiative (swiss-ai.org) - Announced: 2025-09-02T09:00:00+00:00 - Category: new-model - Coverage: not counted - Announcement: yes - Group: models - Source: https://apertus-ai.org/articles/2025-09-apertus-1-0/ - Record: https://forck.live/items/18381-apertus-1-0 - Subject: Apertus EPFL, ETH Zurich and CSCS released Apertus, Switzerland's first large-scale open multilingual language model, available in 8 billion and 70 billion parameter sizes under a permissive open-source license. The model was trained on 15 trillion tokens across over 1,000 languages, with 40% non-English data including underrepresented languages such as Swiss German and Romansh. Developers can access Apertus via Swisscom's sovereign Swiss AI Platform or download it from Hugging Face. ### OpenAI — Building more helpful ChatGPT experiences for everyone - Company: OpenAI (openai.com) - Announced: 2025-09-02T04:00:00+00:00 - Category: capability-change - Coverage: not counted - Announcement: no - Group: routine - Source: https://openai.com/index/building-more-helpful-chatgpt-experiences-for-everyone - Record: https://forck.live/items/482-building-more-helpful-chatgpt-experiences-for-everyone - Subject: GPT / ChatGPT / API OpenAI is adding parental controls for teens, routing sensitive conversations to reasoning models, and partnering with experts to improve ChatGPT. ### Genspark — Introducing Genspark Clip Genius - The AI Employee That Edits Any Video with One Prompt - Company: Genspark (genspark.ai) - Announced: 2025-09-02 - Category: not stated - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://www.genspark.ai/blog/genspark-clip-genius - Record: https://forck.live/items/18269-introducing-genspark-clip-genius-the-ai-employee-that-edits-any-video-with-one - Subject: Genspark ### Amp — Look Ma, No Flicker - Company: Amp (ampcode.com) - Announced: 2025-09-02 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://ampcode.com/news/look-ma-no-flicker - Record: https://forck.live/items/7806-look-ma-no-flicker - Subject: Amp Amp announced that its CLI has evolved into a fullscreen terminal user interface, featuring smooth scrolling, mouse interactions, overlays, popups, and clickable buttons, with no flicker or jarring redraws. ### Mistral AI — Le Chat. Custom MCP connectors. Memories. - Company: Mistral AI (mistral.ai) - Announced: 2025-09-02 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://mistral.ai/news/le-chat-mcp-connectors-memories - Record: https://forck.live/items/3314-le-chat-custom-mcp-connectors-memories - Subject: Mistral / Le Chat Mistral AI announces custom MCP connectors and memories for Le Chat. ### Mistral AI — Make Memory work for you. - Company: Mistral AI (mistral.ai) - Announced: 2025-09-02 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://mistral.ai/news/memory - Record: https://forck.live/items/3313-make-memory-work-for-you - Subject: Mistral / Le Chat Mistral AI announces a product or feature called Memory. ### Hugging Face — Make your ZeroGPU Spaces go brrr with ahead-of-time compilation - Company: Hugging Face (huggingface.co) - Announced: 2025-09-02 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://huggingface.co/blog/zerogpu-aoti - Record: https://forck.live/items/1630-make-your-zerogpu-spaces-go-brrr-with-ahead-of-time-compilation - Subject: Platform Hugging Face announces support for PyTorch ahead-of-time (AoT) compilation in ZeroGPU Spaces, enabling faster model loading and inference with speedups of 1.3×–1.8× on models like Flux, Wan, and LTX. ### Upstage — Upstage Information Extract Gets Document Split Beta - Company: Upstage (upstage.ai) - Announced: 2025-09-01 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://www.upstage.ai/blog/en/information-extract-document-split-beta - Record: https://forck.live/items/18465-upstage-information-extract-gets-document-split-beta - Subject: Solar Pro / Solar Mini Upstage added a Document Split beta feature to its Information Extract API, enabling automatic separation of multiple documents within a single PDF file without custom scripts.