# forck.live — 27 April – 3 May 2026

> 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 2026-W18, 27 April – 3 May 2026, 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/2026-W18
- The current issue, and the catalogue of the other weeks: https://forck.live/briefing.md
- Catalogue: https://forck.live/llms.txt

## Announcements

### Google Research — Catalyzing scientific impact through global partnerships and open resources

- Company: Google Research (research.google)
- Announced: 2026-05-01T16:37:00+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://research.google/blog/catalyzing-scientific-impact-through-global-partnerships-and-open-resources/
- Record: https://forck.live/items/1341-catalyzing-scientific-impact-through-global-partnerships-and-open-resources
- Subject: Research

Its open science philosophy, partnerships with scientific organizations, and a catalog of open-source tools and datasets across genomics, neuroscience, earth modeling, biodiversity, and healthcare. The post highlights existing resources such as DeepVariant, Neuroglancer, Open Buildings, SpeciesNet, and Health AI Developer Foundations, but does not announce any new product, model, or feature.

### Vercel — How GitBook serves 30,000 sites with sub-second content updates

- Company: Vercel (vercel.com)
- Announced: 2026-05-01T04:00:00+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://vercel.com/blog/how-gitbook-serves-30000-sites-with-sub-second-content-updates
- Record: https://forck.live/items/20714-how-gitbook-serves-30-000-sites-with-sub-second-content-updates
- Subject: v0 / AI SDK / AI Gateway

GitBook hosts 30,000 documentation sites on a single Vercel deployment, serving 120 million monthly page views. The platform processes 40,000 cache invalidations daily, each resolved in under 300 milliseconds, using tag-based invalidation triggered by merge events. AI-driven traffic now accounts for 41% of all page views, having surged 5x year-over-year in 2025.

### Cursor — Team Marketplace Updates

- Company: Cursor (cursor.com)
- Announced: 2026-05-01
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://cursor.com/changelog/05-01-26
- Record: https://forck.live/items/2615-team-marketplace-updates
- Subject: Cursor

Admins can now create a team marketplace without needing to connect a repository first. The update also lets them add, remove, and configure install behavior for first-party plugins directly in the team marketplace settings, with three distribution modes: Default Off, Default On, and Required.

### Harvey — How Harvey Secures Embeddings at Scale

- Company: Harvey (harvey.ai)
- Announced: 2026-04-30T15:00:00+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://www.harvey.ai/blog/how-harvey-secures-embeddings-at-scale
- Record: https://forck.live/items/17574-how-harvey-secures-embeddings-at-scale
- Subject: Harvey

Harvey describes how it secures embeddings against reversal attacks in its enterprise RAG systems, explaining the threat and its architectural controls.

### Google DeepMind — Enabling a new model for healthcare with AI co-clinician

- Company: Google DeepMind (deepmind.google)
- Announced: 2026-04-30T12:14:15+00:00
- Category: research-paper
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://deepmind.google/blog/ai-co-clinician/
- Record: https://forck.live/items/1236-enabling-a-new-model-for-healthcare-with-ai-co-clinician
- Subject: Gemini

Google DeepMind is researching the development of an AI co-clinician for healthcare.

### Harvey — Harvey to Open Chicago Office in July

- Company: Harvey (harvey.ai)
- Announced: 2026-04-30T12:00:00+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://www.harvey.ai/blog/harvey-to-open-chicago-office-in-july
- Record: https://forck.live/items/17575-harvey-to-open-chicago-office-in-july
- Subject: Harvey

Harvey announced it will open a new office in Chicago in July 2025, citing the city's diverse legal market and its existing work with several law firms and in-house teams in the region.

### Tencent — Real life is where context gets hard

- Company: Tencent (tencent.com)
- Announced: 2026-04-30T07:00:00+00:00
- Category: research-paper
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://hunyuan.tencent.com/research/100039
- Record: https://forck.live/items/4579-real-life-is-where-context-gets-hard
- Subject: Hunyuan

Tencent Hunyuan introduces CL-bench Life, a benchmark for evaluating context learning ability in real-life settings, containing 405 context-task pairs and 5,348 human-written rubrics. Evaluations of 12 language models show they solve only 14.5% of tasks on average, with the best model (GPT-5.5 High) solving 22.2%.

### Alibaba — Qwen-Scope: Decoding Intelligence, Unleashing Potential

- Company: Alibaba (alibaba.com)
- Announced: 2026-04-30T04:00:00+00:00
- Category: developer-tool-release
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://qwen.ai/blog?id=qwen-scope
- Record: https://forck.live/items/4607-qwen-scope-decoding-intelligence-unleashing-potential
- Subject: Qwen

Alibaba introduces Qwen-Scope, an interpretability toolkit that uses Sparse Autoencoders (SAEs) trained on Qwen3 and Qwen3.5 series models to decompose hidden representations into interpretable features, enabling controllable inference, data classification and synthesis, targeted fine-tuning, and evaluation redundancy analysis.

### Kyutai — MoshiRAG: Asynchronous Knowledge Retrieval for Full-Duplex Speech Language Models

- Company: Kyutai (kyutai.org)
- Announced: 2026-04-30
- Category: not stated
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://kyutai.org/blog/2026-04-30-moshi-rag
- Record: https://forck.live/items/18344-moshirag-asynchronous-knowledge-retrieval-for-full-duplex-speech-language-models
- Subject: Moshi / Pocket TTS

← Back to Blog MoshiRAG: Asynchronous Knowledge Retrieval for Full-Duplex Speech Language Models 30 April 2026

### OpenClaw — How OpenClaw Got Safer in Public

- Company: OpenClaw (openclaw.ai)
- Announced: 2026-04-30
- Category: safety-policy-update
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://openclaw.ai/blog/openclaw-security-in-public/
- Record: https://forck.live/items/17769-how-openclaw-got-safer-in-public
- Subject: OpenClaw

OpenClaw describes how it improved security in response to a high volume of vulnerability reports, including triage processes, hardening, and team expansion. As of April 30, GitHub showed 1,309 security advisories since January 10, with 87% of critical reports closed as invalid. The project pushed functionality to plugins, added observability, and moved secrets to references.

### Cognition — Devin heads east: Cognition opens its Singapore APAC headquarters

- Company: Cognition (cognition.com)
- Announced: 2026-04-30
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://cognition.com/blog/singapore
- Record: https://forck.live/items/7663-devin-heads-east-cognition-opens-its-singapore-apac-headquarters
- Subject: Devin / SWE

Cognition opened its Asia-Pacific headquarters in Singapore, appointing Richard Spence as Vice President and General Manager for the region. The company is building a local team across engineering, go-to-market, and partnerships, and is already working with Singaporean enterprise OCBC, where engineers have seen up to 30% improvement in code and test case generation.

### Baidu — ERNIE-5.1-Preview Tops LMArena Text Leaderboard as No.1 Chinese Model!

- Company: Baidu (baidu.com)
- Announced: 2026-04-30
- Category: new-model
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://ernie.baidu.com/blog/posts/ernie-5.1-preview-0430-release-on-lmarena/
- Record: https://forck.live/items/4641-ernie-5-1-preview-tops-lmarena-text-leaderboard-as-no-1-chinese-model
- Subject: ERNIE

Baidu announced that ERNIE-5.1-Preview topped the LMArena Text leaderboard as the No. 1 Chinese model and achieved high rankings in several category leaderboards. The model compresses parameters compared to ERNIE-5.0 using a decoupled fully-asynchronous reinforcement learning and scaled agentic post-training approach, resulting in improved foundational capabilities and cost-effectiveness at a reduced pre-training cost.

### xAI — Custom Voices

- Company: xAI (x.ai)
- Announced: 2026-04-30
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://x.ai/news/grok-custom-voices
- Record: https://forck.live/items/3471-custom-voices
- Subject: Grok

xAI announced Custom Voices, a feature that allows users to clone a voice from a short recording and use it across Grok Text to Speech and Voice Agent APIs.

### Cursor — Cursor Security Review

- Company: Cursor (cursor.com)
- Announced: 2026-04-30
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://cursor.com/changelog/04-30-26
- Record: https://forck.live/items/2616-cursor-security-review
- Subject: Cursor

Cursor Security Review enters beta on Teams and Enterprise plans, introducing Security Reviewer for PR-level vulnerability checks and Vulnerability Scanner for scheduled codebase scans, with support for custom triggers, instructions, and tooling via MCP servers.

### Together AI — Announcing Together AI and Adaption Partnership

- Company: Together AI (together.ai)
- Announced: 2026-04-30
- Category: partnership-acquisition
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://www.together.ai/blog/announcing-together-ai-and-adaption-partnership
- Record: https://forck.live/items/2353-announcing-together-ai-and-adaption-partnership
- Subject: Inference platform

Together AI announces a partnership with Adaption to integrate Together Fine-Tuning into Adaptive Data platform, enabling users to fine-tune models on optimized datasets directly from the platform.

### Together AI — From 732 bytes to nowhere: shutting down Copy Fail in production

- Company: Together AI (together.ai)
- Announced: 2026-04-30
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://www.together.ai/blog/shutting-down-copy-fail-in-production
- Record: https://forck.live/items/2352-from-732-bytes-to-nowhere-shutting-down-copy-fail-in-production
- Subject: Inference platform

Together AI responded to the Copy Fail kernel vulnerability (CVE-2026-31431) by disabling the vulnerable algif_aead module across its infrastructure, unloading it from running kernels and removing the module file to prevent re-loading, as a fleet-level emergency mitigation.

### OpenAI — Introducing Advanced Account Security

- Company: OpenAI (openai.com)
- Announced: 2026-04-30
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://openai.com/index/advanced-account-security
- Record: https://forck.live/items/157-introducing-advanced-account-security
- Subject: GPT / ChatGPT / API

OpenAI announced Advanced Account Security, which includes phishing-resistant login, stronger recovery, and enhanced protections to safeguard sensitive data and prevent account takeover.

### Google Research — Four ways Google Research scientists have been using Empirical Research Assistance

- Company: Google Research (research.google)
- Announced: 2026-04-29T21:07:00+00:00
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://research.google/blog/four-ways-google-research-scientists-have-been-using-empirical-research-assistance/
- Record: https://forck.live/items/1342-four-ways-google-research-scientists-have-been-using-empirical-research
- Subject: Research

Google Research scientists have been using Empirical Research Assistance (ERA) for real-world applications in epidemiology, cosmology, atmospheric monitoring, and neuroscience, including generating hospitalization forecasts for flu, COVID-19, and RSV, solving a cosmology problem involving cosmic strings, and monitoring CO2 with weather satellites.

### OpenAI — Where the goblins came from

- Company: OpenAI (openai.com)
- Announced: 2026-04-29T20:00:00+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://openai.com/index/where-the-goblins-came-from
- Record: https://forck.live/items/158-where-the-goblins-came-from
- Subject: GPT / ChatGPT / API

The spread of 'goblin outputs' in AI models, specifically focusing on GPT-5, and covers the timeline, root cause, and fixes for these personality-driven quirks.

### Replit — Introducing Replit App Monitoring

- Company: Replit (replit.com)
- Announced: 2026-04-29T17:00:00+00:00
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://replit.com/blog/app-monitoring
- Record: https://forck.live/items/17907-introducing-replit-app-monitoring
- Subject: Replit Agent

Replit introduced App Monitoring for published apps, which notifies users by email when their app goes down and provides an uptime graph. Replit Agent can now read production logs and the production database (with read-only access) to diagnose failures and identify root cause.

### Hugging Face — Granite 4.1 LLMs: How They’re Built

- Company: Hugging Face (huggingface.co)
- Announced: 2026-04-29T15:01:48+00:00
- Category: new-model
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://huggingface.co/blog/ibm-granite/granite-4-1
- Record: https://forck.live/items/1508-granite-4-1-llms-how-they-re-built
- Subject: Platform

IBM's Granite 4.1 family of dense, decoder-only LLMs (3B, 8B, 30B) is introduced, trained from scratch on ~15T tokens with a multi-stage pre-training pipeline including long-context extension to 512K tokens. The models are further refined with supervised fine-tuning on ~4.1M curated samples and reinforcement learning via on-policy GRPO with DAPO loss. All models are released under the Apache 2.0 license. Notably, the 8B instruct model matches or surpasses the previous Granite 4.0-H-Small (32B-A9B MoE) despite using a simpler dense architecture.

### Harvey — Day in the Life: Legal Engineer at Harvey

- Company: Harvey (harvey.ai)
- Announced: 2026-04-29T15:00:00+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://www.harvey.ai/blog/day-in-the-life-legal-engineer
- Record: https://forck.live/items/17576-day-in-the-life-legal-engineer-at-harvey
- Subject: Harvey

Harvey's Legal Engineers are former practicing lawyers who work with customers to apply AI to legal workflows. They identify use cases, build prompts and workflows, and run onboarding and training. The team spans multiple regions and operates across pre-sale, post-sale, and custom solutions tracks.

### OpenAI — Building the compute infrastructure for the Intelligence Age

- Company: OpenAI (openai.com)
- Announced: 2026-04-29T15:00:00+00:00
- Category: infrastructure-release
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://openai.com/index/building-the-compute-infrastructure-for-the-intelligence-age
- Record: https://forck.live/items/159-building-the-compute-infrastructure-for-the-intelligence-age
- Subject: GPT / ChatGPT / API

OpenAI is expanding its Stargate project to build more data center capacity for AGI compute infrastructure.

### ElevenLabs — Introducing ElevenMusic

- Company: ElevenLabs (elevenlabs.io)
- Announced: 2026-04-29T12:00:00+00:00
- Category: product-launch
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://elevenlabs.io/blog/introducing-elevenmusic
- Record: https://forck.live/items/3360-introducing-elevenmusic
- Subject: Voice AI

ElevenLabs announces ElevenMusic, a new product.

### OpenAI — Cybersecurity in the Intelligence Age

- Company: OpenAI (openai.com)
- Announced: 2026-04-29T04:00:00+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://openai.com/index/cybersecurity-in-the-intelligence-age
- Record: https://forck.live/items/160-cybersecurity-in-the-intelligence-age
- Subject: GPT / ChatGPT / API

OpenAI announces a five-part action plan to strengthen cybersecurity in the Intelligence Age, with an emphasis on democratizing AI-powered cyber defense and protecting critical systems.

### Gradium — The most accurate multilingual text-to-speech, by the numbers

- Company: Gradium (gradium.ai)
- Announced: 2026-04-29
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://gradium.ai/blog/word-error-rate-evaluations
- Record: https://forck.live/items/18321-the-most-accurate-multilingual-text-to-speech-by-the-numbers
- Subject: Gradium TTS / Phonon

Gradium published a technical post explaining how it measures word error rate (WER) for its multilingual text-to-speech models, including normalization challenges and benchmark results on the MiniMax Multilingual TTS Test Set. The company reports best scores on Spanish and Portuguese, second place on English and French, and mid-pack on German. Gradium notes that the benchmark has saturated for easy cases and that meaningful progress now requires harder tests.

### Gradium — Gradium Voice Launches on AWS as a SaaS Subscription and a SageMaker Model Image

- Company: Gradium (gradium.ai)
- Announced: 2026-04-29
- Category: availability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://gradium.ai/blog/gradium-aws-launch
- Record: https://forck.live/items/18320-gradium-voice-launches-on-aws-as-a-saas-subscription-and-a-sagemaker-model-image
- Subject: Gradium TTS / Phonon

Gradium made its voice AI platform available as a managed SaaS subscription on AWS Marketplace, billable through AWS accounts and counting toward committed spend. For in-VPC deployment, the same TTS model is also offered as a deployable Amazon SageMaker image with streaming inference. The SaaS L plan includes 45 million credits per month, covering approximately 1,000 hours of TTS or 4,100 hours of STT.

### Factory AI — Which Model Reviews Code Best?

- Company: Factory AI (factory.com)
- Announced: 2026-04-29
- Category: research-paper
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://factory.com/news/code-review-benchmark
- Record: https://forck.live/items/17823-which-model-reviews-code-best
- Subject: Droid

Factory Research tested 13 models on 50 real pull requests from five open-source projects, scoring each on F1 and cost per PR. GPT-5.2 and Claude Opus 4.6 achieved the highest F1 (~60%), but GPT-5.2 did so at $1.25/PR versus $3.11 for Opus. The most cost-efficient models were MiniMax M2.7 ($0.15/PR, 46-52% F1) and Kimi K2.5 ($0.41/PR, 51.9% F1), offering 75-86% of top-tier quality at 10-30x lower cost.

### Cursor — Build programmatic agents with the Cursor SDK

- Company: Cursor (cursor.com)
- Announced: 2026-04-29
- Category: developer-tool-release
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://cursor.com/changelog/sdk-release
- Record: https://forck.live/items/2617-build-programmatic-agents-with-the-cursor-sdk
- Subject: Cursor

Cursor released a software development kit (SDK) for building programmatic agents, along with updates to its Cloud Agents API, available in public beta with token-based pricing.

### Together AI — DeepSeek-V4 Pro now available on Together AI

- Company: Together AI (together.ai)
- Announced: 2026-04-29
- Category: availability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://www.together.ai/blog/deepseek-v4-pro-now-available-on-together-ai
- Record: https://forck.live/items/2354-deepseek-v4-pro-now-available-on-together-ai
- Subject: Inference platform

DeepSeek V4 Pro is now available on Together AI with a 512K-token context window, using a 1.6T-parameter MoE architecture with 49B activated parameters, supporting three reasoning modes (Non-Think, Think High, Think Max), and priced at $2.10 per 1M input tokens, $0.20 per 1M cached input tokens, and $4.40 per 1M output tokens. DeepSeek-V4 Flash is coming soon.

### Hugging Face — DeepInfra on Hugging Face Inference Providers 🔥

- Company: Hugging Face (huggingface.co)
- Announced: 2026-04-29
- Category: partnership-acquisition
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://huggingface.co/blog/inference-providers-deepinfra
- Record: https://forck.live/items/1509-deepinfra-on-hugging-face-inference-providers
- Subject: Platform

DeepInfra is now a supported Inference Provider on Hugging Face Hub, enabling access to models like DeepSeek V4, Kimi-K2.6, GLM-5.1, and DeepSeek V4 Pro through serverless inference.

### Supabase — Introducing the OSSCAR Index

- Company: Supabase (supabase.com)
- Announced: 2026-04-28T19:00:00+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://supabase.com/blog/introducing-osscar-index
- Record: https://forck.live/items/21290-introducing-the-osscar-index
- Subject: Supabase

Supabase launched the OSSCAR Index, a quarterly ranking of the fastest-growing open source organizations based on growth in GitHub stars, unique contributors, and package downloads. The first edition covers Q1 2026 and is open source. Projects are split into Emerging (fewer than 1,000 stars) and Scaling (1,000+ stars) divisions to ensure fair comparisons.

### NVIDIA — Introducing NVIDIA Nemotron 3 Nano Omni: Long-Context Multimodal Intelligence for Documents, Audio and Video Agents

- Company: NVIDIA (nvidia.com)
- Announced: 2026-04-28T15:58:57+00:00
- Category: new-model
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://huggingface.co/blog/nvidia/nemotron-3-nano-omni-multimodal-intelligence
- Record: https://forck.live/items/1510-introducing-nvidia-nemotron-3-nano-omni-long-context-multimodal-intelligence

NVIDIA announces Nemotron 3 Nano Omni, a new omni-modal understanding model that extends the Nemotron multimodal line to handle text, image, video, and audio inputs. The model achieves top accuracy on several benchmarks including MMlongbench-Doc, OCRBenchV2, WorldSense, DailyOmni, and VoiceBench, and offers up to 9x higher throughput and 2.9x faster reasoning speed compared to alternatives. It uses a hybrid Mamba-Transformer Mixture-of-Experts backbone with C-RADIOv4-H vision and Parakeet-TDT-0.6B-v2 audio encoders, and checkpoints are available in BF16, FP8, and NVFP4 formats on HuggingFace.

### Sakana AI — KAME: Tandem Architecture for Enhancing Knowledge in Real-Time Speech-to-Speech Conversational AI

- Company: Sakana AI (sakana.ai)
- Announced: 2026-04-28T15:00:00+00:00
- Category: research-paper
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://sakana.ai/kame-icassp-2026
- Record: https://forck.live/items/4685-kame-tandem-architecture-for-enhancing-knowledge-in-real-time-speech-to-speech
- Subject: Sakana models

Sakana AI introduces KAME, a tandem architecture for real-time speech-to-speech conversational AI that uses a fast speech model and an asynchronous backend LLM to enable 'speak while thinking'.

### ElevenLabs — ElevenLabs is expanding in Spain

- Company: ElevenLabs (elevenlabs.io)
- Announced: 2026-04-28T12:00:00+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://elevenlabs.io/blog/expanding-in-spain
- Record: https://forck.live/items/3362-elevenlabs-is-expanding-in-spain
- Subject: Voice AI

### Vercel — 2026 Vercel AI Accelerator recap

- Company: Vercel (vercel.com)
- Announced: 2026-04-28T04:00:00+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://vercel.com/blog/2026-vercel-ai-accelerator-recap
- Record: https://forck.live/items/20715-2026-vercel-ai-accelerator-recap
- Subject: v0 / AI SDK / AI Gateway

Vercel's 2026 AI Accelerator program concluded with 39 teams pitching at Demo Day. Each team received over $200K in infrastructure and credits from Vercel and partners including AWS, Anthropic, and OpenAI. Alumni from the prior cohort have collectively raised over $100M in venture funding.

### Alibaba — FlashQLA: CP-/Bwd-Friendly Fused Linear Attention Kernels for GDN

- Company: Alibaba (alibaba.com)
- Announced: 2026-04-28T02:00:00+00:00
- Category: developer-tool-release
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://qwen.ai/blog?id=flashqla
- Record: https://forck.live/items/4606-flashqla-cp-bwd-friendly-fused-linear-attention-kernels-for-gdn
- Subject: Qwen

Alibaba's Qwen team open-sourced FlashQLA, a high-performance linear attention kernel library for GDN (Gated Delta Network) that achieves 2-3x forward and 2x backward speedup over the FLA Triton kernel across multiple scenarios on NVIDIA Hopper.

### Kyutai — ARC-Encoder: learning compressed text representations for LLMs

- Company: Kyutai (kyutai.org)
- Announced: 2026-04-28
- Category: not stated
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://kyutai.org/blog/2026-04-28-arc-encoder
- Record: https://forck.live/items/18345-arc-encoder-learning-compressed-text-representations-for-llms
- Subject: Moshi / Pocket TTS

← Back to Blog ARC-Encoder: learning compressed text representations for LLMs 28 April 2026

### Together AI — Together AI Brings NVIDIA Nemotron 3 Nano Omni to Developers on Day 0

- Company: Together AI (together.ai)
- Announced: 2026-04-28
- Category: availability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://www.together.ai/blog/together-ai-brings-nvidia-nemotron-3-nano-omni-to-developers-on-day-0
- Record: https://forck.live/items/2355-together-ai-brings-nvidia-nemotron-3-nano-omni-to-developers-on-day-0
- Subject: Inference platform

Together AI announces that NVIDIA Nemotron 3 Nano Omni is now available on its platform, offering developers immediate access to the multimodal model for production-scale inference.

### OpenAI — Our commitment to community safety

- Company: OpenAI (openai.com)
- Announced: 2026-04-28
- Category: safety-policy-update
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://openai.com/index/our-commitment-to-community-safety
- Record: https://forck.live/items/162-our-commitment-to-community-safety
- Subject: GPT / ChatGPT / API

OpenAI describes how it protects community safety within ChatGPT via model safeguards, misuse detection, policy enforcement, and collaboration with safety experts.

### OpenAI — OpenAI models, Codex, and Managed Agents come to AWS

- Company: OpenAI (openai.com)
- Announced: 2026-04-28
- Category: availability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://openai.com/index/openai-on-aws
- Record: https://forck.live/items/161-openai-models-codex-and-managed-agents-come-to-aws
- Subject: GPT / ChatGPT / API

OpenAI announces availability of GPT models, Codex, and Managed Agents on AWS.

### OpenAI — OpenAI available at FedRAMP Moderate

- Company: OpenAI (openai.com)
- Announced: 2026-04-27T14:00:00+00:00
- Category: availability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://openai.com/index/openai-available-at-fedramp-moderate
- Record: https://forck.live/items/163-openai-available-at-fedramp-moderate
- Subject: GPT / ChatGPT / API

OpenAI has achieved FedRAMP Moderate authorization, making ChatGPT Enterprise and the OpenAI API available to U.S. federal agencies.

### Harvey — A Dallas Office 14 Years in the Making

- Company: Harvey (harvey.ai)
- Announced: 2026-04-27T13:10:00+00:00
- Category: availability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://www.harvey.ai/blog/a-dallas-office-14-years-in-the-making
- Record: https://forck.live/items/17577-a-dallas-office-14-years-in-the-making
- Subject: Harvey

Harvey opened a physical office in Dallas, Texas, with over a dozen employees already serving local customers. The company works with several leading Texas law firms and recently became the Official Legal AI Partner of the Dallas Mavericks and a partner of the Texas Rangers.

### Google DeepMind — Join the new AI Agents Vibe Coding Course from Google and Kaggle

- Company: Google DeepMind (deepmind.google)
- Announced: 2026-04-27T13:00:00+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://blog.google/innovation-and-ai/technology/developers-tools/kaggle-genai-intensive-course-vibe-coding-june-2026/
- Record: https://forck.live/items/1315-join-the-new-ai-agents-vibe-coding-course-from-google-and-kaggle
- Subject: Gemini

Google announces the return of its 5-Day AI Agents Intensive Course in partnership with Kaggle, with registration now open.

### Google DeepMind — Announcing our partnership with the Republic of Korea

- Company: Google DeepMind (deepmind.google)
- Announced: 2026-04-27T07:00:06+00:00
- Category: partnership-acquisition
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://deepmind.google/blog/announcing-our-partnership-with-the-republic-of-korea/
- Record: https://forck.live/items/1237-announcing-our-partnership-with-the-republic-of-korea
- Subject: Gemini

Google DeepMind announced a partnership with the Republic of Korea to accelerate scientific breakthroughs using frontier AI models.

### OpenAI — The next phase of the Microsoft OpenAI partnership

- Company: OpenAI (openai.com)
- Announced: 2026-04-27T06:00:00+00:00
- Category: partnership-acquisition
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://openai.com/index/next-phase-of-microsoft-partnership
- Record: https://forck.live/items/164-the-next-phase-of-the-microsoft-openai-partnership
- Subject: GPT / ChatGPT / API

OpenAI and Microsoft announce an amended agreement that simplifies their partnership, adds long-term clarity, and supports continued AI innovation at scale.

### Lovable — The Lovable mobile app is here

- Company: Lovable (lovable.dev)
- Announced: 2026-04-27
- Category: product-launch
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://lovable.dev/blog/mobile-app
- Record: https://forck.live/items/12086-the-lovable-mobile-app-is-here
- Subject: Lovable / AI app builder

Lovable launched a mobile app for iOS and Android that allows users to build web applications from their phones. The app enables users to queue prompts via voice or text, receive notifications when builds are complete, and seamlessly switch between mobile and desktop devices while continuing projects.

### Runway — No Idle GPUs: Managing Research Compute at Runway

- Company: Runway (runway.com)
- Announced: 2026-04-27
- Category: infrastructure-release
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://runway.com/news/engineering/no-idle-gpus-managing-research-compute-at-runway
- Record: https://forck.live/items/7888-no-idle-gpus-managing-research-compute-at-runway
- Subject: Gen / Aleph

Runway describes how it increased GPU utilization by over 20 percentage points using Kueue as a Kubernetes admission controller, implementing reserved queues for guaranteed capacity and a shared default queue that borrows idle GPUs with preemption.

### Cognition — Devin CLI: Start Local, Hand Off to the Cloud

- Company: Cognition (cognition.com)
- Announced: 2026-04-27
- Category: product-launch
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://cognition.com/blog/devin-for-terminal
- Record: https://forck.live/items/7665-devin-cli-start-local-hand-off-to-the-cloud
- Subject: Devin / SWE

Cognition released Devin CLI, a local coding agent that can hand off sessions to cloud agents. It supports models including Opus 4.7, GPT-5.5, and SWE-1.6, and can be installed via a curl command.

### Cognition — Engineering in the fast lane: Mercedes-Benz partners with Cognition

- Company: Cognition (cognition.com)
- Announced: 2026-04-27
- Category: partnership-acquisition
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://cognition.com/blog/mercedes-benz-cognition
- Record: https://forck.live/items/7664-engineering-in-the-fast-lane-mercedes-benz-partners-with-cognition
- Subject: Devin / SWE

Mercedes-Benz has partnered with Cognition to deploy Devin and Windsurf across its global engineering teams. In a pilot, Devin analyzed over 200,000 lines of COBOL code and reduced modernization time from an estimated eight months to eight days.

### Mistral AI — Workflows for work that runs the business

- Company: Mistral AI (mistral.ai)
- Announced: 2026-04-27
- Category: availability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://mistral.ai/news/workflows
- Record: https://forck.live/items/3299-workflows-for-work-that-runs-the-business
- Subject: Mistral / Le Chat

Mistral AI announced that Workflows is now in public preview.

### Hugging Face — How to build scalable web apps with OpenAI's Privacy Filter

- Company: Hugging Face (huggingface.co)
- Announced: 2026-04-27
- Category: developer-tool-release
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://huggingface.co/blog/openai-privacy-filter-web-apps
- Record: https://forck.live/items/1511-how-to-build-scalable-web-apps-with-openai-s-privacy-filter
- Subject: Platform

OpenAI released Privacy Filter, an open-source PII detector, on Hugging Face Hub. The blog post demonstrates building three web apps (Document Privacy Explorer, Image Anonymizer, SmartRedact Paste) using gradio.Server with the model.

### OpenAI — Choco automates food distribution with AI agents

- Company: OpenAI (openai.com)
- Announced: 2026-04-27
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://openai.com/index/choco
- Record: https://forck.live/items/166-choco-automates-food-distribution-with-ai-agents
- Subject: GPT / ChatGPT / API

This is a customer story about how Choco used OpenAI APIs to improve food distribution, productivity, and growth.

### OpenAI — An open-source spec for orchestration: Symphony

- Company: OpenAI (openai.com)
- Announced: 2026-04-27
- Category: open-weight-release
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://openai.com/index/open-source-codex-orchestration-symphony
- Record: https://forck.live/items/165-an-open-source-spec-for-orchestration-symphony
- Subject: GPT / ChatGPT / API

OpenAI announces Symphony, an open-source specification for orchestrating Codex agents, designed to connect issue trackers to persistent agent systems.