# forck.live — 2–8 June 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-W23, 2–8 June 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-W23 - The current issue, and the catalogue of the other weeks: https://forck.live/briefing.md - Catalogue: https://forck.live/llms.txt ## Announcements ### Sarvam AI — Sarvam Translate - Company: Sarvam AI (sarvam.ai) - Announced: 2025-06-07 - Category: open-weight-release - Coverage: not counted - Announcement: yes - Group: models - Source: https://www.sarvam.ai/blogs/sarvam-translate/ - Record: https://forck.live/items/18400-sarvam-translate - Subject: Sarvam / Bulbul / Saaras Sarvam AI, in partnership with AI4Bharat, released Sarvam-Translate, an open-weights model fine-tuned from Gemma3-4B-IT. It supports paragraph-level translation for 22 Indian languages and structured content translation for 15 languages. In human evaluation, it outperformed larger models like Gemma3-27B-IT, Llama4 Scout, and Llama-3.1-405B-FP8. ### Amp — Amp Tab - Company: Amp (ampcode.com) - Announced: 2025-06-06 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://ampcode.com/news/amp-tab - Record: https://forck.live/items/7834-amp-tab - Subject: Amp Amp released Amp Tab, a new in-editor completion engine, as a research preview free for all Amp users. It uses a custom model trained to anticipate the user's next actions based on recent changes, language server diagnostics, and semantic context. ### Hugging Face — ScreenSuite - The most comprehensive evaluation suite for GUI Agents! - Company: Hugging Face (huggingface.co) - Announced: 2025-06-06 - Category: developer-tool-release - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://huggingface.co/blog/screensuite - Record: https://forck.live/items/1681-screensuite-the-most-comprehensive-evaluation-suite-for-gui-agents - Subject: Platform Hugging Face releases ScreenSuite, a comprehensive evaluation suite for GUI agents that includes 13 benchmarks covering perception, grounding, single-step actions, and multi-step agents, and evaluates leading vision-language models on vision-only tasks. ### OpenAI — How we’re responding to The New York Times’ data demands in order to protect user privacy - Company: OpenAI (openai.com) - Announced: 2025-06-05T16:30:00+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://openai.com/index/response-to-nyt-data-demands - Record: https://forck.live/items/542-how-we-re-responding-to-the-new-york-times-data-demands-in-order-to-protect - Subject: GPT / ChatGPT / API OpenAI is opposing a court order sought by The New York Times and other plaintiffs that would require indefinite retention of user data from ChatGPT and the API, citing privacy and data protection commitments. ### Replicate — Get the most from Google Veo 3 - Company: Replicate (replicate.com) - Announced: 2025-06-05T16:00:00+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://replicate.com/blog/veo-3 - Record: https://forck.live/items/2489-get-the-most-from-google-veo-3 - Subject: Platform Replicate shares a prompting guide for Google Veo 3, highlighting its features like native audio, accurate dialog, and video game world generation. ### OpenAI — Disrupting malicious uses of AI: June 2025 - Company: OpenAI (openai.com) - Announced: 2025-06-05T02:00:00+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://openai.com/global-affairs/disrupting-malicious-uses-of-ai-june-2025 - Record: https://forck.live/items/543-disrupting-malicious-uses-of-ai-june-2025 - Subject: GPT / ChatGPT / API OpenAI published a report on detecting and preventing malicious uses of AI, featuring case studies. ### Harvey — The Brief: June 2025 - Company: Harvey (harvey.ai) - Announced: 2025-06-04T16:00:00+00:00 - Category: not stated - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://www.harvey.ai/blog/the-brief-june-2025 - Record: https://forck.live/items/17708-the-brief-june-2025 - Subject: Harvey Welcome back to The Brief, where we round up what’s new in Harvey each month. This month’s updates focus on refining how you interact with Harvey. ### Vercel — The no-nonsense approach to AI agent development - Company: Vercel (vercel.com) - Announced: 2025-06-04T13:00:00+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://vercel.com/blog/the-no-nonsense-approach-to-ai-agent-development - Record: https://forck.live/items/20867-the-no-nonsense-approach-to-ai-agent-development - Subject: v0 / AI SDK / AI Gateway Vercel published a guide on building AI agents, emphasizing a three-step approach: manually prototyping the task, automating the loop with code, and optimizing for reliability. The guide advises using LLMs only for steps requiring judgment and relying on deterministic code for parsing, calculation, and sorting. ### Runway — Runway Partners with AMC Networks Across Marketing and TV Development - Company: Runway (runway.com) - Announced: 2025-06-04T10:10:10.110000+00:00 - Category: partnership-acquisition - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://runway.com/news/runway-amc-partnership - Record: https://forck.live/items/7907-runway-partners-with-amc-networks-across-marketing-and-tv-development - Subject: Gen / Aleph Runway and AMC Networks are partnering to incorporate Runway's AI models and tools into AMC Networks' marketing and TV development processes, including visual concept development, marketing workflows, and special effects ideation. ### Genspark — Genspark as Your AI Secretary — Managing Your Google Suite and More! - Company: Genspark (genspark.ai) - Announced: 2025-06-04 - Category: not stated - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://www.genspark.ai/blog/genspark-ai-secretary - Record: https://forck.live/items/18283-genspark-as-your-ai-secretary-managing-your-google-suite-and-more - Subject: Genspark ### LG AI Research — The winners of the LG Awards 2025 ━ How AI Became the Driving Force of Industrial Innovation - Company: LG AI Research (lgresearch.ai) - Announced: 2025-06-04 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://www.lgresearch.ai/blog/view?seq=568 - Record: https://forck.live/items/4786-the-winners-of-the-lg-awards-2025-how-ai-became-the-driving-force-of - Subject: EXAONE LG AI Research's NCC Operations Optimization AI Scheduler Development Project won the Customer Satisfaction Award at the 2025 LG Awards. The project developed an AI-based scheduler for LG Chem's Daesan NCC production plant, achieving over 3% performance improvement over human level and increasing marginal profit by more than KRW 10 billion per year at that plant alone. ### Mistral AI — Introducing Mistral Code - Company: Mistral AI (mistral.ai) - Announced: 2025-06-04 - Category: product-launch - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://mistral.ai/news/mistral-code - Record: https://forck.live/items/3322-introducing-mistral-code - Subject: Mistral / Le Chat Mistral AI announces the launch of a new product called Mistral Code. ### Hugging Face — KV Cache from scratch in nanoVLM - Company: Hugging Face (huggingface.co) - Announced: 2025-06-04 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://huggingface.co/blog/kv-cache - Record: https://forck.live/items/1682-kv-cache-from-scratch-in-nanovlm - Subject: Platform The implementation of KV caching from scratch in the nanoVLM repository, a small pure PyTorch codebase for training Vision Language Models. It reports a 38% speedup in generation and discusses the theory behind KV caching and its practical application. ### Hugging Face — Real-Time AI Sound Generation on Arm: A Personal Tool for Creative Freedom - Company: Hugging Face (huggingface.co) - Announced: 2025-06-03T15:04:36+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://huggingface.co/blog/Arm/ai-sound-gen-on-arm - Record: https://forck.live/items/1683-real-time-ai-sound-generation-on-arm-a-personal-tool-for-creative-freedom - Subject: Platform A software engineer and music producer built a personal sound generation app that runs entirely on an Arm-based CPU using the open-source Stable Audio Open model, enabling real-time audio generation from text prompts without cloud or GPU dependency, and seamlessly integrates with Ableton Live. ### H Company — Holo1: New family of GUI automation VLMs powering GUI agent Surfer-H - Company: H Company (hcompany.ai) - Announced: 2025-06-03T13:27:59+00:00 - Category: new-model - Coverage: not counted - Announcement: yes - Group: models - Source: https://huggingface.co/blog/Hcompany/holo1 - Record: https://forck.live/items/1684-holo1-new-family-of-gui-automation-vlms-powering-gui-agent-surfer-h H Company announces the release of Holo1, a new family of open-source Action Vision Language Models designed for GUI automation, including Holo1-3B and Holo1-7B. The models are released on Hugging Face under a permissive open-source license and are accompanied by the WebClick benchmark. Holo1 models are based on Qwen2.5-VL and achieve high accuracy on UI localization benchmarks. The models power the Surfer-H web agent, which achieves low cost per task. ### LG AI Research — EXAONE Path 1.5 Unveiled at ASCO 2025 ━ Strong On-Site Response Highlights Its Transformative Potential - Company: LG AI Research (lgresearch.ai) - Announced: 2025-06-03 - Category: model-update - Coverage: not counted - Announcement: yes - Group: models - Source: https://www.lgresearch.ai/news/view?seq=570 - Record: https://forck.live/items/11004-exaone-path-1-5-unveiled-at-asco-2025-strong-on-site-response-highlights-its - Subject: EXAONE LG AI Research unveiled EXAONE Path 1.5, an AI model for digital pathology, at ASCO 2025. The model is pretrained on about 73,000 whole-slide images and incorporates genetic data from roughly 10,000 slide-RNA pairs, improving its ability to predict genetic mutations and recommend treatments from pathology images alone. ### Suno — A New Level of Creative Control - Company: Suno (suno.com) - Announced: 2025-06-03 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://suno.com/blog/songeditor - Record: https://forck.live/items/10334-a-new-level-of-creative-control - Subject: Music / Studio Suno launched an upgraded Song Editor with section-level editing, stem extraction into 12 tracks, extended audio uploads up to 8 minutes, and creative sliders for controlling generation style. ### LG AI Research — EXAONE Path 1.5 Unveiled at ASCO 2025 ━ Strong On-Site Response Highlights Its Transformative Potential - Company: LG AI Research (lgresearch.ai) - Announced: 2025-06-03 - Category: model-update - Coverage: not counted - Announcement: yes - Group: models - Source: https://www.lgresearch.ai/blog/view?seq=570 - Record: https://forck.live/items/4787-exaone-path-1-5-unveiled-at-asco-2025-strong-on-site-response-highlights-its - Subject: EXAONE LG AI Research announced EXAONE Path 1.5, an updated AI model for digital pathology, at the ASCO 2025 conference. The model features an expanded training dataset of approximately 73,000 whole-slide images, improved attention-based architecture for whole-slide analysis, and integration of slide-RNA genetic data to predict genetic mutations without conventional testing. ### Hugging Face — No GPU left behind: Unlocking Efficiency with Co-located vLLM in TRL - Company: Hugging Face (huggingface.co) - Announced: 2025-06-03 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://huggingface.co/blog/vllm-colocate - Record: https://forck.live/items/1686-no-gpu-left-behind-unlocking-efficiency-with-co-located-vllm-in-trl - Subject: Platform TRL now supports co-located vLLM, allowing training and inference to share the same GPUs, reducing idle time and improving throughput without extra hardware. ### Hugging Face — SmolVLA: Efficient Vision-Language-Action Model trained on Lerobot Community Data - Company: Hugging Face (huggingface.co) - Announced: 2025-06-03 - Category: new-model - Coverage: not counted - Announcement: yes - Group: models - Source: https://huggingface.co/blog/smolvla - Record: https://forck.live/items/1685-smolvla-efficient-vision-language-action-model-trained-on-lerobot-community-data - Subject: Platform SmolVLA is a compact (450M) open-source Vision-Language-Action model for robotics, trained on community data, that outperforms larger models on simulation and real-world tasks. ### Genspark — Seeing AGI - Company: Genspark (genspark.ai) - Announced: 2025-06-02 - Category: not stated - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://www.genspark.ai/blog/seeing-agi-impact - Record: https://forck.live/items/18284-seeing-agi - Subject: Genspark ### Luma — Introducing Modify Video - Company: Luma (lumalabs.ai) - Announced: 2025-06-02 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://lumalabs.ai/news/introducing-modify-video - Record: https://forck.live/items/7984-introducing-modify-video - Subject: Ray / Dream Machine Luma announced Modify Video, a new feature within Dream Machine: Ray 2 that allows users to reimagine environments, lighting, and texture while preserving motion and performance. The feature includes presets (Adhere, Flex, Reimagine) and supports up to 10 seconds duration. Luma claims it outperforms Runway V2V in retaining motion and temporal consistency. ### Amp — Multi-Root Workspaces - Company: Amp (ampcode.com) - Announced: 2025-06-02 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://ampcode.com/news/multi-root-workspaces - Record: https://forck.live/items/7835-multi-root-workspaces - Subject: Amp Amp has improved support for multi-root workspaces. File operations now work consistently across workspace roots, the root name is shown alongside file paths, and the search agent covers all workspace roots. ### Replicate — FLUX.1 Kontext from the community - Company: Replicate (replicate.com) - Announced: 2025-06-02 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://replicate.com/blog/kontext-roundup - Record: https://forck.live/items/2490-flux-1-kontext-from-the-community - Subject: Platform Replicate announces the community's exploration of FLUX.1 Kontext, a new image editing model from Black Forest Labs, highlighting its capabilities for text-based edits, community examples, and the launch of Kontext Chat and ready-to-use Kontext apps.