# forck.live — 9–15 September 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-W37, 9–15 September 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-W37 - The current issue, and the catalogue of the other weeks: https://forck.live/briefing.md - Catalogue: https://forck.live/llms.txt ## Announcements ### Vercel — Deploying dreams: An inside look at a summer internship with Vercel - Company: Vercel (vercel.com) - Announced: 2024-09-13T13:00:00+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://vercel.com/blog/summer-internship-at-vercel - Record: https://forck.live/items/20951-deploying-dreams-an-inside-look-at-a-summer-internship-with-vercel - Subject: v0 / AI SDK / AI Gateway A Vercel intern describes working on the Support Center UX and the new conversational UI for v0, including implementing citations and rolling out beta invitations. ### LG AI Research — [ACL 2024] AI-generated Text Detection Research Trends - Company: LG AI Research (lgresearch.ai) - Announced: 2024-09-13 - Category: research-paper - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://www.lgresearch.ai/blog/view?seq=482 - Record: https://forck.live/items/4829-acl-2024-ai-generated-text-detection-research-trends - Title in English (translated by forck.live from en): [ACL 2024] AI-generated Text Detection Research Trends - Subject: EXAONE This blog post by LG AI Research summarizes research trends in AI-generated text detection presented at ACL 2024, including a discussion of existing techniques like DetectGPT and a new paper proposing selective perturbation to improve detection. ### Hugging Face — Accelerate 1.0.0 - Company: Hugging Face (huggingface.co) - Announced: 2024-09-13 - Category: developer-tool-release - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://huggingface.co/blog/accelerate-v1 - Record: https://forck.live/items/1832-accelerate-1-0-0 - Subject: Platform Hugging Face announces the first release candidate for Accelerate 1.0.0, a library for simplifying multi-GPU and TPU training in PyTorch, with new features like FP8 support, DeepSpeed multi-model orchestration, and torch.compile integration. ### Harvey — Harvey is building legal agents and workflows with OpenAI o1 - Company: Harvey (harvey.ai) - Announced: 2024-09-12T18:26:00+00:00 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://www.harvey.ai/blog/harvey-building-legal-agents-and-workflows-with-openai-s-o1 - Record: https://forck.live/items/17727-harvey-is-building-legal-agents-and-workflows-with-openai-o1 - Subject: Harvey Harvey is building legal agents and workflows using OpenAI's o1 reasoning model. The model outperforms prior foundation models on Harvey's BigLaw Bench and expert preference evaluations, though it still falls short of Harvey's domain-specialized assistant models. Harvey's step-change review highlights o1's capabilities in reasoning, planning, and collaboration as key enablers for next-generation professional service agents. ### OpenAI — OpenAI o1-mini - Company: OpenAI (openai.com) - Announced: 2024-09-12T10:01:00+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://openai.com/index/openai-o1-mini-advancing-cost-efficient-reasoning - Record: https://forck.live/items/711-openai-o1-mini - Subject: GPT / ChatGPT / API OpenAI is advancing cost-efficient reasoning. ### OpenAI — OpenAI o1 Contributions - Company: OpenAI (openai.com) - Announced: 2024-09-12T10:00:00+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://openai.com/openai-o1-contributions - Record: https://forck.live/items/712-openai-o1-contributions - Subject: GPT / ChatGPT / API OpenAI announces contributions related to o1. ### Supabase — Edge Functions are now 2x smaller and boot 3x faster - Company: Supabase (supabase.com) - Announced: 2024-09-12 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://supabase.com/blog/edge-functions-faster-smaller - Record: https://forck.live/items/21398-edge-functions-are-now-2x-smaller-and-boot-3x-faster - Subject: Supabase Supabase reduced Edge Function bundle sizes by 2x and boot times by 3x for functions using npm modules, achieved by lazy integrity checks, switching to rkyv serialization, and using xxHash-3 instead of SHA-256. Users must redeploy with CLI v1.192.5 or later to benefit. ### Suno — Introducing Covers - Company: Suno (suno.com) - Announced: 2024-09-12 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://suno.com/blog/covers - Record: https://forck.live/items/10343-introducing-covers - Subject: Music / Studio Suno introduced Covers, a new feature in early-access beta for Pro and Premier subscribers that allows users to upload any audio and transform it into a different style while preserving the original melody. Users receive 200 free Covers, after which each generation costs 10 credits. ### OpenAI — Economics and reasoning with OpenAI o1 - Company: OpenAI (openai.com) - Announced: 2024-09-12 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://openai.com/index/o1-economics - Record: https://forck.live/items/716-economics-and-reasoning-with-openai-o1 - Subject: GPT / ChatGPT / API Tyler Cowen explains how OpenAI o1 tackles complex economic questions. ### OpenAI — Answering quantum physics questions with OpenAI o1 - Company: OpenAI (openai.com) - Announced: 2024-09-12 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://openai.com/index/o1-quantum-physics - Record: https://forck.live/items/715-answering-quantum-physics-questions-with-openai-o1 - Subject: GPT / ChatGPT / API OpenAI highlights quantum physicist Mario Krenn using OpenAI o1 to answer quantum physics questions. ### OpenAI — Coding with OpenAI o1 - Company: OpenAI (openai.com) - Announced: 2024-09-12 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://openai.com/index/o1-coding - Record: https://forck.live/items/714-coding-with-openai-o1 - Subject: GPT / ChatGPT / API The post features Scott Wu, CEO and Co-Founder of Cognition, explaining how OpenAI o1 makes coding decisions in a more human-like way. ### OpenAI — Decoding genetics with OpenAI o1 - Company: OpenAI (openai.com) - Announced: 2024-09-12 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://openai.com/index/o1-genetics - Record: https://forck.live/items/713-decoding-genetics-with-openai-o1 - Subject: GPT / ChatGPT / API Geneticist Catherine Brownstein demonstrates how OpenAI o1 can speed up diagnosing rare medical challenges. ### Upstage — Upstage to Release Preview of Next-Generation LLM ‘Solar Pro’ - Company: Upstage (upstage.ai) - Announced: 2024-09-11 - Category: new-model - Coverage: not counted - Announcement: yes - Group: models - Source: https://www.upstage.ai/news/solar-pro-preview - Record: https://forck.live/items/18439-upstage-to-release-preview-of-next-generation-llm-solar-pro - Subject: Solar Pro / Solar Mini Upstage released a preview of Solar Pro, its next-generation flagship LLM with 22 billion parameters, available as an open-source model with free API access ahead of an official release in November. The model is optimized to run on a single GPU via Upstage's Depth-Up Scaling method and advanced data recipe, and it shows an average 51% improvement over Solar Mini on key LLM benchmarks, scoring 52.11 on MMLU Pro and 84.37 on IFEval. ### Runway — Pioneering New Interfaces in the Age of Generative Media - Company: Runway (runway.com) - Announced: 2024-09-10T16:22:35.195000+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://runway.com/research/pioneering-new-interfaces-age-generative-media - Record: https://forck.live/items/7925-pioneering-new-interfaces-in-the-age-of-generative-media - Subject: Gen / Aleph Runway shares three prototype interfaces for generative media and outlines design principles (wonder, discovery, control, feedback) for future creative tools. The post argues that text prompts are insufficient and that new interfaces are needed for expressive interaction with AI. ### Upstage — Small models, big impact: How Upstage built a highly accurate SLM for proofreading - Company: Upstage (upstage.ai) - Announced: 2024-09-10 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://www.upstage.ai/blog/en/small-models-big-impact-how-upstage-built-a-highly-accurate-slm-for-proofreading - Record: https://forck.live/items/18479-small-models-big-impact-how-upstage-built-a-highly-accurate-slm-for-proofreading - Subject: Solar Pro / Solar Mini Upstage used Predibase to fine-tune its Solar-Mini small language model into Solar-Proofread, a custom model for proofreading. In evaluations, Solar-Proofread achieved 79% accuracy in identifying errors, compared to 71% for a fine-tuned GPT-4o mini and 25% for the base GPT-4o mini. The model was built for a major international media company that needed to proofread over 1,800 articles per month. ### Upstage — Solar Pro Preview: The most intelligent LLM on a single GPU - Company: Upstage (upstage.ai) - Announced: 2024-09-10 - Category: open-weight-release - Coverage: not counted - Announcement: yes - Group: models - Source: https://www.upstage.ai/blog/en/solar-pro-preview-the-most-intelligent-llm-on-a-single-gpu - Record: https://forck.live/items/18478-solar-pro-preview-the-most-intelligent-llm-on-a-single-gpu - Subject: Solar Pro / Solar Mini Upstage released Solar Pro Preview, an open-source language model designed to run on a single GPU. The company reports that among models deployable on a single GPU, Solar Pro Preview shows superior performance on MMLU-Pro and IFEval benchmarks. The preview is limited to English and a 4k context window, with the full Solar Pro release planned for November 2024 featuring expanded language support and longer context windows. ### Suno — Suno Remix Contest (feat. Flosstradamus) - Company: Suno (suno.com) - Announced: 2024-09-10 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://suno.com/blog/floss - Record: https://forck.live/items/10344-suno-remix-contest-feat-flosstradamus - Subject: Music / Studio Suno announced its first remix contest in partnership with Flosstradamus, offering over $100,000 in cash prizes for remixes of the track Soundclash, with entries based on stems uploaded to Suno's platform. ### LG AI Research — [ACL 2024] Towards Efficient Large Language Models - Company: LG AI Research (lgresearch.ai) - Announced: 2024-09-10 - Category: research-paper - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://www.lgresearch.ai/blog/view?seq=478 - Record: https://forck.live/items/4830-acl-2024-towards-efficient-large-language-models - Subject: EXAONE LG AI Research blog post summarizing the LayerSkip method from Meta presented at ACL 2024, which uses layer dropout, early exit loss, and self-speculative decoding to speed up LLM inference. ### Replicate — Fine-tune FLUX.1 with an API - Company: Replicate (replicate.com) - Announced: 2024-09-09 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://replicate.com/blog/fine-tune-flux-with-an-api - Record: https://forck.live/items/2514-fine-tune-flux-1-with-an-api - Subject: Platform Replicate announced the ability to fine-tune FLUX.1 models via an API, with training under 2 minutes and under $2, and provided a guide.