# forck.live — 26 February – 3 March 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-W09, 26 February – 3 March 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-W09
- The current issue, and the catalogue of the other weeks: https://forck.live/briefing.md
- Catalogue: https://forck.live/llms.txt

## Announcements

### Vercel — Introducing AI SDK 3.0 with Generative UI support

- Company: Vercel (vercel.com)
- Announced: 2024-03-01T13:00:00+00:00
- Category: developer-tool-release
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://vercel.com/blog/ai-sdk-3-generative-ui
- Record: https://forck.live/items/21005-introducing-ai-sdk-3-0-with-generative-ui-support
- Subject: v0 / AI SDK / AI Gateway

Vercel released AI SDK 3.0, open-sourcing the Generative UI technology behind v0.dev. The SDK allows developers to stream React Server Components from LLMs, enabling rich, component-based interfaces beyond plaintext and markdown chatbots. It supports OpenAI-compatible function calling and works with providers like OpenAI, Mistral, and Fireworks.

### Vercel — The Frontend Cloud: Powering resiliency for global web applications

- Company: Vercel (vercel.com)
- Announced: 2024-02-29T13:00:00+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://vercel.com/blog/the-resiliency-of-the-frontend-cloud
- Record: https://forck.live/items/21006-the-frontend-cloud-powering-resiliency-for-global-web-applications
- Subject: v0 / AI SDK / AI Gateway

Vercel's Frontend Cloud provides serverless infrastructure and an Edge Network with 18 compute-capable data centers and over 100 points of presence to automatically scale web applications and handle traffic spikes without manual intervention. It includes built-in high-availability features such as regional failover, durable caching that serves stale content during backend outages, and Edge Config for backend failover. Customers like Under Armour, Sonos, NZXT, and The Washington Post have used the platform to maintain performance during peak events.

### Supabase — PostgREST Aggregate Functions

- Company: Supabase (supabase.com)
- Announced: 2024-02-29
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://supabase.com/blog/postgrest-aggregate-functions
- Record: https://forck.live/items/21444-postgrest-aggregate-functions
- Subject: Supabase

Supabase announced that PostgREST v12 now supports aggregate functions (avg, count, max, min, sum) directly through the API, allowing dynamic data summarization without requiring views. The feature is disabled by default due to potential performance risks, and users must enable it by modifying the PostgREST connection role.

### Upstage — Upstage and Polaris Office Partner to Create World's First On-Device AI Document Software

- Company: Upstage (upstage.ai)
- Announced: 2024-02-29
- Category: partnership-acquisition
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://www.upstage.ai/news/polaris-on-device-ai
- Record: https://forck.live/items/18445-upstage-and-polaris-office-partner-to-create-world-s-first-on-device-ai
- Subject: Solar Pro / Solar Mini

Upstage and Polaris Office announced a partnership to develop and commercialize the world's first on-device AI document software solution. The solution integrates Upstage's Solar LLM into Polaris Office's suite, enabling real-time document translation, summarization, and text generation without internet connectivity. The collaboration targets organizations with strict security protocols or limited internet access.

### Hugging Face — Text-Generation Pipeline on Intel® Gaudi® 2 AI Accelerator

- Company: Hugging Face (huggingface.co)
- Announced: 2024-02-29
- Category: developer-tool-release
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://huggingface.co/blog/textgen-pipe-gaudi
- Record: https://forck.live/items/1931-text-generation-pipeline-on-intel-gaudi-2-ai-accelerator
- Subject: Platform

Hugging Face announces a custom text-generation pipeline class for the Llama 2 family of models (7b, 13b, 70b) on Intel Gaudi 2 AI Accelerator using Optimum Habana, providing easy-to-use scripts and LangChain compatibility.

### LG AI Research — [WACV 2024] UNSPAT: Uncertainty-Guided SpatioTemporal Transformer for 3D Human Pose and Shape Estimation on Videos

- Company: LG AI Research (lgresearch.ai)
- Announced: 2024-02-28
- Category: research-paper
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://www.lgresearch.ai/blog/view?seq=414
- Record: https://forck.live/items/4862-wacv-2024-unspat-uncertainty-guided-spatiotemporal-transformer-for-3d-human
- Subject: EXAONE

LG AI Research proposes a new framework for 3D human pose and shape estimation from video, incorporating Spatial Alignment Module, Space2Batch, and uncertainty-guided attention re-weighting, achieving state-of-the-art results on benchmarks and presented at WACV 2024.

### Hugging Face — StarCoder2 and The Stack v2

- Company: Hugging Face (huggingface.co)
- Announced: 2024-02-28
- Category: new-model
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://huggingface.co/blog/starcoder2
- Record: https://forck.live/items/1932-starcoder2-and-the-stack-v2
- Subject: Platform

BigCode announced the release of the StarCoder2 family of open code LLMs in three sizes (3B, 7B, 15B) and The Stack v2 dataset. The models use a 16,384-token context window and are licensed under BigCode OpenRAIL-M v1.

### Upstage — Upstage Forge Partnership to Drive AI Innovation in Finance

- Company: Upstage (upstage.ai)
- Announced: 2024-02-27
- Category: partnership-acquisition
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://www.upstage.ai/news/kbank-mou
- Record: https://forck.live/items/18446-upstage-forge-partnership-to-drive-ai-innovation-in-finance
- Subject: Solar Pro / Solar Mini

Upstage signed an MOU with KT, kt cloud, and K Bank to collaborate on using large language models in finance. Upstage will fine-tune its Solar LLM with financial data, while KT and kt cloud provide computing infrastructure and K Bank identifies AI applications.

### Hugging Face — TTS Arena: Benchmarking Text-to-Speech Models in the Wild

- Company: Hugging Face (huggingface.co)
- Announced: 2024-02-27
- Category: product-launch
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://huggingface.co/blog/arena-tts
- Record: https://forck.live/items/1933-tts-arena-benchmarking-text-to-speech-models-in-the-wild
- Subject: Platform

Hugging Face announces the TTS Arena, a tool for benchmarking text-to-speech models via human voting, with an initial set of six models including both open-source and proprietary ones.

### Vercel — Deploying safely on Vercel without merge queues

- Company: Vercel (vercel.com)
- Announced: 2024-02-26T13:00:00+00:00
- Category: capability-change
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://vercel.com/blog/deploy-safely-on-vercel-without-merge-queues
- Record: https://forck.live/items/21007-deploying-safely-on-vercel-without-merge-queues
- Subject: v0 / AI SDK / AI Gateway

Vercel added an option to disable automatic promotion of production deployments, allowing teams to run custom CI checks on the built deployment before it becomes visible to end users. The deployment is promoted only when the user runs a promote command via CLI, API, or UI. This enables workflows that remove the need for a merge queue while still ensuring safety.

### Mistral AI — Le Chat

- Company: Mistral AI (mistral.ai)
- Announced: 2024-02-26T08:00:00+00:00
- Category: product-launch
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://mistral.ai/news/le-chat-mistral/
- Record: https://forck.live/items/7376-le-chat
- Subject: Mistral / Le Chat

Mistral AI released Le Chat, a new conversational assistant that can use Mistral Large, Mistral Small, or the prototype Mistral Next. It also includes an enterprise version called Le Chat Enterprise.

### Mistral AI — Au Large

- Company: Mistral AI (mistral.ai)
- Announced: 2024-02-26T08:00:00+00:00
- Category: new-model
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://mistral.ai/news/mistral-large/
- Record: https://forck.live/items/7375-au-large
- Subject: Mistral / Le Chat

Mistral AI released Mistral Large, a new flagship language model with top-tier reasoning and multilingual capabilities, and Mistral Small, optimized for low latency. Both are available through la Plateforme and Azure.

### Supabase — Build a content recommendation app with Flutter and OpenAI

- Company: Supabase (supabase.com)
- Announced: 2024-02-26
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://supabase.com/blog/content-recommendation-with-flutter
- Record: https://forck.live/items/21445-build-a-content-recommendation-app-with-flutter-and-openai
- Subject: Supabase

Supabase published a guide on building a content recommendation app using Flutter, Supabase as a vector database, and OpenAI's embeddings API. The app stores movie data and embeddings in Supabase, then performs vector similarity searches to recommend related movies. The guide covers creating a database table, using Supabase Edge Functions to fetch movie data from TMDB and generate embeddings via OpenAI, and querying similar movies from a Flutter interface.

### LG AI Research — Opening Up New Possibilities for AI, Lab Leader Soonyoung Lee of the Multimodal Lab

- Company: LG AI Research (lgresearch.ai)
- Announced: 2024-02-26
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://www.lgresearch.ai/blog/view?seq=412
- Record: https://forck.live/items/4863-opening-up-new-possibilities-for-ai-lab-leader-soonyoung-lee-of-the-multimodal
- Subject: EXAONE

LG AI Research's Multimodal Lab, led by Soonyoung Lee, focuses on multimodal AI research including image generation, image editing, image understanding, and medical data analysis. The lab is collaborating with LG H&H on design applications and hosts the NICE challenge at CVPR 2024.

### Hugging Face — AI Watermarking 101: Tools and Techniques

- Company: Hugging Face (huggingface.co)
- Announced: 2024-02-26
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://huggingface.co/blog/watermarking
- Record: https://forck.live/items/1934-ai-watermarking-101-tools-and-techniques
- Subject: Platform

AI watermarking techniques, including visible and invisible watermarks, data poisoning, and signing methods, as well as tools available on the Hugging Face Hub.
