# forck.live — 29 April – 5 May 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-W18, 29 April – 5 May 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-W18 - The current issue, and the catalogue of the other weeks: https://forck.live/briefing.md - Catalogue: https://forck.live/llms.txt ## Announcements ### Hugging Face — Introducing the Open Leaderboard for Hebrew LLMs! - Company: Hugging Face (huggingface.co) - Announced: 2024-05-05 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://huggingface.co/blog/leaderboard-hebrew - Record: https://forck.live/items/1896-introducing-the-open-leaderboard-for-hebrew-llms - Subject: Platform Hugging Face introduces an open leaderboard for evaluating Hebrew language models, with four benchmarks designed to assess Hebrew-specific linguistic capabilities. ### Vercel — How Dub grew to 3,000 active domains with Vercel’s multi-tenant SaaS toolkit - Company: Vercel (vercel.com) - Announced: 2024-05-03T13:00:00+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://vercel.com/blog/how-dub-grew-to-3000-active-domains-with-vercels-multi-tenant-saas-toolkit - Record: https://forck.live/items/20985-how-dub-grew-to-3-000-active-domains-with-vercel-s-multi-tenant-saas-toolkit - Subject: v0 / AI SDK / AI Gateway Vercel's case study documents how Dub, an open-source link management platform, uses Vercel's Domains API and managed infrastructure to handle over 3,000 active domains and 5 million link redirects per month, growing at 25% and 40% month-over-month respectively. The API simplifies reverse proxy setup and SSL certificate issuance, enabling Dub to provision custom domains for users via a single API call. ### Vercel — Vercel Functions are now faster—and powered by Rust - Company: Vercel (vercel.com) - Announced: 2024-05-03T13:00:00+00:00 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://vercel.com/blog/vercel-functions-are-now-faster-and-powered-by-rust - Record: https://forck.live/items/20984-vercel-functions-are-now-faster-and-powered-by-rust - Subject: v0 / AI SDK / AI Gateway Vercel has migrated its serverless Functions runtime to Rust, resulting in faster cold starts and lower costs. The company reports 30% faster cold starts for smaller workloads and 80ms (average) and 500ms (p99) faster execution for larger workloads. The improvements are rolling out gradually to all customers with no action required. ### LG AI Research — The Latest Research Trends in Time Series Foundation Models - Company: LG AI Research (lgresearch.ai) - Announced: 2024-05-03 - Category: research-paper - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://www.lgresearch.ai/blog/view?seq=428 - Record: https://forck.live/items/4855-the-latest-research-trends-in-time-series-foundation-models - Subject: EXAONE This blog post from LG AI Research discusses the latest research trends in time series foundation models, categorizing methodologies into those based on large language models and those building dedicated time series models, and describing subcategories such as Without Adaptation, Adapt LLM, and Adapt to LLM. ### Hugging Face — Bringing the Artificial Analysis LLM Performance Leaderboard to Hugging Face - Company: Hugging Face (huggingface.co) - Announced: 2024-05-03 - Category: developer-tool-release - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://huggingface.co/blog/leaderboard-artificial-analysis - Record: https://forck.live/items/1897-bringing-the-artificial-analysis-llm-performance-leaderboard-to-hugging-face - Subject: Platform Artificial Analysis LLM Performance Leaderboard is now available on Hugging Face, providing comprehensive metrics on quality, price, speed, and latency for over 100 serverless LLM API endpoints. ### Replit — Introducing Object Storage - Company: Replit (replit.com) - Announced: 2024-05-02T16:00:00+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://replit.com/blog/object-storage - Record: https://forck.live/items/17996-introducing-object-storage - Subject: Replit Agent Replit announced Object Storage, a new service for persisting files and unstructured data within the Replit workspace and Deployments. It stores data in Buckets, handles simultaneous reads and writes from multiple processes, and is built on Google Cloud Storage with 99.999999999% durability. Python and TypeScript libraries are available with zero configuration, and users can also use official GCP Client Libraries for other languages. ### Vercel — Vercel AI SDK 3.1: ModelFusion joins the team - Company: Vercel (vercel.com) - Announced: 2024-05-02T13:00:00+00:00 - Category: developer-tool-release - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://vercel.com/blog/vercel-ai-sdk-3-1-modelfusion-joins-the-team - Record: https://forck.live/items/20986-vercel-ai-sdk-3-1-modelfusion-joins-the-team - Subject: v0 / AI SDK / AI Gateway Vercel released AI SDK 3.1, a TypeScript framework for building AI applications. The SDK is organized into three parts: AI SDK Core for generating text, structured objects, and tool calls; AI SDK UI for chat interfaces; and AI SDK RSC for streaming generative UIs with React Server Components. It supports providers including OpenAI, Anthropic, Google Gemini, and Mistral. ### Supabase — What's new in pgvector v0.7.0 - Company: Supabase (supabase.com) - Announced: 2024-05-02 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://supabase.com/blog/pgvector-0-7-0 - Record: https://forck.live/items/21426-what-s-new-in-pgvector-v0-7-0 - Subject: Supabase pgvector 0.7.0 introduces float16 (half-precision) HNSW indexes that consume half the memory of float32 indexes, along with sparse vector and bit vector representations. The update also adds L1, Hamming, and Jaccard distance functions. Float16 indexes can be built on existing float32 tables without changing the underlying data, or on float16 tables for additional disk space savings. ### Vercel — Vercel supports HIPAA compliance - Company: Vercel (vercel.com) - Announced: 2024-05-01T13:00:00+00:00 - Category: safety-policy-update - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://vercel.com/blog/vercel-supports-hipaa-compliance - Record: https://forck.live/items/20987-vercel-supports-hipaa-compliance - Subject: v0 / AI SDK / AI Gateway Vercel announced support for HIPAA compliance for its enterprise customers, enabling healthcare organizations and entities processing protected health information to use the Vercel Frontend Cloud while maintaining regulatory compliance. The company underwent an independent third-party audit to support compliance with the HIPAA Security Rule and HITECH Breach Notification Requirements, and will sign business associate agreements with covered entities. ### Hugging Face — Powerful ASR + diarization + speculative decoding with Hugging Face Inference Endpoints - Company: Hugging Face (huggingface.co) - Announced: 2024-05-01 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://huggingface.co/blog/asr-diarization - Record: https://forck.live/items/1898-powerful-asr-diarization-speculative-decoding-with-hugging-face-inference - Subject: Platform To deploy a custom inference handler on Hugging Face Inference Endpoints that combines Whisper for ASR, a Pyannote model for diarization, and an assistant model for speculative decoding, with code examples and configuration details. ### OpenAI — "Bad Grammar": Russian-linked Telegram comment activity - Company: OpenAI (openai.com) - Announced: 2024-05-01 - Category: safety-policy-update - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://openai.com/index/disrupting-malicious-uses-of-ai-bad-grammar - Record: https://forck.live/items/778-bad-grammar-russian-linked-telegram-comment-activity - Subject: GPT / ChatGPT / API OpenAI banned accounts linked to a Russia-origin operation called Bad Grammar that used AI to generate Telegram comments on political topics. ### OpenAI — Operation "Spamouflage": China-linked influence activity - Company: OpenAI (openai.com) - Announced: 2024-05-01 - Category: safety-policy-update - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://openai.com/index/disrupting-malicious-uses-of-ai-spamouflage - Record: https://forck.live/items/777-operation-spamouflage-china-linked-influence-activity - Subject: GPT / ChatGPT / API OpenAI banned accounts linked to the PRC-origin influence operation 'Spamouflage', which used AI to research social media, generate posts, and debug a previously unreported website. ### OpenAI — Operation "Doppelganger": Russian influence activity targeting Ukraine - Company: OpenAI (openai.com) - Announced: 2024-05-01 - Category: safety-policy-update - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://openai.com/index/disrupting-malicious-uses-of-ai-doppelganger - Record: https://forck.live/items/776-operation-doppelganger-russian-influence-activity-targeting-ukraine - Subject: GPT / ChatGPT / API OpenAI banned accounts linked to the Russian influence operation 'Doppelganger', which used AI to create anti-Ukraine content. ### OpenAI — Operation "Zero Zeno": Israel-linked influence activity - Company: OpenAI (openai.com) - Announced: 2024-05-01 - Category: safety-policy-update - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://openai.com/index/disrupting-malicious-uses-of-ai-zero-zeno - Record: https://forck.live/items/775-operation-zero-zeno-israel-linked-influence-activity - Subject: GPT / ChatGPT / API OpenAI banned accounts linked to a previously unreported Israel-linked influence operation called 'Zero Zeno' that used AI to generate content with various political stances. ### OpenAI — IUVM: Iran-linked influence content network - Company: OpenAI (openai.com) - Announced: 2024-05-01 - Category: safety-policy-update - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://openai.com/index/disrupting-malicious-uses-of-ai-iuvm - Record: https://forck.live/items/774-iuvm-iran-linked-influence-content-network - Subject: GPT / ChatGPT / API OpenAI banned accounts associated with the Iran-origin operation 'IUVM', which used AI to generate and translate pro-Iran, anti-Israel and anti-US website content. ### Supabase — Open Source Hackathon 2024 winners - Company: Supabase (supabase.com) - Announced: 2024-04-30 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://supabase.com/blog/supabase-oss-hackathon-winners - Record: https://forck.live/items/21427-open-source-hackathon-2024-winners - Subject: Supabase Supabase announced the winners of its Open Source Hackathon 2024, with vdbs winning best overall project. Winners and runner-ups in categories such as best use of AI, most fun, most technically impressive, and most visually pleasing received prizes including an Apple AirPods and Supabase swag kits. ### Hugging Face — Improving Prompt Consistency with Structured Generations - Company: Hugging Face (huggingface.co) - Announced: 2024-04-30 - Category: research-paper - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://huggingface.co/blog/evaluation-structured-outputs - Record: https://forck.live/items/1899-improving-prompt-consistency-with-structured-generations - Subject: Platform Hugging Face's research team, with collaborators at Dottxt, investigates how LLM evaluation results are highly sensitive to prompt format changes, showing that performance can vary by 10 points and model rankings can be inconsistent across different prompt formats. They describe experiments on MMLU subsets with 8 prompt formats and 5 models, highlighting the need for greater prompt consistency. ### Hugging Face — StarCoder2-Instruct: Fully Transparent and Permissive Self-Alignment for Code Generation - Company: Hugging Face (huggingface.co) - Announced: 2024-04-29 - Category: new-model - Coverage: not counted - Announcement: yes - Group: models - Source: https://huggingface.co/blog/sc2-instruct - Record: https://forck.live/items/1900-starcoder2-instruct-fully-transparent-and-permissive-self-alignment-for-code - Subject: Platform Hugging Face announces StarCoder2-15B-Instruct-v0.1, a self-aligned code generation model trained using a fully permissive and transparent pipeline. The model achieves a 72.6 HumanEval score, surpassing CodeLlama-70B-Instruct. ### OpenAI — We’re bringing the Financial Times’ world-class journalism to ChatGPT - Company: OpenAI (openai.com) - Announced: 2024-04-29 - Category: partnership-acquisition - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://openai.com/index/content-partnership-with-financial-times - Record: https://forck.live/items/779-we-re-bringing-the-financial-times-world-class-journalism-to-chatgpt - Subject: GPT / ChatGPT / API OpenAI and the Financial Times will collaborate on new AI experiences for FT readers.