# forck.live — 12–18 February 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-W07, 12–18 February 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-W07 - The current issue, and the catalogue of the other weeks: https://forck.live/briefing.md - Catalogue: https://forck.live/llms.txt ## Announcements ### Hugging Face — Synthetic data: save money, time and carbon with open source - Company: Hugging Face (huggingface.co) - Announced: 2024-02-16 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://huggingface.co/blog/synthetic-data-save-costs - Record: https://forck.live/items/1941-synthetic-data-save-money-time-and-carbon-with-open-source - Subject: Platform A method for creating synthetic data using open-source LLMs, such as Mixtral-8x7B-Instruct-v0.1, to fine-tune smaller, more efficient models. It includes a case study on financial sentiment analysis showing significant cost, time, and carbon savings compared to using proprietary LLM APIs like GPT-4. ### Replit — Replit + pip - Company: Replit (replit.com) - Announced: 2024-02-15T17:00:00+00:00 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://replit.com/blog/pip - Record: https://forck.live/items/18004-replit-pip - Subject: Replit Agent Replit announced first-class pip support for its workspace, enabling users to switch from poetry to pip or import repositories from GitHub with out-of-the-box functionality. The change addresses usability issues where pip-installed packages were usable interactively but caused import errors during deployment. Replit built a requirements.txt parser and a custom pip UPM backend to provide a consistent experience. ### OpenAI — Video generation models as world simulators - Company: OpenAI (openai.com) - Announced: 2024-02-15T08:00:00+00:00 - Category: research-paper - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://openai.com/index/video-generation-models-as-world-simulators - Record: https://forck.live/items/807-video-generation-models-as-world-simulators - Subject: GPT / ChatGPT / API OpenAI discusses training video generation models on video and image data, introducing Sora, a model that can generate a minute of high fidelity video. ### LG AI Research — [WACV 2024] Industrial Anomaly Detection and Utilization in the Vision Inspection - Company: LG AI Research (lgresearch.ai) - Announced: 2024-02-15 - Category: research-paper - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://www.lgresearch.ai/blog/view?seq=402 - Record: https://forck.live/items/4865-wacv-2024-industrial-anomaly-detection-and-utilization-in-the-vision-inspection - Subject: EXAONE LG AI Research's Vision Lab presented ReConPatch, a method for industrial anomaly detection, at the WACV 2024 conference. ### Replit — Sharding Infrastructure: The Regional Goval Project - Company: Replit (replit.com) - Announced: 2024-02-14T22:00:00+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://replit.com/blog/regional-goval - Record: https://forck.live/items/18006-sharding-infrastructure-the-regional-goval-project - Subject: Replit Agent Replit's infrastructure team describes the Regional Goval project, which rearchitects the company's cloud infrastructure into uniformly sized, single-region clusters managed with CDK for Terraform (Typescript). The new design replaces a previous multi-region, user-membership-based cluster scheme that led to unbalanced clusters and cross-region latency. The project aims to improve reliability, simplify provisioning, and enable single-tenant clusters for enterprise customers. ### Replit — Introducing Multiplayer AI Chat - Company: Replit (replit.com) - Announced: 2024-02-14T17:00:00+00:00 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://replit.com/blog/multiplayer-ai - Record: https://forck.live/items/18007-introducing-multiplayer-ai-chat - Subject: Replit Agent Replit added multi-session chat with persistence and collaborative AI chat to its platform, allowing users to create multiple chat threads and collaborate with team members in the AI Chat Pane. ### Vercel — Vercel + WPP: World-class creativity enabled by technology - Company: Vercel (vercel.com) - Announced: 2024-02-14T13:00:00+00:00 - Category: partnership-acquisition - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://vercel.com/blog/vercel-wpp-creativity-enabled-by-technology - Record: https://forck.live/items/21010-vercel-wpp-world-class-creativity-enabled-by-technology - Subject: v0 / AI SDK / AI Gateway Vercel announced a strategic partnership with WPP to invest in co-developing an AI-first approach to digital experience development and delivery, facilitating quarterly global technology leader forums, optimizing project delivery efficiency through a standardized Next.js and Vercel development process, and providing early product roadmap access to WPP and its selected customers. ### Vercel — Evolving Vercel Functions - Company: Vercel (vercel.com) - Announced: 2024-02-14T13:00:00+00:00 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://vercel.com/blog/evolving-vercel-functions - Record: https://forck.live/items/21009-evolving-vercel-functions - Subject: v0 / AI SDK / AI Gateway Vercel announced a major iteration of Vercel Functions featuring increased concurrency (up to 30,000 concurrent invocations for Pro and 100,000 for Enterprise, with higher options), support for Web standard Request/Response API signatures, zero-config HTTP streaming, and longer function durations (up to 5 minutes for Pro and 15 minutes for Enterprise). Cold starts are now faster, with Vercel reporting twice the cold boot speed when comparing Next.js 13 to Next.js 14. Enterprise customers also gain automatic regional failover for increased resiliency. ### LG AI Research — [NeurIPS 2023] Trends on Score-Based Generative Modeling - Company: LG AI Research (lgresearch.ai) - Announced: 2024-02-14 - Category: research-paper - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://www.lgresearch.ai/blog/view?seq=405 - Record: https://forck.live/items/4866-neurips-2023-trends-on-score-based-generative-modeling - Subject: EXAONE LG AI Research blogs about research trends in score-based generative modeling from NeurIPS 2023, summarizing selected papers on diffusion model architectures and text-to-image diffusion models for downstream tasks. ### Hugging Face — AMD Pervasive AI Developer Contest! - Company: Hugging Face (huggingface.co) - Announced: 2024-02-14 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://huggingface.co/blog/amd_pervasive_developer_ai_contest - Record: https://forck.live/items/1942-amd-pervasive-ai-developer-contest - Subject: Platform AMD and Hugging Face announce the Pervasive AI Developer Contest, a global competition offering developers free access to AMD hardware and cash prizes up to $10,000 USD in three categories: Generative AI, Robotics AI, and PC AI. ### Supabase — Matryoshka embeddings: faster OpenAI vector search using Adaptive Retrieval - Company: Supabase (supabase.com) - Announced: 2024-02-13 - Category: new-model - Coverage: not counted - Announcement: yes - Group: models - Source: https://supabase.com/blog/matryoshka-embeddings - Record: https://forck.live/items/21447-matryoshka-embeddings-faster-openai-vector-search-using-adaptive-retrieval - Subject: Supabase OpenAI released text-embedding-3-small and text-embedding-3-large, which outperform the previous text-embedding-ada-002 model on MTEB and MIRACL benchmarks. The models include a dimensions API parameter that allows shortening the embedding vector to any size, enabled by Matryoshka Representation Learning (MRL), a training technique that preserves semantic meaning in truncated sub-vectors. ### LG AI Research — [NeurIPS 2023] Out-of-Distribution Detection via Synthetic Outlier Generation Part 1 - Company: LG AI Research (lgresearch.ai) - Announced: 2024-02-13 - Category: research-paper - Coverage: not counted - Announcement: no - Group: routine - Source: https://www.lgresearch.ai/blog/view?seq=406 - Record: https://forck.live/items/4868-neurips-2023-out-of-distribution-detection-via-synthetic-outlier-generation - Subject: EXAONE LG AI Research's DI Lab presented a research paper at NeurIPS 2023 on generating synthetic outliers using diffusion models for out-of-distribution detection, introducing Semantic-Discrepant outliers and a novel OOD detection method. ### LG AI Research — [NeurIPS 2023] Out-of-Distribution Detection via Synthetic Outlier Generation Part 2 - Company: LG AI Research (lgresearch.ai) - Announced: 2024-02-13 - Category: research-paper - Coverage: not counted - Announcement: no - Group: routine - Source: https://www.lgresearch.ai/blog/view?seq=407 - Record: https://forck.live/items/4867-neurips-2023-out-of-distribution-detection-via-synthetic-outlier-generation - Subject: EXAONE Two notable papers on outlier generation for out-of-distribution detection: one on textual outlier exposure using CLIP and LLMs, and another on learning with unreliable OOD sources. ### OpenAI — Memory and new controls for ChatGPT - Company: OpenAI (openai.com) - Announced: 2024-02-13 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://openai.com/index/memory-and-new-controls-for-chatgpt - Record: https://forck.live/items/809-memory-and-new-controls-for-chatgpt - Subject: GPT / ChatGPT / API OpenAI is testing a memory feature for ChatGPT that allows it to remember information from past conversations to improve future interactions, with user controls. ### Replit — Flexible Credits and Usage-Based Billing - Company: Replit (replit.com) - Announced: 2024-02-12T17:00:00+00:00 - Category: pricing-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://replit.com/blog/flexible-credits-and-usage-based-billing - Record: https://forck.live/items/18008-flexible-credits-and-usage-based-billing - Subject: Replit Agent Replit introduced usage-based billing for usage exceeding plan allotments, replacing the previous Power Up system. The Core plan now includes $8 in flexible credits per month usable across cloud services like Autoscale Deployments, Reserved VM Deployments, Outbound Data Transfer, and Postgres. An advanced usage dashboard at replit.com/usage provides per-Repl metrics for storage and data transfer. ### Vercel — Finishing Turborepo's migration from Go to Rust - Company: Vercel (vercel.com) - Announced: 2024-02-12T13:00:00+00:00 - Category: developer-tool-release - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://vercel.com/blog/finishing-turborepos-migration-from-go-to-rust - Record: https://forck.live/items/21011-finishing-turborepo-s-migration-from-go-to-rust - Subject: v0 / AI SDK / AI Gateway Vercel has completed porting Turborepo, its JavaScript and TypeScript build system, from Go to Rust. The migration involved building an all-Rust version of the run command and porting components such as file hashing, lockfile analysis, and cache signature verification. The new Rust-based implementation uses enums for better type safety and efficiency compared to the previous Go version. ### LG AI Research — [NeurIPS 2023] Large Language Model-Based Autonomous Agents - Company: LG AI Research (lgresearch.ai) - Announced: 2024-02-12 - Category: research-paper - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://www.lgresearch.ai/blog/view?seq=409 - Record: https://forck.live/items/4869-neurips-2023-large-language-model-based-autonomous-agents - Subject: EXAONE LLM-based autonomous agent research presented at NeurIPS 2023, covering general-purpose agents (Tree of Thoughts, Reflexion, AdaPlanner, SwiftSage) and domain-specific agents including web agents (Mind2Web), coding agents (InterCode), and embodied agents (DEPS).