# forck.live — 18–24 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-W21, 18–24 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-W21
- The current issue, and the catalogue of the other weeks: https://forck.live/briefing.md
- Catalogue: https://forck.live/llms.txt

## Announcements

### Mistral AI — Emmi joins Mistral to accelerate the AI-native industry

- Company: Mistral AI (mistral.ai)
- Announced: 2026-05-23
- Category: partnership-acquisition
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://mistral.ai/news/accelerate-ai-native-industry
- Record: https://forck.live/items/3297-emmi-joins-mistral-to-accelerate-the-ai-native-industry
- Subject: Mistral / Le Chat

### ElevenLabs — $22 million earned by voice creators on ElevenLabs

- Company: ElevenLabs (elevenlabs.io)
- Announced: 2026-05-22T12:00:00+00:00
- Category: funding-company-news
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://elevenlabs.io/blog/22-million-earned-by-voice-creators-on-elevenlabs
- Record: https://forck.live/items/3357-22-million-earned-by-voice-creators-on-elevenlabs
- Subject: Voice AI

ElevenLabs reports that voice creators have earned $22 million on the platform.

### Krea — Made with Krea 2 moodboards

- Company: Krea (krea.ai)
- Announced: 2026-05-22
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://www.krea.ai/blog/made-with-krea-2-moodboards
- Record: https://forck.live/items/15105-made-with-krea-2-moodboards
- Subject: Krea 2 / AI creative platform

Krea 2 introduces moodboards, a feature that encodes a curated set of images into a single creative direction—color, mood, composition, and materials—so that generated images with different subjects share a consistent look. The post showcases example moodboards such as 'Prismatic airbrush' and 'Painterly graffiti' and explains that the aesthetic comes from the board, not the prompt.

### Krea — Krea 2 and Nano Banana 2, side by side

- Company: Krea (krea.ai)
- Announced: 2026-05-22
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://www.krea.ai/blog/krea-2-and-nano-banana-2-side-by-side
- Record: https://forck.live/items/15104-krea-2-and-nano-banana-2-side-by-side
- Subject: Krea 2 / AI creative platform

Krea published a comparison of its own foundation model Krea 2 and Google's Nano Banana 2, showing how each handles the same prompts differently. Krea 2 is described as opinionated and focused on composition and mood, while Nano Banana 2 defaults to photorealistic literal interpretation. Both models are available inside Krea's Image tool, and users can switch between them on the same canvas.

### Mistral AI — Connect the dots: Build with built-in and custom MCPs in Studio

- Company: Mistral AI (mistral.ai)
- Announced: 2026-05-22
- Category: capability-change
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://mistral.ai/news/connectors
- Record: https://forck.live/items/3298-connect-the-dots-build-with-built-in-and-custom-mcps-in-studio
- Subject: Mistral / Le Chat

Mistral AI announced a new capability in Studio that allows building with built-in and custom MCPs, featuring reusable connectors, direct tool calling, and human-in-the-loop approval controls for connecting enterprise data to AI applications.

### Mistral AI — Remote agents in Vibe. Powered by Mistral Medium 3.5.

- Company: Mistral AI (mistral.ai)
- Announced: 2026-05-22
- Category: model-update
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://mistral.ai/news/vibe-remote-agents-mistral-medium-3-5
- Record: https://forck.live/items/3285-remote-agents-in-vibe-powered-by-mistral-medium-3-5
- Subject: Mistral / Le Chat

Mistral AI announced the release of Mistral Medium 3.5, along with remote coding agents in Vibe and a new Work mode in Le Chat for complex tasks.

### OpenAI — OpenAI named a Leader in enterprise coding agents by Gartner

- Company: OpenAI (openai.com)
- Announced: 2026-05-22
- Category: funding-company-news
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://openai.com/index/gartner-2026-agentic-coding-leader
- Record: https://forck.live/items/113-openai-named-a-leader-in-enterprise-coding-agents-by-gartner
- Subject: GPT / ChatGPT / API

OpenAI is named a leader in the 2026 Gartner Magic Quadrant for Enterprise AI Coding Agents, with Codex recognized for innovation and enterprise-scale deployment.

### OpenAI — How Virgin Atlantic ships faster with Codex

- Company: OpenAI (openai.com)
- Announced: 2026-05-22
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://openai.com/index/virgin-atlantic
- Record: https://forck.live/items/112-how-virgin-atlantic-ships-faster-with-codex
- Subject: GPT / ChatGPT / API

Virgin Atlantic used OpenAI's Codex to build its mobile app, achieving high test coverage and zero critical defects under a tight deadline.

### Replit — Replit Enterprise, Now Self-Serve

- Company: Replit (replit.com)
- Announced: 2026-05-21T15:45:00+00:00
- Category: availability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://replit.com/blog/enterprise-self-serve
- Record: https://forck.live/items/17901-replit-enterprise-now-self-serve
- Subject: Replit Agent

Replit made its Enterprise plan self-serve, allowing any organization to purchase directly on the website, configure SSO and SCIM, and start building production apps in minutes without demo requests or contract negotiations. The self-serve plan includes unlimited seats with an annual credit commitment, and every customer is assigned a dedicated account manager.

### Harvey — Introducing Contract Intelligence in Harvey

- Company: Harvey (harvey.ai)
- Announced: 2026-05-21T13:30:00+00:00
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://www.harvey.ai/blog/introducing-contract-intelligence
- Record: https://forck.live/items/17563-introducing-contract-intelligence-in-harvey
- Subject: Harvey

Harvey announced Contract Intelligence, a system for in-house legal teams to manage contract intake, review, and negotiation. The system adapts to each team's playbooks and risk appetite, and improves with every executed agreement. It is being developed with a cohort of design partners.

### OpenAI — AdventHealth advances whole-person care with OpenAI

- Company: OpenAI (openai.com)
- Announced: 2026-05-21T12:00:00+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://openai.com/index/adventhealth
- Record: https://forck.live/items/114-adventhealth-advances-whole-person-care-with-openai
- Subject: GPT / ChatGPT / API

AdventHealth is using ChatGPT for Healthcare to streamline workflows and reduce administrative burden.

### Runway — Introducing Aleph 2.0 and Edit Studio

- Company: Runway (runway.com)
- Announced: 2026-05-21T10:00:00+00:00
- Category: model-update
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://runway.com/news/introducing-aleph-2-and-edit-studio
- Record: https://forck.live/items/7879-introducing-aleph-2-0-and-edit-studio
- Subject: Gen / Aleph

Runway launched Aleph 2.0, an upgrade to its video editing model, along with Edit Studio, a new product experience. Aleph 2.0 supports editing up to 30-second 1080p videos, enables localized edits with precise preservation of input video, allows image-level control over edits, and applies edits across multiple shots. Edit Studio is designed for marketing teams, filmmakers, and small business owners and is available on all paid Runway plans.

### Black Forest Labs — FLUX Erase: Remove anything, leave no trace

- Company: Black Forest Labs (bfl.ai)
- Announced: 2026-05-21T08:37:00+00:00
- Category: product-launch
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://bfl.ai/blog/flux-erase-remove-anything-leave-no-trace
- Record: https://forck.live/items/3391-flux-erase-remove-anything-leave-no-trace
- Subject: FLUX

FLUX Erase product announced, with description 'Remove anything, leave no trace'.

### Lovable — We Gave Our Agent a Vent Tool

- Company: Lovable (lovable.dev)
- Announced: 2026-05-21T07:21:14+00:00
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://lovable.dev/blog/we-gave-our-agent-a-vent-tool
- Record: https://forck.live/items/12080-we-gave-our-agent-a-vent-tool
- Subject: Lovable / AI app builder

Lovable introduced a 'Vent Tool' and a knowledge base called Lovable Stack Overflow (LSO) to detect when users are stuck and inject relevant solutions into the agent's context. An early version of LSO reduced stuck rates by 5% and increased publish rates by 2%.

### Krea — Krea 2 LoRA training is now available

- Company: Krea (krea.ai)
- Announced: 2026-05-21
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://www.krea.ai/blog/krea-2-lora-training
- Record: https://forck.live/items/15079-krea-2-lora-training-is-now-available
- Subject: Krea 2 / AI creative platform

Krea released LoRA training for Krea 2 in beta, allowing subscribers to fine-tune style, character, or object identity from a set of images. The trained LoRAs work directly in the Image tool, support strength adjustment and multi-LoRA combination, and remain private by default with optional sharing.

### Krea — Krea 2 Prompting Guide: Exploration, Style References, Moodboards

- Company: Krea (krea.ai)
- Announced: 2026-05-21
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://www.krea.ai/blog/krea-2-deep-dive-walkthrough
- Record: https://forck.live/items/15078-krea-2-prompting-guide-exploration-style-references-moodboards
- Subject: Krea 2 / AI creative platform

Krea's guide explains how to prompt its Krea 2 foundation model across three layers: open-ended exploration with vague prompts, style references with adjustable strength, and moodboards. It details how the style transfer system extracts and applies stylistic components, and how the strength slider controls the balance between the reference and the base model.

### Amp — Rush, 2.0

- Company: Amp (ampcode.com)
- Announced: 2026-05-21
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://ampcode.com/news/rush-2.0
- Record: https://forck.live/items/7747-rush-2-0
- Subject: Amp

Amp released Rush 2.0, a new version of its agent mode that uses GPT-5.5 instead of Haiku 4.5. The update improved solve rate from 39% to 44% and reduced average completion time from 2 minutes 59 seconds to 1 minute 32 seconds, with a slight cost increase to $0.58 per task.

### Amp — Amp Labs

- Company: Amp (ampcode.com)
- Announced: 2026-05-21
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://ampcode.com/news/amp-labs
- Record: https://forck.live/items/7746-amp-labs
- Subject: Amp

Amp is assembling small teams of software builders to bring AI to a single company per industry and region, with CEO mandates. The company profits only through warrants unlocked if the partner company's stock appreciates.

### Cognition — Devin is Getting a Windows PC

- Company: Cognition (cognition.com)
- Announced: 2026-05-21
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://cognition.com/blog/devin-is-getting-a-windows-pc
- Record: https://forck.live/items/7659-devin-is-getting-a-windows-pc
- Subject: Devin / SWE

Cognition announced that Devin, its AI software engineer, can now operate natively in a Windows environment, building, running, and testing in its own Windows VM. The capability is in beta for Enterprise Cloud and Dedicated Deployment customers.

### xAI — Use Grok in OpenCode

- Company: xAI (x.ai)
- Announced: 2026-05-21
- Category: availability-change
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://x.ai/news/grok-opencode
- Record: https://forck.live/items/3464-use-grok-in-opencode
- Subject: Grok

xAI announces that Grok is now available in OpenCode, accessible via SuperGrok or X Premium subscription.

### Replicate — How to prompt Grok Imagine Video 1.5

- Company: Replicate (replicate.com)
- Announced: 2026-05-21
- Category: capability-change
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://replicate.com/blog/grok-imagine
- Record: https://forck.live/items/2465-how-to-prompt-grok-imagine-video-1-5
- Subject: Platform

Replicate published a blog post with prompting tips for xAI's Grok Imagine Video 1.5 model, which can generate video with synchronized audio in a single pass. The post includes example prompts and guidance on adding sound, using intensity modifiers, describing camera movement, and keeping prompts focused. Users can run the model on Replicate's platform and start from a generated still image for best results. No pricing, license, or availability changes are announced; the post focuses on improving usage of the existing model release.

### Stability AI — Meet Stable Audio 3.0, the model family built for artistic experimentation with open-weight models

- Company: Stability AI (stability.ai)
- Announced: 2026-05-20T14:59:27+00:00
- Category: new-model
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://stability.ai/news-updates/meet-stable-audio-3-the-model-family-built-for-artistic-experimentation-with-open-weight-models
- Record: https://forck.live/items/17475-meet-stable-audio-3-0-the-model-family-built-for-artistic-experimentation-with
- Subject: Stable Diffusion

Stability AI released Stable Audio 3.0, a family of open-weight music models trained on licensed data. The models support variable-length generation up to six minutes and full song composition on portable devices. Three models are available as open weights on Hugging Face, while the Large model is accessible via the Stability AI API.

### Harvey — Introducing Command Center

- Company: Harvey (harvey.ai)
- Announced: 2026-05-20T13:30:00+00:00
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://www.harvey.ai/blog/introducing-command-center
- Record: https://forck.live/items/17564-introducing-command-center
- Subject: Harvey

Harvey announced Command Center, a dashboard providing usage analytics, peer benchmarking, an agent for natural-language queries over deployment data, and intelligent recommendations for new capabilities. The feature is available through early access with a waitlist.

### Alibaba — Qwen3.7: The Agent Frontier

- Company: Alibaba (alibaba.com)
- Announced: 2026-05-20T02:00:00+00:00
- Category: new-model
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://qwen.ai/blog?id=qwen3.7
- Record: https://forck.live/items/4612-qwen3-7-the-agent-frontier
- Subject: Qwen

Alibaba introduces Qwen3.7-Max, a proprietary model designed for the agent era, with strong capabilities in coding, general-purpose agents, reasoning, and multilingual tasks, as demonstrated by various benchmark results.

### Google DeepMind — Running Guide agent: A step towards running unbounded

- Company: Google DeepMind (deepmind.google)
- Announced: 2026-05-20T00:01:00+00:00
- Category: capability-change
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://blog.google/innovation-and-ai/models-and-research/google-deepmind/running-guide-agent/
- Record: https://forck.live/items/1309-running-guide-agent-a-step-towards-running-unbounded
- Subject: Gemini

Google DeepMind announces Running Guide agent, an AI agent that provides real-time audio navigation and obstacle detection to help blind and low-vision athletes run independently.

### H Company — Handicap International finds its funders among 13,000 foundations

- Company: H Company (hcompany.ai)
- Announced: 2026-05-20
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://hcompany.ai/newsroom/handicap-international
- Record: https://forck.live/items/18370-handicap-international-finds-its-funders-among-13-000-foundations
- Subject: Holo

H Company built AI agents for Handicap International that automate the identification and scoring of Swiss foundations, replacing weeks of manual prospecting per cycle. The agents run continuously, scoring foundations against a custom relevance framework, and the project projects a 30% increase in philanthropic revenue. The agent is now being explored for expansion to additional markets.

### Kyutai — Introducing KE:SAI

- Company: Kyutai (kyutai.org)
- Announced: 2026-05-20
- Category: partnership-acquisition
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://kyutai.org/blog/2026-05-20-kesai-launch
- Record: https://forck.live/items/18342-introducing-ke-sai
- Subject: Moshi / Pocket TTS

Kyutai announced KE:SAI, a new open-science lab focused on physical AI, established in partnership with the ELLIS Institute Tübingen.

### Factory AI — Deferred Context Engine

- Company: Factory AI (factory.com)
- Announced: 2026-05-20
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://factory.com/news/deferred-context-engine
- Record: https://forck.live/items/17827-deferred-context-engine
- Subject: Droid

Factory AI introduced the Deferred Context Engine for its Droid assistant, which defers loading of tool schemas and instructions until needed, reducing input tokens. Production telemetry shows average token reduction of 15.1% and up to 50.8% for sessions with large tool catalogs.

### Cohere — Introducing Command A+: Making sovereign agentic capabilities available to all

- Company: Cohere (cohere.com)
- Announced: 2026-05-20
- Category: new-model
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://cohere.com/blog/command-a-plus
- Record: https://forck.live/items/3268-introducing-command-a-making-sovereign-agentic-capabilities-available-to-all
- Subject: Command

Cohere announces Command A+, making sovereign agentic capabilities available to all.

### Cursor — Improvements to Cursor Automations

- Company: Cursor (cursor.com)
- Announced: 2026-05-20
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://cursor.com/changelog/05-20-26
- Record: https://forck.live/items/2605-improvements-to-cursor-automations
- Subject: Cursor

Cursor Automations now available in the Agents Window, support multi-repo and no-repo configurations, and offer a 50% discount on agent runs for new automations for the next 7 days. Five new no-repo templates are added to the Marketplace.

### Cursor — Shared Canvases and /loop Skill

- Company: Cursor (cursor.com)
- Announced: 2026-05-20
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://cursor.com/changelog/shared-canvases
- Record: https://forck.live/items/2604-shared-canvases-and-loop-skill
- Subject: Cursor

Cursor announced two new capabilities: shared canvases for team collaboration and the /loop skill for running prompts repeatedly on a schedule.

### OpenAI — The next phase of OpenAI’s Education for Countries

- Company: OpenAI (openai.com)
- Announced: 2026-05-20
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://openai.com/index/the-next-phase-of-education-for-countries
- Record: https://forck.live/items/117-the-next-phase-of-openai-s-education-for-countries
- Subject: GPT / ChatGPT / API

OpenAI is expanding its Education for Countries initiative with new partnerships, teacher training, and tools to increase AI adoption in schools.

### OpenAI — How Ramp engineers accelerate code review with Codex

- Company: OpenAI (openai.com)
- Announced: 2026-05-20
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://openai.com/index/ramp
- Record: https://forck.live/items/116-how-ramp-engineers-accelerate-code-review-with-codex
- Subject: GPT / ChatGPT / API

Ramp engineers use Codex with GPT-5.5 to review code and ship improvements, getting substantive feedback in minutes instead of hours.

### OpenAI — An OpenAI model has disproved a central conjecture in discrete geometry

- Company: OpenAI (openai.com)
- Announced: 2026-05-20
- Category: research-paper
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://openai.com/index/model-disproves-discrete-geometry-conjecture
- Record: https://forck.live/items/115-an-openai-model-has-disproved-a-central-conjecture-in-discrete-geometry
- Subject: GPT / ChatGPT / API

An OpenAI model solved the 80-year-old unit distance problem, disproving a major conjecture in discrete geometry and marking a milestone in AI-driven mathematics.

### fal — fal and AWS: Building for the Next Phase of Generative Media

- Company: fal (fal.ai)
- Announced: 2026-05-19T21:24:32+00:00
- Category: partnership-acquisition
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://blog.fal.ai/fal-and-aws-building-for-the-next-phase-of-generative-media/
- Record: https://forck.live/items/7997-fal-and-aws-building-for-the-next-phase-of-generative-media
- Subject: fal platform

fal announced a strategic partnership with Amazon Web Services (AWS) to use AWS's global scale and reliability for generative media inference workloads. The collaboration aims to serve fal's over 2.5 million developers and enterprise customers including Amazon MGM Studios, Canva, and Adobe.

### OpenAI — Introducing OpenAI for Singapore

- Company: OpenAI (openai.com)
- Announced: 2026-05-19T20:30:00+00:00
- Category: availability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://openai.com/index/introducing-openai-for-singapore
- Record: https://forck.live/items/118-introducing-openai-for-singapore
- Subject: GPT / ChatGPT / API

OpenAI announces a multi-year AI partnership in Singapore to expand deployment, build local talent, and support businesses and public services with AI.

### Ai2 — OlmoEarth v1.1: A more efficient family of Earth observation models

- Company: Ai2 (allenai.org)
- Announced: 2026-05-19T18:38:09+00:00
- Category: model-update
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://huggingface.co/blog/allenai/olmoearth-v1-1
- Record: https://forck.live/items/1500-olmoearth-v1-1-a-more-efficient-family-of-earth-observation-models

Ai2 released OlmoEarth v1.1, a new family of remote sensing foundation models that cuts compute costs by up to 3x compared to OlmoEarth v1 while maintaining similar performance. The improvement comes from merging tokens across Sentinel-2 resolutions into a single token, reducing sequence length, and modifying the pretraining regimen. The family includes Base, Tiny, and Nano sizes.

### Google Research — Empirical Research Assistance (ERA): From Nature publication to catalyzing Computational Discovery

- Company: Google Research (research.google)
- Announced: 2026-05-19T17:52:00+00:00
- Category: research-paper
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://research.google/blog/empirical-research-assistance-era-from-nature-publication-to-catalyzing-computational-discovery/
- Record: https://forck.live/items/1340-empirical-research-assistance-era-from-nature-publication-to-catalyzing
- Subject: Research

Google Research announces the publication of Empirical Research Assistance (ERA) in Nature, an AI tool that uses Gemini to write and optimize scientific code for expert-level computational modeling. The tool has been applied to open scientific problems including epidemiological forecasting, runoff prediction, CO2 mapping, and solar energy optimization. ERA is also used to build the Computational Discovery prototype, now available through a trusted tester program in Google Labs.

### Google Antigravity — Google Antigravity @ I/O 2026

- Company: Google Antigravity (antigravity.google)
- Announced: 2026-05-19T17:45:00+00:00
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://antigravity.google/blog/google-io-2026
- Record: https://forck.live/items/4548-google-antigravity-i-o-2026
- Subject: Antigravity

Google launched Antigravity 2.0, a standalone desktop application for agent interaction.

### Google DeepMind — Gemini for Science: AI experiments and tools for a new era of discovery

- Company: Google DeepMind (deepmind.google)
- Announced: 2026-05-19T17:45:00+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://blog.google/innovation-and-ai/technology/research/gemini-for-science-io-2026/
- Record: https://forck.live/items/1312-gemini-for-science-ai-experiments-and-tools-for-a-new-era-of-discovery
- Subject: Gemini

Announced a new collection of science tools and experiments called Gemini for Science, aimed at expanding the scale and precision of scientific exploration.

### Google DeepMind — Simulate real-world places with Project Genie and Street View

- Company: Google DeepMind (deepmind.google)
- Announced: 2026-05-19T17:45:00+00:00
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://blog.google/innovation-and-ai/models-and-research/google-deepmind/project-genie-expands/
- Record: https://forck.live/items/1311-simulate-real-world-places-with-project-genie-and-street-view
- Subject: Gemini

Google DeepMind is connecting Project Genie with Google Street View imagery to enable users to create new worlds anchored in reality.

### Google DeepMind — Making it easier to understand how content was created and edited

- Company: Google DeepMind (deepmind.google)
- Announced: 2026-05-19T17:45:00+00:00
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://blog.google/innovation-and-ai/products/identifying-ai-generated-media-online/
- Record: https://forck.live/items/1310-making-it-easier-to-understand-how-content-was-created-and-edited
- Subject: Gemini

Google DeepMind is expanding its tools to help users understand how content was created and edited across the web.

### Google Antigravity — Introducing Google Antigravity 2.0

- Company: Google Antigravity (antigravity.google)
- Announced: 2026-05-19T17:44:00+00:00
- Category: availability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://antigravity.google/blog/introducing-google-antigravity-2
- Record: https://forck.live/items/4549-introducing-google-antigravity-2-0
- Subject: Antigravity

Google Antigravity 2.0 is a new standalone desktop application for agent-optimized experiences, available on macOS, Linux, and Windows. It features dynamic subagents, asynchronous task management, JSON hooks, scheduled tasks, project-based grouping, live voice transcription, and various slash commands. It is powered by Gemini models and available to enterprises.

### Google Antigravity — Google Antigravity Built an OS (and more)

- Company: Google Antigravity (antigravity.google)
- Announced: 2026-05-19T17:43:00+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://antigravity.google/blog/google-antigravity-built-an-os
- Record: https://forck.live/items/4550-google-antigravity-built-an-os-and-more
- Subject: Antigravity

Google Antigravity used Gemini 3.5 Flash and a team of agents to build a functional operating system from a single prompt, costing $916.92 at API pricing.

### Google Antigravity — Google Antigravity CLI

- Company: Google Antigravity (antigravity.google)
- Announced: 2026-05-19T17:42:00+00:00
- Category: product-launch
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://antigravity.google/blog/introducing-google-antigravity-cli
- Record: https://forck.live/items/4551-google-antigravity-cli
- Subject: Antigravity

Google introduced the Antigravity CLI, a lightweight terminal interface for invoking, monitoring, and interacting with Antigravity agents, complementing Antigravity 2.0 and sharing a unified agent harness across developer surfaces.

### Google Antigravity — Google Antigravity SDK

- Company: Google Antigravity (antigravity.google)
- Announced: 2026-05-19T17:41:00+00:00
- Category: developer-tool-release
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://antigravity.google/blog/introducing-google-antigravity-sdk
- Record: https://forck.live/items/4552-google-antigravity-sdk
- Subject: Antigravity

Google Antigravity announced the preview launch of the Antigravity SDK, a Python library that provides programmatic access to the Antigravity coding agent. The SDK includes the same agent runtime powering Antigravity 2.0 and the Antigravity CLI, with built-in tools, safety policies, lifecycle hooks, and support for multimodal input, streaming, structured output, and custom tools. It is designed for AI-native development and allows developers to build agentic applications with minimal code.

### Google Antigravity — Google Antigravity in Gemini Enterprise

- Company: Google Antigravity (antigravity.google)
- Announced: 2026-05-19T17:40:00+00:00
- Category: availability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://antigravity.google/blog/google-antigravity-for-enterprises
- Record: https://forck.live/items/4553-google-antigravity-in-gemini-enterprise
- Subject: Antigravity

Google Antigravity is now available to Google Cloud customers, allowing enterprises to use Antigravity with their GCP projects and access Gemini 3.5 models through the Gemini Enterprise Agent Platform.

### Google Antigravity — Gemini 3.5 Flash in Google Antigravity

- Company: Google Antigravity (antigravity.google)
- Announced: 2026-05-19T17:39:00+00:00
- Category: model-update
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://antigravity.google/blog/gemini-3-5-flash-in-google-antigravity
- Record: https://forck.live/items/4554-gemini-3-5-flash-in-google-antigravity
- Subject: Antigravity

Gemini 3.5 Flash is now the default Flash model in Google Antigravity, delivering frontier performance for agents and coding, with co-optimization for speed and efficiency.

### Google Antigravity — Subagents, Hooks, Scheduled Tasks, Agent Management, Voice, and Much More

- Company: Google Antigravity (antigravity.google)
- Announced: 2026-05-19T17:38:00+00:00
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://antigravity.google/blog/google-io-2026-feature-deep-dive
- Record: https://forck.live/items/4555-subagents-hooks-scheduled-tasks-agent-management-voice-and-much-more
- Subject: Antigravity

Google Antigravity 2.0 and CLI are launched with new capabilities: subagents, asynchronous task management, hooks, project-based agent management, scheduled tasks (fixed to Gemini 3.5 Flash), and voice input with live transcription using Gemini Audio models.

### Google Antigravity — Changes to Antigravity Plans

- Company: Google Antigravity (antigravity.google)
- Announced: 2026-05-19T17:37:00+00:00
- Category: pricing-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://antigravity.google/blog/changes-to-antigravity-plans
- Record: https://forck.live/items/4556-changes-to-antigravity-plans
- Subject: Antigravity

Google Antigravity introduces a new $100/month Ultra plan, reduces the top-tier Ultra plan from $250 to $200/month, merges Gemini Flash and Pro quotas into a shared pool, and removes AI credits from base plans, making them an overage-only mechanism.

### Lovable — Connect Lovable to Google Workspace and Gemini Enterprise to build apps and more on data you already have

- Company: Lovable (lovable.dev)
- Announced: 2026-05-19T14:48:00+00:00
- Category: capability-change
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://lovable.dev/blog/already-using-google-now-you-can-build-with-it-in-lovable
- Record: https://forck.live/items/12081-connect-lovable-to-google-workspace-and-gemini-enterprise-to-build-apps-and
- Subject: Lovable / AI app builder

Lovable announced connectors to Google Workspace (Gmail, Calendar, Drive, Sheets, Slides, Maps Platform) and Gemini Enterprise, enabling users to build apps that read and write data from these services. The connectors are available to all Lovable users. The integration also supports BigQuery and allows companies on Gemini Enterprise to ground internal tools in their indexed data.

### ElevenLabs — Bringing voice AI into the classroom with ElevenLabs

- Company: ElevenLabs (elevenlabs.io)
- Announced: 2026-05-19T12:00:00+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://elevenlabs.io/blog/bringing-voice-ai-into-the-classroom-with-elevenlabs
- Record: https://forck.live/items/3361-bringing-voice-ai-into-the-classroom-with-elevenlabs
- Subject: Voice AI

### OpenAI — Advancing content provenance for a safer, more transparent AI ecosystem

- Company: OpenAI (openai.com)
- Announced: 2026-05-19T10:45:00+00:00
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://openai.com/index/advancing-content-provenance
- Record: https://forck.live/items/119-advancing-content-provenance-for-a-safer-more-transparent-ai-ecosystem
- Subject: GPT / ChatGPT / API

OpenAI is advancing content provenance by implementing Content Credentials, SynthID, and a verification tool to help identify and trust AI-generated media.

### Alibaba — Qwen3.5-LiveTranslate: From Sound to Sight, From Word to Right

- Company: Alibaba (alibaba.com)
- Announced: 2026-05-19T09:40:00+00:00
- Category: new-model
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://qwen.ai/blog?id=qwen3.5-livetranslate
- Record: https://forck.live/items/4610-qwen3-5-livetranslate-from-sound-to-sight-from-word-to-right
- Subject: Qwen

Alibaba announces Qwen3.5-LiveTranslate-Flash, a real-time multimodal translation model with expanded language coverage, ultra-low latency, voice cloning, and hotword enhancement, built on the Qwen3.5-Omni architecture and available via DashScope API.

### Thinking Machines Lab — Announcing Interactivity Research Grants

- Company: Thinking Machines Lab (thinkingmachines.ai)
- Announced: 2026-05-19
- Category: funding-company-news
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://thinkingmachines.ai/news/interactivity-research-grants
- Record: https://forck.live/items/18558-announcing-interactivity-research-grants
- Subject: Thinking Machines / Inkling

Thinking Machines Lab is offering multiple grants of $100,000 plus $25,000 in Tinker credits for research proposals advancing interactivity in AI. Proposals are due by June 19, 2026, and the lab encourages submissions on topics including evaluations for real-time multimodal interaction, safety for multimodal real-time AI, explaining complex information with generative UI, and steering agents during long-horizon tasks.

### Laminar — How Laminar compresses agent traces by 20x

- Company: Laminar (laminar.sh)
- Announced: 2026-05-19
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://laminar.sh/blog/laminar-20x-agent-trace-compression
- Record: https://forck.live/items/16850-how-laminar-compresses-agent-traces-by-20x
- Subject: flow / agent observability

Laminar implemented trace deduplication: instead of storing every LLM call's full message history (which grows quadratically with turn count), the system hashes each message and stores unique messages only once per trace, storing a compact array of hashes in spans. At query time, the original input array is reconstructed byte-identically via the SQL engine. On realistic long-running agents the approach reduces data from 5× to 10×, and on a 50-turn coding agent with stable system prompt and long tool outputs it approaches 50×.

### LG AI Research — Solving Bio-Challenges: Juseung Yun, the Mind Behind EXAONE Path

- Company: LG AI Research (lgresearch.ai)
- Announced: 2026-05-19
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://www.lgresearch.ai/news/view?seq=649
- Record: https://forck.live/items/10991-solving-bio-challenges-juseung-yun-the-mind-behind-exaone-path
- Subject: EXAONE

LG AI Research's Bio Intelligence Lab developed EXAONE Path, a pathology foundation model with 135 million parameters that achieves world-class performance despite being smaller than competitors. The model ranked 1st in internal datasets and 2nd in the global public benchmark Patho-bench. A subsequent version, EXAONE Path 2.5, integrates pathology images with multi-omics data including RNA, CNV, SNP, and methylation.

### Anthropic — New in Claude Managed Agents: dreaming, outcomes, and multiagent orchestration

- Company: Anthropic (anthropic.com)
- Announced: 2026-05-19
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://claude.com/blog/new-in-claude-managed-agents
- Record: https://forck.live/items/7406-new-in-claude-managed-agents-dreaming-outcomes-and-multiagent-orchestration
- Subject: Claude

Anthropic launched dreaming in Claude Managed Agents as a research preview, a memory-refining feature that reviews past sessions to find patterns and help agents self-improve. They also made outcomes, multiagent orchestration, and webhooks available to developers building with Managed Agents.

### Anthropic — New in Claude Managed Agents: self-hosted sandboxes and MCP tunnels

- Company: Anthropic (anthropic.com)
- Announced: 2026-05-19
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://claude.com/blog/claude-managed-agents-updates
- Record: https://forck.live/items/7405-new-in-claude-managed-agents-self-hosted-sandboxes-and-mcp-tunnels
- Subject: Claude

Anthropic announced that Claude Managed Agents can now run in self-hosted sandboxes (public beta) and connect to private MCP servers via MCP tunnels (research preview). The sandbox executes tools on the customer's own infrastructure or with managed providers such as Cloudflare, Daytona, Modal, or Vercel, while the agent loop remains on Anthropic's infrastructure. MCP tunnels allow agents to reach internal databases, private APIs, and other services without exposing them to the public internet.

### LG AI Research — Solving Bio-Challenges: Juseung Yun, the Mind Behind EXAONE Path

- Company: LG AI Research (lgresearch.ai)
- Announced: 2026-05-19
- Category: research-paper
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://www.lgresearch.ai/blog/view?seq=649
- Record: https://forck.live/items/4749-solving-bio-challenges-juseung-yun-the-mind-behind-exaone-path
- Subject: EXAONE

LG AI Research introduces EXAONE Path, a pathology foundation model with 135M parameters that achieves world-class performance despite being smaller than competitors. The model integrates pathology images with multi-omics data (RNA, CNV, SNP, methylation) in version 2.5, and is designed for cancer subtype classification, tumor microenvironment interpretation, and treatment response prediction.

### xAI — Use Grok in OpenClaw

- Company: xAI (x.ai)
- Announced: 2026-05-19
- Category: availability-change
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://x.ai/news/grok-openclaw
- Record: https://forck.live/items/3465-use-grok-in-openclaw
- Subject: Grok

xAI announces that SuperGrok and X Premium subscriptions can now be used inside OpenClaw, an open-source agent and personal assistant. This makes Grok available in a new local-first environment.

### Cursor — Cursor in Jira

- Company: Cursor (cursor.com)
- Announced: 2026-05-19
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://cursor.com/changelog/05-19-26
- Record: https://forck.live/items/2606-cursor-in-jira
- Subject: Cursor

Cursor is now integrated with Jira, allowing users to assign work items to Cursor or mention @Cursor in comments to initiate a cloud agent that uses the work item details and repository settings to perform tasks like fixing bugs or adding features. Installation requires Cursor admin access and Jira Commercial Cloud with Rovo enabled.

### Together AI — Benchmarking inference at scale: coding agents

- Company: Together AI (together.ai)
- Announced: 2026-05-19
- Category: research-paper
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://www.together.ai/blog/coding-agent-benchmarks
- Record: https://forck.live/items/2345-benchmarking-inference-at-scale-coding-agents
- Subject: Inference platform

A benchmark comparing Together Inference Engine against TensorRT-LLM and SGLang on a coding agent workload using Kimi K2.5. It reports that Together Inference Engine achieves 31% higher tokens per second (TPS) and 2× better time-to-first-token (TTFT) at saturation, attributed to full-stack optimizations including ThunderMLA and custom kernel rewrites.

### Hugging Face — Introducing the Ettin Reranker Family

- Company: Hugging Face (huggingface.co)
- Announced: 2026-05-19
- Category: new-model
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://huggingface.co/blog/ettin-reranker
- Record: https://forck.live/items/1501-introducing-the-ettin-reranker-family
- Subject: Platform

Hugging Face released six new CrossEncoder reranker models, the Ettin Reranker Family, built on Ettin ModernBERT encoders.

### Google DeepMind — Fast-tracking genetic leads to reverse cellular aging

- Company: Google DeepMind (deepmind.google)
- Announced: 2026-05-18T18:21:39+00:00
- Category: research-paper
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://deepmind.google/blog/fast-tracking-genetic-leads-to-reverse-cellular-aging/
- Record: https://forck.live/items/1220-fast-tracking-genetic-leads-to-reverse-cellular-aging
- Subject: Gemini

Biologists used Co-Scientist to identify novel factors that successfully rejuvenate human cells.

### Hugging Face — PaddleOCR 3.5: Running OCR and Document Parsing Tasks with a Transformers Backend

- Company: Hugging Face (huggingface.co)
- Announced: 2026-05-18T15:12:46+00:00
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://huggingface.co/blog/PaddlePaddle/paddleocr-transformers
- Record: https://forck.live/items/1502-paddleocr-3-5-running-ocr-and-document-parsing-tasks-with-a-transformers-backend
- Subject: Platform

PaddleOCR 3.5 introduces a Transformers inference backend, allowing supported OCR and document parsing models to run with Hugging Face Transformers, reducing integration friction for RAG and Document AI workflows.

### Lovable — Turn repeated instructions into reusable skills in Lovable

- Company: Lovable (lovable.dev)
- Announced: 2026-05-18T15:05:00+00:00
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://lovable.dev/blog/introducing-skills
- Record: https://forck.live/items/12082-turn-repeated-instructions-into-reusable-skills-in-lovable
- Subject: Lovable / AI app builder

Lovable introduced Skills, a feature that lets users create reusable markdown-based instructions for the AI agent. Skills are task-specific playbooks that activate only when relevant, complementing existing persistent knowledge fields. They are portable across tools like Anthropic and OpenAI, and can be personal or shared among team members.

### OpenAI — OpenAI and Dell partner to bring Codex to hybrid and on-premise enterprise environments

- Company: OpenAI (openai.com)
- Announced: 2026-05-18T10:00:00+00:00
- Category: partnership-acquisition
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://openai.com/index/dell-codex-enterprise-partnership
- Record: https://forck.live/items/120-openai-and-dell-partner-to-bring-codex-to-hybrid-and-on-premise-enterprise
- Subject: GPT / ChatGPT / API

OpenAI and Dell have partnered to bring Codex to hybrid and on-premise enterprise environments, enabling secure deployment of AI coding agents.

### Odyssey — Starchild-1: The First Real-Time Multimodal World Model

- Company: Odyssey (odyssey.systems)
- Announced: 2026-05-18
- Category: new-model
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://odyssey.systems/introducing-starchild-1
- Record: https://forck.live/items/18596-starchild-1-the-first-real-time-multimodal-world-model
- Subject: Odyssey world models

Odyssey previewed Starchild-1, a multimodal world model that autoregressively generates synchronized audio and video in real-time while responding to streaming user input. The model uses a causal multimodal training and inference stack, including a causal distillation pipeline and an asynchronous KV-cache architecture, to maintain coherent generation over extended interactions. Starchild-1 is positioned as an early step toward general world intelligence, with potential applications in robotics, education, gaming, healthcare, and defense.

### Odyssey — Agora-1: The Multi-Agent World Model

- Company: Odyssey (odyssey.systems)
- Announced: 2026-05-18
- Category: research-paper
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://odyssey.systems/introducing-agora-1
- Record: https://forck.live/items/18595-agora-1-the-multi-agent-world-model
- Subject: Odyssey world models

Odyssey released Agora-1, a multi-agent world model that allows up to four participants to share and interact within the same generated world simulation in real time. The model decouples simulation dynamics from rendering, learning a shared world state that evolves from player actions and rendering that state visually. Agora-1 is trained on the internal state of GoldenEye and functions as a learned game engine.

### Krea — Krea 2 Prompting Guide: Start Vague, Get Better Results

- Company: Krea (krea.ai)
- Announced: 2026-05-18
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://www.krea.ai/blog/explorative-prompting-krea-2
- Record: https://forck.live/items/15106-krea-2-prompting-guide-start-vague-get-better-results
- Subject: Krea 2 / AI creative platform

Krea published a prompting guide for Krea 2 that recommends starting with a vague prompt, reviewing the variety of outputs, and then narrowing down with a style hint. The guide explains that the model is designed to explore multiple interpretations rather than locking onto a single one.

### Runway — Why Distributed Training Is Hard: DTensor, Correctness and the Costs of Abstraction

- Company: Runway (runway.com)
- Announced: 2026-05-18
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://runway.com/news/engineering/dtensor-distributed-training
- Record: https://forck.live/items/7880-why-distributed-training-is-hard-dtensor-correctness-and-the-costs-of
- Subject: Gen / Aleph

Runway published a technical post explaining the challenges of distributed training with DTensor, including gradient correctness issues that arise when sharding tensors across process groups.

### Cognition — Introducing Auto-Triage

- Company: Cognition (cognition.com)
- Announced: 2026-05-18
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://cognition.com/blog/auto-triage
- Record: https://forck.live/items/7660-introducing-auto-triage
- Subject: Devin / SWE

Cognition announced Auto-Triage for Devin, a new capability that automatically monitors incoming alerts, investigates them, and provides investigation results, next steps, or a pull request. It integrates with systems like Slack, Linear, GitHub, Sentry, and Datadog, and is available now in Devin Automations.

### xAI — Skills in web, iOS, and Android

- Company: xAI (x.ai)
- Announced: 2026-05-18
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://x.ai/news/grok-skills
- Record: https://forck.live/items/3466-skills-in-web-ios-and-android
- Subject: Grok

xAI announced new persistent skills for Grok on web, iOS, and Android, enabling document generation, workflow automation, and the ability to build and share custom skills.

### Cursor — Composer 2.5

- Company: Cursor (cursor.com)
- Announced: 2026-05-18
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://cursor.com/changelog/composer-2-5
- Record: https://forck.live/items/2607-composer-2-5
- Subject: Cursor

Composer 2.5 is now available in Cursor, offering substantial improvements in intelligence and behavior over Composer 2, with pricing at Standard $0.50/M input, $2.50/M output tokens and Fast $3.00/M input, $15.00/M output tokens.