# forck.live — 25 November – 1 December 2024 > 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 2024-W48, 25 November – 1 December 2024, 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/2024-W48 - The current issue, and the catalogue of the other weeks: https://forck.live/briefing.md - Catalogue: https://forck.live/llms.txt ## Announcements ### Supabase — OrioleDB Public Alpha - Company: Supabase (supabase.com) - Announced: 2024-12-01 - Category: availability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://supabase.com/blog/orioledb-launch - Record: https://forck.live/items/21390-orioledb-public-alpha - Subject: Supabase Supabase released the Public Alpha of OrioleDB, a Postgres storage extension designed as a drop-in replacement for Heap storage. The alpha is restricted to Free organizations and does not support production workloads or non-B-Tree indexes. ### LG AI Research — AI at the Center of the Global Human Rights Discussions, at the UN Forum on Business and Human Rights - Company: LG AI Research (lgresearch.ai) - Announced: 2024-11-28 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://www.lgresearch.ai/news/view?seq=505 - Record: https://forck.live/items/11016-ai-at-the-center-of-the-global-human-rights-discussions-at-the-un-forum-on - Subject: EXAONE LG AI Research participated in the UN Forum on Business and Human Rights, sharing its AI ethics practices and impact assessment methodology. The company highlighted its AI Ethical Impact Assessments, which are conducted for all research projects, and called for unified global AI norms from international organizations. ### LG AI Research — AI at the Center of the Global Human Rights Discussions, at the UN Forum on Business and Human Rights - Company: LG AI Research (lgresearch.ai) - Announced: 2024-11-28 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://www.lgresearch.ai/blog/view?seq=505 - Record: https://forck.live/items/4817-ai-at-the-center-of-the-global-human-rights-discussions-at-the-un-forum-on - Subject: EXAONE LG AI Research participated in the UN Forum on Business and Human Rights as a company representative, sharing its experiences and expertise in AI ethics and impact assessments. ### Harvey — Lawyers at Harvey: Applied Legal Research - Company: Harvey (harvey.ai) - Announced: 2024-11-27T18:00:00+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://www.harvey.ai/blog/lawyers-at-harvey-applied-legal-research - Record: https://forck.live/items/17718-lawyers-at-harvey-applied-legal-research - Subject: Harvey Harvey employs applied legal researchers (ALRs), former lawyers from top law firms with no prior AI experience, to collaborate with AI and product teams. ALRs define training data, evaluate model outputs, and improve reasoning for legal tasks. The team has grown to nearly a dozen members supporting functions from foundational AI research to client relationships. ### Alibaba — QwQ: Reflect Deeply on the Boundaries of the Unknown - Company: Alibaba (alibaba.com) - Announced: 2024-11-27T16:00:03+00:00 - Category: new-model - Coverage: not counted - Announcement: yes - Group: models - Source: https://qwenlm.github.io/blog/qwq-32b-preview/ - Record: https://forck.live/items/2302-qwq-reflect-deeply-on-the-boundaries-of-the-unknown - Subject: Qwen Alibaba announced QwQ, a new AI model that approaches problems with genuine wonder and doubt, embodying philosophical curiosity. ### Vercel — Retailer sees $10M increase in sales on Vercel - Company: Vercel (vercel.com) - Announced: 2024-11-27T13:00:00+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://vercel.com/blog/retailer-sees-10m-increase-in-sales-on-vercel - Record: https://forck.live/items/20918-retailer-sees-10m-increase-in-sales-on-vercel - Subject: v0 / AI SDK / AI Gateway A global sportswear retailer migrated from Salesforce Commerce Cloud to Vercel, achieving a 33% increase in average orders per minute, a 500ms reduction in Time to First Byte, and a $10 million revenue increase during the holiday weekend, with $48 million in sales compared to $38 million the prior year. The retailer experienced its first Black Friday without site downtime. ### LG AI Research — [AI Ethics Seminar EP.4] Building a Responsible AI Society through AI Agents - Company: LG AI Research (lgresearch.ai) - Announced: 2024-11-27 - Category: research-paper - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://www.lgresearch.ai/blog/view?seq=503 - Record: https://forck.live/items/4818-ai-ethics-seminar-ep-4-building-a-responsible-ai-society-through-ai-agents - Subject: EXAONE LG AI Research's AI Ethics Seminar discussed the current state and future of AI agents, focusing on a research paper that tested cooperation among LLM agents (GPT-3.5, GPT-4, Claude-3, Llama-3) using the GovSim simulation. The study found that agents failed to maintain sustainable cooperation but that applying a universalizing law prompt improved survival time and resource usage. ### Replit — So you suspect you have a memory leak... - Company: Replit (replit.com) - Announced: 2024-11-26T14:28:53+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://replit.com/blog/finding-and-solving-memory-leaks - Record: https://forck.live/items/17987-so-you-suspect-you-have-a-memory-leak - Subject: Replit Agent Replit engineers identified and investigated a memory leak in the Replit Agent processes, which were running out of memory roughly once an hour. They used the memray profiler and Python's gc module to inspect heap allocations and find the cause of the leak. ### Vercel — From minutes to seconds: How Meter accelerates delivery with Vercel and Next.js - Company: Vercel (vercel.com) - Announced: 2024-11-26T13:00:00+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://vercel.com/blog/from-minutes-to-seconds-how-meter-accelerates-delivery-with-vercel-and-next - Record: https://forck.live/items/20919-from-minutes-to-seconds-how-meter-accelerates-delivery-with-vercel-and-next-js - Subject: v0 / AI SDK / AI Gateway Meter, a networking solutions provider, migrated its Dashboard product to Vercel, reducing build times from over 10 minutes to under a minute. The company also built its Command generative UI on Next.js and Vercel, enabling rapid iteration and natural-language network management. The case study documents how Vercel's git workflow and previews improved CI/CD speed and deployment reliability. ### Replicate — FLUX fine-tunes are now fast - Company: Replicate (replicate.com) - Announced: 2024-11-26 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://replicate.com/blog/fast-flux-fine-tunes - Record: https://forck.live/items/2506-flux-fine-tunes-are-now-fast - Subject: Platform Replicate announced that fine-tunes of FLUX models are now significantly faster, matching base model speeds, and the optimizations are open-sourced. The update enables loading LoRAs from Hugging Face, Civitai, and arbitrary URLs, and introduces a go_fast=true flag for full optimizations. ### Hugging Face — SmolVLM - small yet mighty Vision Language Model - Company: Hugging Face (huggingface.co) - Announced: 2024-11-26 - Category: open-weight-release - Coverage: not counted - Announcement: yes - Group: models - Source: https://huggingface.co/blog/smolvlm - Record: https://forck.live/items/1792-smolvlm-small-yet-mighty-vision-language-model - Subject: Platform Hugging Face introduces SmolVLM, a 2B vision language model that is small, fast, memory-efficient, and fully open-source. The release includes three model variants (Base, Synthetic, Instruct), all model checkpoints, VLM datasets, training recipes, and tools under the Apache 2.0 license. ### Hugging Face — Rearchitecting Hugging Face Uploads and Downloads - Company: Hugging Face (huggingface.co) - Announced: 2024-11-26 - Category: infrastructure-release - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://huggingface.co/blog/rearchitecting-uploads-and-downloads - Record: https://forck.live/items/1791-rearchitecting-hugging-face-uploads-and-downloads - Subject: Platform Hugging Face announces a redesign of the Hub's upload and download architecture, introducing a content-addressed store (CAS) for a custom protocol that analyzes files at the byte level to improve transfer speeds and enable deduplication. ### Runway — Introducing Frames - Company: Runway (runway.com) - Announced: 2024-11-25T16:22:35.195000+00:00 - Category: new-model - Coverage: not counted - Announcement: yes - Group: models - Source: https://runway.com/research/introducing-frames - Record: https://forck.live/items/7919-introducing-frames - Subject: Gen / Aleph Runway released Frames, a new base model for image generation that emphasizes stylistic control and visual fidelity. The model is being gradually rolled out to users of Gen-3 Alpha and the Runway API. ### Vercel — How Notion powers rapid and performant experimentation - Company: Vercel (vercel.com) - Announced: 2024-11-25T13:00:00+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://vercel.com/blog/how-notion-powers-rapid-and-performant-experimentation - Record: https://forck.live/items/20920-how-notion-powers-rapid-and-performant-experimentation - Subject: v0 / AI SDK / AI Gateway Notion's marketing site team adopted Vercel and Next.js, reducing hotfix deployment time from one hour to 15 minutes. They integrated Statsig with Vercel Edge Config for server-side experimentation, achieving an average Cumulative Layout Shift score of 0.09. The team now runs hundreds of experiments per year. ### Hugging Face — You could have designed state of the art positional encoding - Company: Hugging Face (huggingface.co) - Announced: 2024-11-25 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://huggingface.co/blog/designing-positional-encoding - Record: https://forck.live/items/1793-you-could-have-designed-state-of-the-art-positional-encoding - Subject: Platform The step-by-step discovery of state-of-the-art positional encoding in transformer models, iteratively improving the approach to arrive at RoPE, used in Llama 3.2 and modern transformers.