# forck.live — 23–29 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-W26, 23–29 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-W26 - The current issue, and the catalogue of the other weeks: https://forck.live/briefing.md - Catalogue: https://forck.live/llms.txt ## Announcements ### Genspark — Seeing AGI (4) - Vibe Working - Company: Genspark (genspark.ai) - Announced: 2025-06-28 - Category: not stated - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://www.genspark.ai/blog/seeing-agi-impact4 - Record: https://forck.live/items/18278-seeing-agi-4-vibe-working - Subject: Genspark ### Harvey — Harvey's First Brand Campaign - Company: Harvey (harvey.ai) - Announced: 2025-06-27T22:23:00+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://www.harvey.ai/blog/harveys-first-ooh-launch - Record: https://forck.live/items/17703-harvey-s-first-brand-campaign - Subject: Harvey Harvey launched its first out-of-home brand campaign in New York, featuring ads on streets and in the subway. The creative is based on brand foundation work by the founders and focuses on early customers, a challenge to embrace change, and the promise to 'Set a New Precedent.' ### NVIDIA — Welcome the NVIDIA Llama Nemotron Nano VLM to Hugging Face Hub - Company: NVIDIA (nvidia.com) - Announced: 2025-06-27T21:09:27+00:00 - Category: new-model - Coverage: not counted - Announcement: yes - Group: models - Source: https://huggingface.co/blog/nvidia/llama-nemotron-nano-vl - Record: https://forck.live/items/1671-welcome-the-nvidia-llama-nemotron-nano-vlm-to-hugging-face-hub NVIDIA announces Llama Nemotron Nano VL, an 8B vision language model optimized for intelligent document processing with high accuracy OCR and multimodal understanding. Built on Llama-3.1-8B-Instruct and C-RADIOv2-VLM-H, it excels in text recognition, table extraction, and element parsing on complex documents. ### Alibaba — Time to Speak Some Dialects, Qwen-TTS! - Company: Alibaba (alibaba.com) - Announced: 2025-06-27T07:01:30+00:00 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://qwenlm.github.io/blog/qwen-tts/ - Record: https://forck.live/items/2287-time-to-speak-some-dialects-qwen-tts - Subject: Qwen Qwen-TTS now supports generating speech in three Chinese dialects: Pekingese, Shanghainese, and Sichuanese, and offers seven Chinese-English bilingual voices. ### Amp — Better, Faster, Cheaper Summaries - Company: Amp (ampcode.com) - Announced: 2025-06-27 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://ampcode.com/news/better-faster-cheaper-summaries - Record: https://forck.live/items/7831-better-faster-cheaper-summaries - Subject: Amp Amp switched to a different model for compacting or summarizing threads, resulting in 4-6x faster performance, roughly 30x lower cost, and improved summary quality. ### Cognition — Coding Agents 101: The Art of Actually Getting Things Done - Company: Cognition (cognition.com) - Announced: 2025-06-27 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://cognition.com/blog/coding-agents-101-the-art-of-actually-getting-things-done - Record: https://forck.live/items/7698-coding-agents-101-the-art-of-actually-getting-things-done - Subject: Devin / SWE Cognition shares top lessons from customers and internal experience on using coding agents effectively. ### Harvey — Time100 Most Influential Companies - Company: Harvey (harvey.ai) - Announced: 2025-06-26T15:59:00+00:00 - Category: funding-company-news - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://www.harvey.ai/blog/time-100-most-influential-companies-2025 - Record: https://forck.live/items/17704-time100-most-influential-companies - Subject: Harvey Harvey was named one of the Time100 Most Influential Companies by Time magazine. ### Alibaba — Qwen VLo: From "Understanding" the World to "Depicting" It - Company: Alibaba (alibaba.com) - Announced: 2025-06-26T14:00:04+00:00 - Category: new-model - Coverage: not counted - Announcement: yes - Group: models - Source: https://qwenlm.github.io/blog/qwen-vlo/ - Record: https://forck.live/items/2288-qwen-vlo-from-understanding-the-world-to-depicting-it - Subject: Qwen Alibaba introduces Qwen VLo, a unified multimodal understanding and generation model that can both understand images and generate high-quality recreations. ### Vercel — Vercel Ship 2025 recap - Company: Vercel (vercel.com) - Announced: 2025-06-26T13:00:00+00:00 - Category: capability-change - Coverage: not counted - Announcement: no - Group: routine - Source: https://vercel.com/blog/vercel-ship-2025-recap - Record: https://forck.live/items/20857-vercel-ship-2025-recap - Subject: v0 / AI SDK / AI Gateway Vercel announced several platform updates at Ship 2025, including AI Gateway for unified model access across providers, Fluid compute with Active CPU pricing that charges only for actual execution time, Vercel Sandbox for running untrusted code in isolated microVMs, and Rolling Releases for safe incremental deployments. The default execution time for all projects using Fluid compute increased to 300 seconds. ### Runway — Parsons School of Design Develops New Course Powered by Runway - Company: Runway (runway.com) - Announced: 2025-06-26T10:10:10.110000+00:00 - Category: partnership-acquisition - Coverage: not counted - Announcement: no - Group: routine - Source: https://runway.com/news/parsons-school-design-develops-new-course-powered-by-runway - Record: https://forck.live/items/7906-parsons-school-of-design-develops-new-course-powered-by-runway - Subject: Gen / Aleph Parsons School of Design will offer a new course, 'Hopepunk and Runway,' during the Fall 2025 semester, giving students access to Runway's tools to build longer-form creative projects. The course is open to all New School students with priority to Parsons students, and no prior experience with Runway is required. ### OpenAI — Customizable, no-code voice agent automation with GPT-4o - Company: OpenAI (openai.com) - Announced: 2025-06-26T10:00:00+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://openai.com/index/retell-ai - Record: https://forck.live/items/535-customizable-no-code-voice-agent-automation-with-gpt-4o - Subject: GPT / ChatGPT / API Retell AI is using GPT-4o and GPT-4.1 to provide no-code voice agent automation for call centers, reducing costs and improving customer satisfaction. ### Amp — Queued Messages - Company: Amp (ampcode.com) - Announced: 2025-06-26 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://ampcode.com/news/queued-messages - Record: https://forck.live/items/7832-queued-messages - Subject: Amp Amp added the ability to queue messages that are delivered to an agent only when it becomes idle, activated by holding shift while submitting. ### Hugging Face — Gemma 3n fully available in the open-source ecosystem! - Company: Hugging Face (huggingface.co) - Announced: 2025-06-26 - Category: new-model - Coverage: not counted - Announcement: yes - Group: models - Source: https://huggingface.co/blog/gemma3n - Record: https://forck.live/items/1672-gemma-3n-fully-available-in-the-open-source-ecosystem - Subject: Platform Gemma 3n, a natively multimodal on-device model supporting image, text, audio, and video inputs, is now fully available in the open-source ecosystem. Two sizes (E2B and E4B) are released with base and instruct variants, achieving competitive performance (e.g., E4B scores 1300+ on LMArena) and efficient memory usage (E2B runs in ~2GB GPU RAM, E4B in ~3GB). The models are supported by transformers, timm, MLX, llama.cpp, and other libraries. ### Vercel — Introducing Active CPU pricing for Fluid compute - Company: Vercel (vercel.com) - Announced: 2025-06-25T13:00:00+00:00 - Category: pricing-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://vercel.com/blog/introducing-active-cpu-pricing-for-fluid-compute - Record: https://forck.live/items/20859-introducing-active-cpu-pricing-for-fluid-compute - Subject: v0 / AI SDK / AI Gateway Vercel introduced Active CPU pricing for its Fluid compute platform, charging for CPU only when code is actively executing on a vCPU. The model adds two new metrics—Active CPU and Provisioned Memory—alongside existing invocation billing. Vercel states the change can bring additional cost savings of up to 90% for workloads with high idle time, such as AI inference, and is enabled by default for Hobby, Pro, and new Enterprise teams. ### Vercel — ​Introducing BotID, invisible bot filtering for critical routes - Company: Vercel (vercel.com) - Announced: 2025-06-25T13:00:00+00:00 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://vercel.com/blog/introducing-botid - Record: https://forck.live/items/20858-introducing-botid-invisible-bot-filtering-for-critical-routes - Subject: v0 / AI SDK / AI Gateway Vercel introduced BotID, a new protection layer that blocks browser automation tools like Playwright and Puppeteer from reaching critical routes such as checkouts, logins, signups, and LLM-powered endpoints. BotID is available in Basic (enabled by default) and Deep Analysis modes, with Deep Analysis powered by Kasada available to Pro and Enterprise teams. Setup requires no configuration files or API keys, and server-side verification uses a single function call. ### Supabase — Build a Personalized AI Assistant with Postgres - Company: Supabase (supabase.com) - Announced: 2025-06-25 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://supabase.com/blog/natural-db - Record: https://forck.live/items/21354-build-a-personalized-ai-assistant-with-postgres - Subject: Supabase Supabase's guide explains how to build a personalized AI assistant using Postgres, combining an LLM-controlled database schema, scheduled tasks via pg_cron, web search, and Zapier MCP integrations for long-term memory and autonomous actions. The system stores structured data in a dedicated memories schema with restricted permissions, uses three memory types (message history, semantic vector search, and structured SQL), and supports evolving user preferences through a self-evolving system prompt. ### Upstage — Upstage to Power Its LLM on FuriosaAI’s Next-Gen AI Chip - Company: Upstage (upstage.ai) - Announced: 2025-06-25 - Category: partnership-acquisition - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://www.upstage.ai/news/upstage-furiosa-llm - Record: https://forck.live/items/18432-upstage-to-power-its-llm-on-furiosaai-s-next-gen-ai-chip - Subject: Solar Pro / Solar Mini Upstage signed an MOU with FuriosaAI to optimize its Solar LLM for FuriosaAI's next-generation AI inference chip, RNGD. The companies also plan to co-develop on-premise generative AI infrastructure for Korean and international markets. ### Vercel — WPP and Vercel: Bringing AI to the creative process - Company: Vercel (vercel.com) - Announced: 2025-06-24T13:00:00+00:00 - Category: partnership-acquisition - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://vercel.com/blog/wpp-and-vercel-bringing-ai-to-the-creative-process - Record: https://forck.live/items/20860-wpp-and-vercel-bringing-ai-to-the-creative-process - Subject: v0 / AI SDK / AI Gateway Vercel expanded its partnership with WPP, making v0 and AI SDK available to WPP's global creative teams. WPP internal pilots showed up to 25% faster digital production times. The partnership also includes priority access to new Vercel product releases and joint investment in product development within WPP Open. ### Harvey — Introducing Agent Builder: Previously Known as Workflow Builder - Company: Harvey (harvey.ai) - Announced: 2025-06-24T07:00:00+00:00 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://www.harvey.ai/blog/introducing-workflow-builder - Record: https://forck.live/items/17705-introducing-agent-builder-previously-known-as-workflow-builder - Subject: Harvey Harvey announced that its Agent Builder (previously called Workflow Builder) is now available to all Harvey customers. The tool allows legal teams to design and deploy custom AI workflows that incorporate their firm's proprietary knowledge, processes, and expertise. It was developed in partnership with several law firms and legal departments. ### OpenAI — Driving scalable growth with OpenAI o3, GPT-4.1, and CUA - Company: OpenAI (openai.com) - Announced: 2025-06-24 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://openai.com/index/unify - Record: https://forck.live/items/536-driving-scalable-growth-with-openai-o3-gpt-4-1-and-cua - Subject: GPT / ChatGPT / API The AI-powered GTM platform Unify uses OpenAI's o3, GPT-4.1, and CUA to automate prospecting, research, and outreach. ### Harvey — Harvey Raises $300M Series E Co-led by Kleiner Perkins and Coatue - Company: Harvey (harvey.ai) - Announced: 2025-06-23T14:00:00+00:00 - Category: funding-company-news - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://www.harvey.ai/blog/harvey-raises-series-e - Record: https://forck.live/items/17706-harvey-raises-300m-series-e-co-led-by-kleiner-perkins-and-coatue - Subject: Harvey Harvey announced a $300 million Series E funding round at a $5 billion valuation, co-led by Kleiner Perkins and Coatue, with participation from Sequoia, GV, DST Global, Conviction, Elad Gil, Open AI Startup Fund, Elemental, SV Angel, Kris Fredrickson, and REV. ### Vercel — Keith Messick joins Vercel as CMO - Company: Vercel (vercel.com) - Announced: 2025-06-23T13:00:00+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://vercel.com/blog/keith-messick-joins-vercel-as-cmo - Record: https://forck.live/items/20861-keith-messick-joins-vercel-as-cmo - Subject: v0 / AI SDK / AI Gateway Vercel appointed Keith Messick as its first Chief Marketing Officer. Messick previously served as CMO at Redis and led marketing at LaunchDarkly, DialPad, and Lucidworks. ### Genspark — Seeing AGI (3) - Time To Build AI Native Team - Company: Genspark (genspark.ai) - Announced: 2025-06-23 - Category: not stated - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://www.genspark.ai/blog/seeing-agi-impact3 - Record: https://forck.live/items/18279-seeing-agi-3-time-to-build-ai-native-team - Subject: Genspark ### Cua — Sandboxed Python Execution: Run Code Safely in Cua Containers - Company: Cua (cua.ai) - Announced: 2025-06-23 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://cua.ai/blog/sandboxed-python-execution - Record: https://forck.live/items/16838-sandboxed-python-execution-run-code-safely-in-cua-containers - Subject: Cua Driver / Lume / Cua Bench / Cua Fleets Cua introduced a @sandboxed decorator that allows AI agents to run Python code inside the same virtual environment as their UI automation, enabling data processing, API calls, and complex logic alongside desktop interactions. ### Cognition — Blockdiff: How we built our own file format for VM disk snapshots - Company: Cognition (cognition.com) - Announced: 2025-06-23 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://cognition.com/blog/blockdiff - Record: https://forck.live/items/7699-blockdiff-how-we-built-our-own-file-format-for-vm-disk-snapshots - Subject: Devin / SWE Cognition open-sourced Blockdiff, a file format for instant block-level diffs of VM disks, built for their VM hypervisor Otterlink. It enables fast snapshot creation (a couple of seconds) compared to EC2's 30+ minutes. ### Hugging Face — Transformers backend integration in SGLang - Company: Hugging Face (huggingface.co) - Announced: 2025-06-23 - Category: developer-tool-release - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://huggingface.co/blog/transformers-backend-sglang - Record: https://forck.live/items/1673-transformers-backend-integration-in-sglang - Subject: Platform SGLang now supports Hugging Face Transformers as a backend, allowing any Transformers-compatible model to run with high-performance inference out of the box, with automatic fallback when a model is not natively supported.