# forck.live — 4–10 August 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-W32, 4–10 August 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-W32 - The current issue, and the catalogue of the other weeks: https://forck.live/briefing.md - Catalogue: https://forck.live/llms.txt ## Announcements ### Replicate — Announcing Replicate's remote MCP server - Company: Replicate (replicate.com) - Announced: 2025-08-10 - Category: developer-tool-release - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://replicate.com/blog/remote-mcp-server - Record: https://forck.live/items/2480-announcing-replicate-s-remote-mcp-server - Subject: Platform Replicate announces a hosted remote MCP server that allows users to interact with its HTTP APIs via natural language in apps like Claude Desktop, Claude Code, Cursor, and VS Code. ### Reflection AI — Building Frontier Open Intelligence - Company: Reflection AI (reflection.ai) - Announced: 2025-08-09 - Category: funding-company-news - Coverage: not counted - Announcement: no - Group: routine - Source: https://reflection.ai/blog/building-frontier-open-intelligencence - Record: https://forck.live/items/18514-building-frontier-open-intelligence - Subject: Reflection Beam Reflection AI announced it has raised $2 billion, assembled a team with experience on PaLM, Gemini, AlphaGo, and other projects, and built a large-scale LLM and reinforcement learning platform for training Mixture-of-Experts models. The company plans to develop open frontier models combining large-scale pretraining and advanced reinforcement learning, aiming to make capable AI openly accessible. ### Vercel — How Zapier scales product partnerships with v0 - Company: Vercel (vercel.com) - Announced: 2025-08-08T13:00:00+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://vercel.com/blog/how-zapier-scales-product-partnerships-with-v0 - Record: https://forck.live/items/20841-how-zapier-scales-product-partnerships-with-v0 - Subject: v0 / AI SDK / AI Gateway Zapier's Powered by Zapier team uses Vercel's v0 to generate high-fidelity, interactive prototypes of embedded integration experiences in under two hours, replacing a process that previously took weeks. This has enabled the team to deliver 100x more partner demos and save over 12 hours of resources per prototype. The team is also exploring v0's Platform API to automate prototype generation based on inbound partner interest. ### Z.ai — GLM Slide/Poster Agent(beta) - Company: Z.ai (z.ai) - Announced: 2025-08-08 - Category: product-launch - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://docs.z.ai/release-notes/new-released#2025-08-08 - Record: https://forck.live/items/4461-glm-slide-poster-agent-beta - Subject: GLM Z.ai launched GLM Slide/Poster Agent, an AI-powered creation agent that generates professional-grade slides and posters from natural language instructions, combining information retrieval, content structuring, and visual layout design. ### Hugging Face — Accelerate ND-Parallel: A guide to Efficient Multi-GPU Training - Company: Hugging Face (huggingface.co) - Announced: 2025-08-08 - Category: capability-change - Coverage: not counted - Announcement: no - Group: routine - Source: https://huggingface.co/blog/accelerate-nd-parallel - Record: https://forck.live/items/1641-accelerate-nd-parallel-a-guide-to-efficient-multi-gpu-training - Subject: Platform Hugging Face announces integration of ND-Parallel (multi-dimensional parallelism) into Accelerate and Axolotl, allowing users to easily combine data, tensor, context, and fully sharded data parallelism strategies for efficient multi-GPU training. ### Hugging Face — Introducing AI Sheets: a tool to work with datasets using open AI models! - Company: Hugging Face (huggingface.co) - Announced: 2025-08-08 - Category: developer-tool-release - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://huggingface.co/blog/aisheets - Record: https://forck.live/items/1640-introducing-ai-sheets-a-tool-to-work-with-datasets-using-open-ai-models - Subject: Platform Hugging Face announces AI Sheets, an open-source no-code tool for building, enriching, and transforming datasets using AI models. It integrates with the Hugging Face Hub and supports thousands of open models, including gpt-oss from OpenAI. Users can deploy locally or on the Hub. ### Lovable — GPT-5 Meets Lovable - Company: Lovable (lovable.dev) - Announced: 2025-08-07T18:23:00+00:00 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://lovable.dev/blog/gpt5-meets-lovable - Record: https://forck.live/items/12136-gpt-5-meets-lovable - Subject: Lovable / AI app builder Lovable has integrated GPT-5 into its platform, making it available in a limited preview until August 12. The company also announced a future initiative with OpenAI to provide free access to Lovable for student builders starting this fall. ### Harvey — Building a Legal Coworker with GPT-5 - Company: Harvey (harvey.ai) - Announced: 2025-08-07T17:00:00+00:00 - Category: capability-change - Coverage: not counted - Announcement: no - Group: routine - Source: https://www.harvey.ai/blog/building-a-legal-coworker-with-gpt-5 - Record: https://forck.live/items/17688-building-a-legal-coworker-with-gpt-5 - Subject: Harvey Harvey announced that GPT-5 is live on its platform via model selector, with plans to make it the default model where it outperforms others. The model showed a 5.09% improvement over OpenAI's previous models on Harvey's BigLaw Bench evaluation, and is described as the first orchestration model capable of combining research and drafting tasks in a single agent. ### Vercel — Vercel is the only vendor to be recognized as a Visionary in the 2025 Gartner® Magic Quadrant™ for Cloud-Native Application Platforms - Company: Vercel (vercel.com) - Announced: 2025-08-07T13:00:00+00:00 - Category: funding-company-news - Coverage: not counted - Announcement: no - Group: routine - Source: https://vercel.com/blog/gartner-mq-visionary-2025 - Record: https://forck.live/items/20843-vercel-is-the-only-vendor-to-be-recognized-as-a-visionary-in-the-2025-gartner - Subject: v0 / AI SDK / AI Gateway Vercel was named a Visionary in the 2025 Gartner Magic Quadrant for Cloud-Native Application Platforms, the only vendor to receive that designation. The company attributes the recognition to its AI Cloud platform, which serves over 4 million weekly active domains and 115 billion weekly requests. ### Vercel — Vercel collaborates with OpenAI for GPT-5 launch - Company: Vercel (vercel.com) - Announced: 2025-08-07T13:00:00+00:00 - Category: availability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://vercel.com/blog/vercel-collaborates-with-openai-for-gpt-5-launch - Record: https://forck.live/items/20842-vercel-collaborates-with-openai-for-gpt-5-launch - Subject: v0 / AI SDK / AI Gateway Vercel announced that OpenAI's GPT-5 family of models is now available through its AI Gateway and in production on v0.dev. Vercel tested the models over the past few weeks and found GPT-5 noticeably better at frontend design, with strong performance in agent-based workflows. The smaller GPT-5 mini and nano variants are already used for speed-sensitive codegen workloads. ### Harvey — Harvey to Expand Team with New Toronto Office - Company: Harvey (harvey.ai) - Announced: 2025-08-07T11:00:00+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://www.harvey.ai/blog/harvey-to-expand-team-with-new-toronto-office - Record: https://forck.live/items/17689-harvey-to-expand-team-with-new-toronto-office - Subject: Harvey Harvey announced it will open a new office in Toronto, Canada, in October to support its sales and engineering teams as part of its global expansion strategy. The office aims to tap into local engineering talent and deepen collaboration with Canadian law firms. ### OpenAI — GPT-5 and the new era of work - Company: OpenAI (openai.com) - Announced: 2025-08-07T10:00:00+00:00 - Category: new-model - Coverage: not counted - Announcement: yes - Group: models - Source: https://openai.com/index/gpt-5-new-era-of-work - Record: https://forck.live/items/496-gpt-5-and-the-new-era-of-work - Subject: GPT / ChatGPT / API OpenAI announces GPT-5 as its most advanced model, focusing on enterprise AI, automation, and workforce productivity. ### OpenAI — Introducing GPT-5 for developers - Company: OpenAI (openai.com) - Announced: 2025-08-07T10:00:00+00:00 - Category: new-model - Coverage: not counted - Announcement: yes - Group: models - Source: https://openai.com/index/introducing-gpt-5-for-developers - Record: https://forck.live/items/495-introducing-gpt-5-for-developers - Subject: GPT / ChatGPT / API OpenAI announces GPT-5 in its API platform, highlighting high reasoning performance, new developer controls, and best-in-class coding results. ### OpenAI — Coding and design with GPT-5 - Company: OpenAI (openai.com) - Announced: 2025-08-07T00:03:00+00:00 - Category: new-model - Coverage: not counted - Announcement: yes - Group: models - Source: https://openai.com/index/gpt-5-coding-design - Record: https://forck.live/items/497-coding-and-design-with-gpt-5 - Subject: GPT / ChatGPT / API GPT-5 enables new capabilities in coding and design. ### OpenAI — Creative writing with GPT-5 - Company: OpenAI (openai.com) - Announced: 2025-08-07T00:02:00+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://openai.com/index/gpt-5-creative-writing - Record: https://forck.live/items/498-creative-writing-with-gpt-5 - Subject: GPT / ChatGPT / API The source text is a brief description of a guide on using GPT-5 for creative writing, but it does not announce any new model, product, or change. ### OpenAI — Medical research with GPT-5 - Company: OpenAI (openai.com) - Announced: 2025-08-07T00:01:00+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://openai.com/index/gpt-5-medical-research - Record: https://forck.live/items/499-medical-research-with-gpt-5 - Subject: GPT / ChatGPT / API GPT-5 is used for medical research. ### Amp — Tool-Level Permissions - Company: Amp (ampcode.com) - Announced: 2025-08-07 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://ampcode.com/news/tool-level-permissions - Record: https://forck.live/items/7823-tool-level-permissions - Subject: Amp Amp now allows users to define granular tool-level permissions, specifying which tools it may run and with which arguments, and how to handle each call (allow, reject, ask, or delegate). ### Amp — JetBrains in the CLI - Company: Amp (ampcode.com) - Announced: 2025-08-07 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://ampcode.com/news/jetbrains-in-the-cli - Record: https://forck.live/items/7822-jetbrains-in-the-cli - Subject: Amp Amp has released a new integration that connects the Amp CLI to JetBrains IDEs, providing context about the current file and selection and allowing direct file editing through the IDE. Users need to update the Amp CLI and run `amp --jetbrains` from the project root. ### Amp — Try GPT-5 With Us - Company: Amp (ampcode.com) - Announced: 2025-08-07 - Category: new-model - Coverage: not counted - Announcement: yes - Group: models - Source: https://ampcode.com/news/gpt-5 - Record: https://forck.live/items/7821-try-gpt-5-with-us - Subject: Amp Amp is evaluating GPT-5 and offers users the ability to try GPT-5 as the oracle via a command-line flag or editor setting for a limited time. ### Hugging Face — Vision Language Model Alignment in TRL ⚡️ - Company: Hugging Face (huggingface.co) - Announced: 2025-08-07 - Category: developer-tool-release - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://huggingface.co/blog/trl-vlm-alignment - Record: https://forck.live/items/1642-vision-language-model-alignment-in-trl - Subject: Platform Hugging Face announces new alignment methods for Vision Language Models in TRL: Mixed Preference Optimization (MPO), Group Relative Policy Optimization (GRPO), Group Sequence Policy Optimization (GSPO), plus extensions for RLOO and Online DPO, and native SFT support for VLMs. ### OpenAI — How Cursor uses GPT-5 - Company: OpenAI (openai.com) - Announced: 2025-08-07 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://openai.com/index/gpt-5-cursor - Record: https://forck.live/items/505-how-cursor-uses-gpt-5 - Subject: GPT / ChatGPT / API OpenAI published a post about how Cursor uses GPT-5. ### OpenAI — How Amgen uses GPT-5 - Company: OpenAI (openai.com) - Announced: 2025-08-07 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://openai.com/index/gpt-5-amgen - Record: https://forck.live/items/504-how-amgen-uses-gpt-5 - Subject: GPT / ChatGPT / API OpenAI announces a post about Amgen's use of GPT-5. ### OpenAI — Introducing GPT-5 - Company: OpenAI (openai.com) - Announced: 2025-08-07 - Category: new-model - Coverage: not counted - Announcement: yes - Group: models - Source: https://openai.com/index/introducing-gpt-5 - Record: https://forck.live/items/503-introducing-gpt-5 - Subject: GPT / ChatGPT / API OpenAI announces GPT-5, a new flagship model that represents a significant leap in intelligence over previous models, with state-of-the-art performance across coding, math, writing, health, and visual perception. ### OpenAI — First look at GPT-5 - Company: OpenAI (openai.com) - Announced: 2025-08-07 - Category: new-model - Coverage: not counted - Announcement: yes - Group: models - Source: https://openai.com/index/gpt-5-first-look - Record: https://forck.live/items/502-first-look-at-gpt-5 - Subject: GPT / ChatGPT / API OpenAI provides a first look at GPT-5, featuring leading developers using the model for the first time. ### OpenAI — GPT-5 System Card - Company: OpenAI (openai.com) - Announced: 2025-08-07 - Category: research-paper - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://openai.com/index/gpt-5-system-card - Record: https://forck.live/items/501-gpt-5-system-card - Subject: GPT / ChatGPT / API OpenAI publishes the GPT-5 system card, detailing a unified model routing system that uses multiple model variants (gpt-5-main, gpt-5-thinking, gpt-5-thinking-nano) for different tasks and developer use cases. ### OpenAI — From hard refusals to safe-completions: toward output-centric safety training - Company: OpenAI (openai.com) - Announced: 2025-08-07 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://openai.com/index/gpt-5-safe-completions - Record: https://forck.live/items/500-from-hard-refusals-to-safe-completions-toward-output-centric-safety-training - Subject: GPT / ChatGPT / API OpenAI describes a new safe-completions approach in GPT-5 that replaces hard refusals with output-centric safety training, aiming to improve both safety and helpfulness on dual-use prompts. ### Harvey — Harvey’s Outlook Add-In - Company: Harvey (harvey.ai) - Announced: 2025-08-06T13:55:00+00:00 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://www.harvey.ai/blog/harvey-outlook-add-in - Record: https://forck.live/items/17690-harvey-s-outlook-add-in - Subject: Harvey Harvey announced an Outlook Add-In that provides AI-powered drafting and summarization within the email client, aiming to reduce context switching for legal professionals. The add-in is scheduled for early Q4 availability, with beta access in the coming weeks. ### Vercel — Introducing Vercel MCP: Connect Vercel to your AI tools - Company: Vercel (vercel.com) - Announced: 2025-08-06T13:00:00+00:00 - Category: developer-tool-release - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://vercel.com/blog/introducing-vercel-mcp-connect-vercel-to-your-ai-tools - Record: https://forck.live/items/20844-introducing-vercel-mcp-connect-vercel-to-your-ai-tools - Subject: v0 / AI SDK / AI Gateway Vercel launched the official Vercel MCP server in Public Beta, a secure, OAuth-compliant interface that lets AI clients interact with Vercel projects. The server is read-only and supports tools like Cursor, Claude, and VS Code for tasks such as searching docs, retrieving deployment logs, and fetching teams or projects. ### OpenAI — Providing ChatGPT to the Entire U.S. Federal Workforce - Company: OpenAI (openai.com) - Announced: 2025-08-06 - Category: availability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://openai.com/index/providing-chatgpt-to-the-entire-us-federal-workforce - Record: https://forck.live/items/506-providing-chatgpt-to-the-entire-u-s-federal-workforce - Subject: GPT / ChatGPT / API OpenAI announces a partnership with the U.S. General Services Administration to provide ChatGPT Enterprise to the entire federal executive branch workforce for one year at essentially no cost. ### Harvey — The Brief: August 2025 - Company: Harvey (harvey.ai) - Announced: 2025-08-05T14:00:59.678000+00:00 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://www.harvey.ai/blog/the-brief-aug-2025 - Record: https://forck.live/items/17691-the-brief-august-2025 - Subject: Harvey Harvey introduced a unified interface for legal work, a new homepage with personalized suggestions, combined drafting and chat, an in-app document editor, natural-language querying over review tables, and tools for assigning and flagging review cells. The update also includes a Words to Workflows feature that auto-generates workflow drafts from natural language, a Playbook Workflow creator, and expanded Vault and Assistant functionality in the Word Add-In. Harvey is now available in Australia-based Azure regions and added nine new curated knowledge sources for Austria, Finland, Germany, Spain, Netherlands, India, Switzerland, Norway, and France. ### Stability AI — Introducing Stability AI Solutions: Generative AI Solutions to Accelerate Enterprise Creative Production - Company: Stability AI (stability.ai) - Announced: 2025-08-05T13:56:13+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://stability.ai/news-updates/introducing-stability-ai-solutions - Record: https://forck.live/items/17485-introducing-stability-ai-solutions-generative-ai-solutions-to-accelerate - Subject: Stable Diffusion Stability AI launched Stability AI Solutions, a new offering for enterprises to scale creative production with generative AI. The initial suite includes solutions for Product Photography, Brand Style, Product Concepting & Design, and Digital Twins, targeting Marketing, Advertising, and Design verticals. The offering can be deployed on-premises, via secure API endpoints, or through web-based applications, and will also be available through WPP Open. ### Ant Ling — Introducing Ring-lite-2507 - Company: Ant Ling (ant-ling.com) - Announced: 2025-08-05 - Category: model-update - Coverage: not counted - Announcement: yes - Group: models - Source: https://developer.ant-ling.com/en/blogs/ring-lite-2507 - Record: https://forck.live/items/22340-introducing-ring-lite-2507 - Subject: Ling / Ring / Ming Ant Ling released Ring-lite-2507, an upgraded version of its lightweight reasoning model Ring-lite (2506). The model is built on a 16.8B Mixture-of-Experts architecture with 2.75B activated parameters and demonstrates improved performance across general text understanding, alignment, coding, logical, and agentic tasks. The company also introduced a novel training approach called Constrained Contextual Computation Policy Optimization (C3PO) and released the training recipe and model weights publicly. ### Factory AI — GPT-5 is Available in Factory - Company: Factory AI (factory.com) - Announced: 2025-08-05 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://factory.com/news/gpt5 - Record: https://forck.live/items/17853-gpt-5-is-available-in-factory - Subject: Droid Factory has made GPT-5 the default model across all its droids, including Code Droid, Knowledge Droid, and Reliability Droid. The company states that GPT-5 is highly agentic and detail-oriented, with built-in behaviors such as scanning context, forming and executing plans, testing results, and checking completeness. Factory collaborated with OpenAI to integrate GPT-5 into its platform. ### Cua — Bringing Computer-Use to the Web - Company: Cua (cua.ai) - Announced: 2025-08-05 - Category: developer-tool-release - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://cua.ai/blog/bringing-computer-use-to-the-web - Record: https://forck.live/items/16837-bringing-computer-use-to-the-web - Subject: Cua Driver / Lume / Cua Bench / Cua Fleets Cua released the @trycua/computer Web SDK, a library for controlling Cua cloud containers from JavaScript or TypeScript projects. The SDK supports clicking, typing, and screenshot capture without extra servers, and works with Node.js, React, Vue, or any web framework. It can be used with OpenAI's computer-use-preview model or other models like Anthropic's Computer Use or UI-TARS. ### Together AI — Announcing the Availability of OpenAI's Open Models on Together AI - Company: Together AI (together.ai) - Announced: 2025-08-05 - Category: availability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://www.together.ai/blog/announcing-the-availability-of-openais-open-models-on-together-ai - Record: https://forck.live/items/2417-announcing-the-availability-of-openai-s-open-models-on-together-ai - Subject: Inference platform Together AI announces the availability of OpenAI's new open-weight reasoning models gpt-oss-120B and gpt-oss-20B on its platform, with pricing at $0.15 per 1M input tokens and $0.60 per 1M output tokens, released under Apache 2.0 license. ### Hugging Face — Welcome GPT OSS, the new open-source model family from OpenAI! - Company: Hugging Face (huggingface.co) - Announced: 2025-08-05 - Category: open-weight-release - Coverage: not counted - Announcement: yes - Group: models - Source: https://huggingface.co/blog/welcome-openai-gpt-oss - Record: https://forck.live/items/1643-welcome-gpt-oss-the-new-open-source-model-family-from-openai - Subject: Platform OpenAI releases GPT OSS, an open-weights model family with two mixture-of-experts models (gpt-oss-120b and gpt-oss-20b), licensed under Apache 2.0, featuring a 128K context window and 4-bit quantization for efficient deployment. ### OpenAI — Open Weights and AI for All - Company: OpenAI (openai.com) - Announced: 2025-08-05 - Category: open-weight-release - Coverage: not counted - Announcement: yes - Group: models - Source: https://openai.com/global-affairs/open-weights-and-ai-for-all - Record: https://forck.live/items/510-open-weights-and-ai-for-all - Subject: GPT / ChatGPT / API OpenAI announces the release of their most capable open-weights models, aiming to make advanced AI more open, flexible, and accessible globally. ### OpenAI — gpt-oss-120b & gpt-oss-20b Model Card - Company: OpenAI (openai.com) - Announced: 2025-08-05 - Category: open-weight-release - Coverage: not counted - Announcement: yes - Group: models - Source: https://openai.com/index/gpt-oss-model-card - Record: https://forck.live/items/509-gpt-oss-120b-gpt-oss-20b-model-card - Subject: GPT / ChatGPT / API OpenAI introduces two open-weight reasoning models, gpt-oss-120b and gpt-oss-20b, under Apache 2.0 license and gpt-oss usage policy. ### OpenAI — Introducing gpt-oss - Company: OpenAI (openai.com) - Announced: 2025-08-05 - Category: open-weight-release - Coverage: not counted - Announcement: yes - Group: models - Source: https://openai.com/index/introducing-gpt-oss - Record: https://forck.live/items/508-introducing-gpt-oss - Subject: GPT / ChatGPT / API OpenAI released two open-weight language models, gpt-oss-120b and gpt-oss-20b, under the Apache 2.0 license, claiming strong real-world performance at low cost. ### OpenAI — Estimating worst case frontier risks of open weight LLMs - Company: OpenAI (openai.com) - Announced: 2025-08-05 - Category: research-paper - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://openai.com/index/estimating-worst-case-frontier-risks-of-open-weight-llms - Record: https://forck.live/items/507-estimating-worst-case-frontier-risks-of-open-weight-llms - Subject: GPT / ChatGPT / API OpenAI studies worst-case frontier risks of releasing gpt-oss by introducing malicious fine-tuning to elicit maximum capabilities in biology and cybersecurity. ### Stability AI — Stability AI Achieves SOC 2 Type II and SOC 3 Compliance, Reaching New Industry Standard for Enterprise-Grade Security - Company: Stability AI (stability.ai) - Announced: 2025-08-04T20:03:04+00:00 - Category: safety-policy-update - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://stability.ai/news-updates/stability-ai-achieves-soc-2-type-ii-and-soc-3-compliance - Record: https://forck.live/items/17486-stability-ai-achieves-soc-2-type-ii-and-soc-3-compliance-reaching-new-industry - Subject: Stable Diffusion Stability AI completed SOC 2 Type II and SOC 3 certifications, validating its security controls and data protection practices through independent third-party auditing. The certifications cover five trust service criteria: security, availability, processing integrity, confidentiality, and privacy. Both reports are available in the Stability AI Trust Center Portal. ### NVIDIA — Measuring Open-Source Llama Nemotron Models on DeepResearch Bench - Company: NVIDIA (nvidia.com) - Announced: 2025-08-04T19:51:50+00:00 - Category: research-paper - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://huggingface.co/blog/nvidia/ai-q-top-ranking-open-portable-deep-research-agent - Record: https://forck.live/items/1644-measuring-open-source-llama-nemotron-models-on-deepresearch-bench NVIDIA's AI-Q Blueprint, combining Llama 3.3-70B Instruct and Llama-3.3-Nemotron-Super-49B-v1.5, achieved the top spot on the Hugging Face DeepResearch Bench for open-source LLMs with search, demonstrating strong reasoning, efficiency, and transparency. ### Alibaba — Qwen-Image: Crafting with Native Text Rendering - Company: Alibaba (alibaba.com) - Announced: 2025-08-04T14:08:30+00:00 - Category: new-model - Coverage: not counted - Announcement: yes - Group: models - Source: https://qwenlm.github.io/blog/qwen-image/ - Record: https://forck.live/items/2283-qwen-image-crafting-with-native-text-rendering - Subject: Qwen Alibaba releases Qwen-Image, a 20B MMDiT image foundation model with advanced text rendering and image editing capabilities. ### Harvey — Harvey's Three Year Anniversary - Company: Harvey (harvey.ai) - Announced: 2025-08-04T13:10:00+00:00 - Category: funding-company-news - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://www.harvey.ai/blog/harveys-three-year-anniversary - Record: https://forck.live/items/17692-harvey-s-three-year-anniversary - Subject: Harvey Harvey reported it now serves over 500 customers in 54 countries, has surpassed $100M ARR, and its Weekly Active Users grew 4x in the past year. The company also noted that 42% of AmLaw 100 firms use Harvey, and its team has grown to 350 employees. ### Vercel — v0: vibe coding, securely - Company: Vercel (vercel.com) - Announced: 2025-08-04T13:00:00+00:00 - Category: not stated - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://vercel.com/blog/v0-vibe-coding-securely - Record: https://forck.live/items/20845-v0-vibe-coding-securely - Subject: v0 / AI SDK / AI Gateway When code is written by AI, secure defaults must be the foundation, not a feature Vibe coding has changed how software gets built. Tools like v0 make it possible to turn ideas into working prototypes in seconds. ### OpenAI — What we’re optimizing ChatGPT for - Company: OpenAI (openai.com) - Announced: 2025-08-04 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://openai.com/index/optimizing-chatgpt - Record: https://forck.live/items/511-what-we-re-optimizing-chatgpt-for - Subject: GPT / ChatGPT / API OpenAI is improving ChatGPT to help users thrive, including better support for tough moments, reminders to take breaks, and improved life advice guided by expert input.