# forck.live — 20–26 July 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-W30, 20–26 July 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-W30
- The current issue, and the catalogue of the other weeks: https://forck.live/briefing.md
- Catalogue: https://forck.live/llms.txt

## Announcements

### Together AI — Kimi K3 vs GPT-5.6 Sol on DeepSWE: Cost, Coding, and Routing

- Company: Together AI (together.ai)
- Announced: 2026-07-26
- Category: research-paper
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://www.together.ai/blog/kimi-k3-vs-gpt-5-6-sol-on-deepswe-cost-coding-and-routing
- Record: https://forck.live/items/2330-kimi-k3-vs-gpt-5-6-sol-on-deepswe-cost-coding-and-routing
- Subject: Inference platform

GPT-5.6 Sol edges Kimi K3 on single-shot quality (72.7% vs 68.5% pass@1), but Kimi K3 wins on pass@k with k>1 (89.4% vs 85.8% pass@4) and costs 64% less per rollout ($4.65 vs $8.37). The models diverge in strengths (correlation 0.46), enabling a cascade strategy that covers 95.6% of DeepSWE tasks at lower cost than either model alone.

### Tencent — Tencent Brings Together AI and Games to Help Preserve and Share Cultural Heritage of New UNESCO Site in Jingdezhen

- Company: Tencent (tencent.com)
- Announced: 2026-07-25T01:25:00+00:00
- Category: product-launch
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://www.tencent.com/tencent-uses-ai-and-games-to-help-preserve-cultural-heritage-of-jingdezhen/
- Record: https://forck.live/items/4572-tencent-brings-together-ai-and-games-to-help-preserve-and-share-cultural
- Subject: Hunyuan

Tencent launched Digital Jingdezhen: Porcelain Craft Adventure, an AI-powered cultural heritage game.

### Midjourney — Version 8.2

- Company: Midjourney (midjourney.com)
- Announced: 2026-07-24T22:18:48+00:00
- Category: model-update
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://updates.midjourney.com/version-8-2/
- Record: https://forck.live/items/18583-version-8-2
- Subject: Midjourney V8

Midjourney released version 8.2 of its image model, an update that focuses on aesthetics, image quality, and personalization. The company states that low-quality images should be 'dramatically reduced' and that personalization profiles now draw from a larger pool of rated images.

### Amazon — Introducing Claude Opus 5 on AWS: Anthropic’s most capable Opus model

- Company: Amazon (amazon.com)
- Announced: 2026-07-24T17:59:03+00:00
- Category: new-model
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://aws.amazon.com/blogs/machine-learning/introducing-claude-opus-5-on-aws-anthropics-most-capable-opus-model/
- Record: https://forck.live/items/2442-introducing-claude-opus-5-on-aws-anthropic-s-most-capable-opus-model
- Subject: Bedrock / Nova

Amazon announces availability of Claude Opus 5 on Amazon Bedrock and Claude Platform on AWS. The model is Anthropic's most advanced Opus model and the first in the fifth generation, with improvements in coding, agentic tasks, visual understanding, and long-running tasks. It matches Claude Fable 5's intelligence in many domains at Opus-tier pricing. On Bedrock, it benefits from zero data retention by default, and is powered by Bedrock's next-generation inference engine.

### Meta — Meta AI Doesn’t Just Think, It Acts

- Company: Meta (meta.com)
- Announced: 2026-07-24T17:00:55+00:00
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://about.fb.com/news/2026/07/meta-ai-muse-spark-doesnt-just-think-it-acts/
- Record: https://forck.live/items/9917-meta-ai-doesn-t-just-think-it-acts
- Subject: Meta AI

Meta announced that its Meta AI assistant, powered by the Muse Spark 1.1 model, can now plan, follow through on tasks, and take action on behalf of users. New capabilities include creating training schedules, researching topics, generating slides, and providing daily briefings. The features begin rolling out today in select markets on the Meta AI app and meta.ai.

### Harvey — Claude Opus 5, Now Live in Harvey

- Company: Harvey (harvey.ai)
- Announced: 2026-07-24T17:00:00+00:00
- Category: not stated
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://www.harvey.ai/blog/opus-5-in-harvey
- Record: https://forck.live/items/17539-claude-opus-5-now-live-in-harvey
- Subject: Harvey

Announcing Opus 5 in Harvey. Opus 5, the latest Anthropic model, is now available in Harvey.

### Amazon — Build an explainable next-best-product recommendation system for banking on AWS

- Company: Amazon (amazon.com)
- Announced: 2026-07-24T15:42:11+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://aws.amazon.com/blogs/machine-learning/build-an-explainable-next-best-product-recommendation-system-for-banking-on-aws/
- Record: https://forck.live/items/2443-build-an-explainable-next-best-product-recommendation-system-for-banking-on-aws
- Subject: Bedrock / Nova

An architectural overview and design decisions for building a Next-Best-Product recommendation system for banking using Amazon SageMaker AI and PyTorch, with a multi-tower neural network and attention mechanism.

### Amazon — Get started with OpenAI GPT-5.6 Sol, Terra, and Luna on Amazon Bedrock

- Company: Amazon (amazon.com)
- Announced: 2026-07-24T15:40:08+00:00
- Category: availability-change
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://aws.amazon.com/blogs/machine-learning/get-started-with-openai-gpt-5-6-sol-terra-and-luna-on-amazon-bedrock/
- Record: https://forck.live/items/2444-get-started-with-openai-gpt-5-6-sol-terra-and-luna-on-amazon-bedrock
- Subject: Bedrock / Nova

OpenAI GPT-5.6 Sol, Terra, and Luna are now generally available on Amazon Bedrock, accessible through the OpenAI Responses API on the bedrock-mantle endpoint, with pricing matching OpenAI first-party rates and usage counting toward existing AWS commitments.

### Lovable — How we run swarms of AI hacking agents against ourselves in a game of capture the flag

- Company: Lovable (lovable.dev)
- Announced: 2026-07-24T14:30:00+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://lovable.dev/blog/how-we-run-swarms-of-ai-hacking-agents-against-ourselves
- Record: https://forck.live/items/12068-how-we-run-swarms-of-ai-hacking-agents-against-ourselves-in-a-game-of-capture
- Subject: Lovable / AI app builder

Lovable describes its internal offensive security program that uses swarms of AI agents to probe its infrastructure for vulnerabilities. Agents must capture a flag—a string from an inaccessible system—as deterministic proof of a real vulnerability, not a model's guess. The program does not replace human security researchers; agents still require human coordination and understanding.

### Lovable — Enterprise features July roundup

- Company: Lovable (lovable.dev)
- Announced: 2026-07-24T14:30:00+00:00
- Category: capability-change
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://lovable.dev/blog/enterprise-features-july-roundup
- Record: https://forck.live/items/12067-enterprise-features-july-roundup
- Subject: Lovable / AI app builder

Lovable introduced several enterprise features including AIUC-1 certification, reusable authentication, publishing controls, Workspace Insights for governance, automated security scanning (Basic and Deep Scan), and cleanup of abandoned apps.

### ElevenLabs — Strengthening and protecting elections

- Company: ElevenLabs (elevenlabs.io)
- Announced: 2026-07-24T12:00:00+00:00
- Category: partnership-acquisition
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://elevenlabs.io/blog/elections
- Record: https://forck.live/items/3345-strengthening-and-protecting-elections
- Subject: Voice AI

ElevenLabs announces partnerships with Brazil's Superior Electoral Court (TSE) to add voice AI to its 'Electoral Assistant', with Comprova for detection training, and with Google DeepMind and Reality Defender for watermarking and detection tools. It also describes safeguards like No Go Voices and Voice CAPTCHA to prevent misuse of voice AI in elections.

### Meta — Introducing Facebook Verified

- Company: Meta (meta.com)
- Announced: 2026-07-24T11:00:54+00:00
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://about.fb.com/news/2026/07/introducing-facebook-verified/
- Record: https://forck.live/items/9918-introducing-facebook-verified
- Subject: Meta AI

Meta introduced Facebook Verified, a free badge that uses a selfie-based process to confirm a profile belongs to a real person. The badge appears on Marketplace, Dating, Groups, and Profile, with plans to expand to Feed posts. It is available to eligible users 18 and older in select markets, with a global rollout planned.

### NVIDIA — At AI Summit, South Korea Outlines Its AI Future With NVIDIA and Partners

- Company: NVIDIA (nvidia.com)
- Announced: 2026-07-24T04:34:27+00:00
- Category: partnership-acquisition
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://blogs.nvidia.com/blog/ai-summit-korea-partners-and-nvidia/
- Record: https://forck.live/items/1433-at-ai-summit-south-korea-outlines-its-ai-future-with-nvidia-and-partners
- Subject: AI platform

NVIDIA and South Korean partners, including KAIST, NAVER, SK Group, and Hyundai Motor Group, announced joint AI research labs, expansion of AI factories, and partnerships to advance AI infrastructure and physical AI.

### Ant Ling — Ling-3.0-flash: More Useful Work per Token

- Company: Ant Ling (ant-ling.com)
- Announced: 2026-07-24
- Category: new-model
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://developer.ant-ling.com/en/blogs/ling-3.0-flash-release
- Record: https://forck.live/items/22330-ling-3-0-flash-more-useful-work-per-token
- Subject: Ling / Ring / Ming

Ant Ling released Ling-3.0-flash, a 124B-parameter Mixture-of-Experts model with 5.1B active parameters per token, designed for token efficiency in agent workflows. The model features native 256K context extendable to 1M, hybrid linear attention (KDA + MLA), and agent-focused training across over 10,000 environments. It achieves a 25.3% pass rate on MiniAppBench, competitive with open-source models roughly twice its size.

### Ai2 — Who gets to understand AI?

- Company: Ai2 (allenai.org)
- Announced: 2026-07-24
- Category: not stated
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://allenai.org/blog/who-gets-to-understand-ai
- Record: https://forck.live/items/18544-who-gets-to-understand-ai
- Subject: Olmo / Molmo / Asta

As AI becomes critical infrastructure for science, medicine, industry, and government, one question matters more than ever: who gets to understand how these systems work?

### Swiss AI Initiative — Apertus 1.5

- Company: Swiss AI Initiative (swiss-ai.org)
- Announced: 2026-07-24
- Category: model-update
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://apertus-ai.org/articles/2026-07-apertus-1-5/
- Record: https://forck.live/items/18373-apertus-1-5
- Subject: Apertus

The Swiss AI Initiative released Apertus 1.5, a continued pretraining of the Apertus 1.0 8B and 70B models. The update adds native image understanding, an optional thinking mode, a 262,144-token context window (four times longer than Apertus 1.0), improved instruction-following, and better tool use. The models are fully open with open weights, open data, open values, and full training details.

### Manus — Build, Migrate, and Operate Your Supabase Database with Manus

- Company: Manus (manus.im)
- Announced: 2026-07-24
- Category: capability-change
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://manus.im/blog/manus-supabase-connector
- Record: https://forck.live/items/14703-build-migrate-and-operate-your-supabase-database-with-manus
- Subject: Manus

Manus released a Supabase connector that lets users query, migrate, and manage Supabase databases using natural language. The connector can execute SQL, apply schema migrations, deploy Edge Functions, and check security recommendations. It is available through Manus Connectors and requires a Supabase account with appropriate permissions.

### xAI — Grok in Google Workspace

- Company: xAI (x.ai)
- Announced: 2026-07-24
- Category: availability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://x.ai/news/introducing-google-workspace-addon
- Record: https://forck.live/items/3436-grok-in-google-workspace
- Subject: Grok

xAI made Grok available as an add-on for Google Workspace, allowing users to work with it directly in Sheets, Slides, and Docs.

### Anthropic — Introducing Claude Opus 5

- Company: Anthropic (anthropic.com)
- Announced: 2026-07-24
- Category: new-model
- Coverage: 2 outlets
- Announcement: yes
- Group: models
- Source: https://www.anthropic.com/news/claude-opus-5
- Record: https://forck.live/items/3249-introducing-claude-opus-5
- Subject: Claude

Anthropic announces Claude Opus 5, a new model that is available today. It approaches the frontier intelligence of Claude Fable 5 at half the price, achieves state-of-the-art results on coding and knowledge work benchmarks like Frontier-Bench and GDPval-AA, and provides improved performance over its predecessor Opus 4.8 at the same cost. It also excels in scientific research and visual output tasks, and demonstrates strong agency and thoroughness in evaluations.

### Together AI — Kimi K3 vs Claude Fable 5 on DeepSWE: Cost and Coding

- Company: Together AI (together.ai)
- Announced: 2026-07-24
- Category: research-paper
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://www.together.ai/blog/kimi-k3-vs-claude-fable-5-on-deepswe-cost-and-coding
- Record: https://forck.live/items/2331-kimi-k3-vs-claude-fable-5-on-deepswe-cost-and-coding
- Subject: Inference platform

The post compares Kimi K3 and Claude Fable 5 on the DeepSWE benchmark, finding that Kimi K3 matches Claude Fable 5 on quality at a third of the cost per solved task, with Kimi K3 being an open-weight model that gives teams full control over deployment.

### Perplexity — MCP Server 1.0: now backed by the Agent API

- Company: Perplexity (perplexity.ai)
- Announced: 2026-07-23T19:36:29+00:00
- Category: developer-tool-release
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://docs.perplexity.ai/docs/resources/changelog#july-2026
- Record: https://forck.live/items/5069-mcp-server-1-0-now-backed-by-the-agent-api
- Subject: Sonar

Perplexity released MCP Server v1.0.0, moving its model-backed tools from legacy Sonar models to Agent API presets (fast, medium, high). The update brings faster and cheaper performance, streaming progress for long-running research, and cancellation support. The strip_thinking and reasoning_effort parameters were removed.

### AMD — AAI 2026: AMD Delivers Full-Stack Compute for the Agentic AI Era

- Company: AMD (amd.com)
- Announced: 2026-07-23T18:30:00+00:00
- Category: not stated
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://ir.amd.com/news-events/press-releases/detail/1294/aai-2026-amd-delivers-full-stack-compute-for-the-agentic-ai-era
- Record: https://forck.live/items/15035-aai-2026-amd-delivers-full-stack-compute-for-the-agentic-ai-era
- Subject: AMD Instinct / ROCm

At Advancing AI 2026, AMD launches 6th Gen AMD EPYC™ CPUs, AMD Instinct™ MI400 Series GPUs, AMD Helios™ AI rackscale solutions, AMD Ryzen™ AI Embedded X100 processors and the AMD Kria™ AI SOM and Robotics Developer Platform.

### Microsoft — AT&T and Microsoft scale trillion-token workloads with Microsoft Foundry and AMD

- Company: Microsoft (microsoft.com)
- Announced: 2026-07-23T18:30:00+00:00
- Category: infrastructure-release
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://azure.microsoft.com/en-us/blog/att-and-microsoft-scale-trillion-token-workloads-with-microsoft-foundry-and-amd/
- Record: https://forck.live/items/2446-at-t-and-microsoft-scale-trillion-token-workloads-with-microsoft-foundry-and-amd
- Subject: Azure AI

Microsoft announces that AT&T used Microsoft Foundry Managed Compute with AMD Instinct MI300X GPUs to train its OTel2.0 telecom AI model, processing about 1 trillion tokens and saving tens of millions of dollars by using open models like Phi-4 instead of frontier models.

### AMD — AMD and Cerebras Announce Industry-Leading Ultra-Low-Latency and High Throughput AI Inference Solution

- Company: AMD (amd.com)
- Announced: 2026-07-23T17:45:00+00:00
- Category: partnership-acquisition
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://ir.amd.com/news-events/press-releases/detail/1293/amd-and-cerebras-announce-industry-leading-ultra-low-latency-and-high-throughput-ai-inference-solution
- Record: https://forck.live/items/15036-amd-and-cerebras-announce-industry-leading-ultra-low-latency-and-high
- Subject: AMD Instinct / ROCm

AMD and Cerebras announced a technical partnership to deliver a disaggregated AI inference solution combining AMD Helios rackscale systems with the Cerebras Wafer-Scale Engine. The joint solution is expected to deliver up to 5x higher tokens per second per watt, with availability initially through Cerebras Cloud in the second half of 2026.

### Runway — Introducing Runway Media Router

- Company: Runway (runway.com)
- Announced: 2026-07-23T17:00:23.330000+00:00
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://runway.com/news/company-news/introducing-runway-media-router
- Record: https://forck.live/items/7863-introducing-runway-media-router
- Subject: Gen / Aleph

Runway introduced Media Router, a new capability in Runway Dev that automatically selects the best video, image or audio model for each request based on user-defined preferences for cost, quality and latency. The router allows developers to set configs with price caps, allow/deny lists, and optimization preferences, and it returns metadata explaining which model was selected and why. Media Router is live for video, image and audio models in Runway Dev.

### Runway — Introducing Runway Dev

- Company: Runway (runway.com)
- Announced: 2026-07-23T16:00:00+00:00
- Category: product-launch
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://runway.com/news/company-news/introducing-runway-dev
- Record: https://forck.live/items/7864-introducing-runway-dev
- Subject: Gen / Aleph

Runway launched Runway Dev, an AI media platform for professional developers and enterprise teams. The platform provides access to Runway's frontier models along with third-party models, a model router, pre-built recipes, custom workflows, and interactive avatars. It is already used by companies including Adobe, ElevenLabs, and Shutterstock.

### Databricks — Why A Frontier Data Agent Outperforms General Coding Agents in Quality and Cost

- Company: Databricks (databricks.com)
- Announced: 2026-07-23T15:33:58+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://www.databricks.com/blog/why-frontier-data-agent-outperforms-general-coding-agents-quality-and-cost
- Record: https://forck.live/items/17502-why-a-frontier-data-agent-outperforms-general-coding-agents-in-quality-and-cost
- Subject: Mosaic AI

Databricks evaluated Genie Code against three general coding agents on 401 real data tasks and found it achieved the highest accuracy (76.6%) at the lowest mean cost ($0.55 per task), roughly half the cost of its closest competitor. The advantage is attributed to Genie Code's semantic search and persistent memory, which reduce inefficient workspace exploration and tool calls.

### Harvey — Harvey Launches Ethical Wall Enforcement With Intapp

- Company: Harvey (harvey.ai)
- Announced: 2026-07-23T15:30:00+00:00
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://www.harvey.ai/blog/harvey-intapp-ethical-walls
- Record: https://forck.live/items/17540-harvey-launches-ethical-wall-enforcement-with-intapp
- Subject: Harvey

Harvey made its integration with Intapp generally available, allowing law firms to sync existing Intapp Walls policies into Harvey for automatic enforcement across surfaces such as Threads, Vault, Review Tables, and Shared Spaces. The integration provides audit-ready logging and governs collaboration surfaces like Shared Spaces, Workflows, and Playbooks. It requires both a cloud license and a separate Harvey connector license.

### Sakana AI — Announcing Fugu-Ultra v1.1 and Claude Code interface for Fugu

- Company: Sakana AI (sakana.ai)
- Announced: 2026-07-23T15:00:00+00:00
- Category: model-update
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://sakana.ai/fugu-1-1-claude-code-interface
- Record: https://forck.live/items/4673-announcing-fugu-ultra-v1-1-and-claude-code-interface-for-fugu
- Subject: Sakana models

Sakana AI announces Fugu-Ultra v1.1, an upgrade to their orchestration engine that improves performance by up to 7.9 points over v1.0, beating leading models like Fable 5 in complex coding and reasoning tasks. They also announce a Claude Code-compatible interface for Fugu.

### Black Forest Labs — FLUX 3 x mimic: The Next Generation of Video-Action Models

- Company: Black Forest Labs (bfl.ai)
- Announced: 2026-07-23T15:00:00+00:00
- Category: new-model
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://bfl.ai/blog/flux-3-mimic
- Record: https://forck.live/items/3385-flux-3-x-mimic-the-next-generation-of-video-action-models
- Subject: FLUX

Black Forest Labs announces FLUX 3, a new multimodal foundation model that generates images, video, and audio jointly, and serves as the backbone for FLUX-mimic, a video-action model developed with mimic robotics for robot control. The model learns a world model that enables both content creation and physical AI, with early versions already running on robots at Audi.

### Black Forest Labs — FLUX 3 - Real World Models: Towards Multimodal Flow Models as the Backbone of Visual Intelligence.

- Company: Black Forest Labs (bfl.ai)
- Announced: 2026-07-23T15:00:00+00:00
- Category: new-model
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://bfl.ai/blog/flux-3
- Record: https://forck.live/items/3384-flux-3-real-world-models-towards-multimodal-flow-models-as-the-backbone-of
- Subject: FLUX

Black Forest Labs announces FLUX 3, a new multimodal foundation model that jointly learns from images, videos, and audio. It is available in Early Access for video generation, with image generation early access to follow. The model demonstrates capabilities in video, image, and action prediction, with preliminary evaluations showing competitive performance against other models.

### NVIDIA — NVIDIA AI Supercomputer Comes Online at Naval Postgraduate School

- Company: NVIDIA (nvidia.com)
- Announced: 2026-07-23T02:00:46+00:00
- Category: infrastructure-release
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://blogs.nvidia.com/blog/naval-postgraduate-school-dgx-ai-supercomputer/
- Record: https://forck.live/items/1435-nvidia-ai-supercomputer-comes-online-at-naval-postgraduate-school
- Subject: AI platform

NVIDIA commissioned an NVIDIA DGX GB300 system at the Naval Postgraduate School, bringing a large-scale AI supercomputer online for students, researchers, and faculty. The system includes NVIDIA Mission Control software and supports model training, inference, weather prediction, cybersecurity, and disaster resilience. The deployment is part of an ongoing collaboration and involves partners DDN, VAST, and Vertiv.

### Supabase — Observability for every Supabase project with Grafana Cloud

- Company: Supabase (supabase.com)
- Announced: 2026-07-23
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://supabase.com/blog/observability-for-every-supabase-project-with-grafana-cloud
- Record: https://forck.live/items/21278-observability-for-every-supabase-project-with-grafana-cloud
- Subject: Supabase

Supabase added a one-click Grafana Cloud integration to its Dashboard, providing a pre-built observability dashboard with over 200 metrics on any plan, including free. The integration reads from Supabase's Metrics API and supports self-hosted Grafana and existing Grafana Cloud setups with the same dashboard definitions. Log support is in progress.

### Midjourney — Midjourney’s First Acquisition

- Company: Midjourney (midjourney.com)
- Announced: 2026-07-23
- Category: partnership-acquisition
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://updates.midjourney.com/midjourneys-first-acquisition/
- Record: https://forck.live/items/18584-midjourney-s-first-acquisition
- Subject: Midjourney V8

Midjourney announced its first acquisition, the astrology app Co–Star, whose founder Banu and her team will join Midjourney to help build a broader organization around Product and Design. Co–Star will remain under Banu's control and direction with expanded resources.

### Harvey — Harvey Expands Collaboration with Microsoft on Legal AI

- Company: Harvey (harvey.ai)
- Announced: 2026-07-23
- Category: partnership-acquisition
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://www.harvey.ai/blog/harvey-expands-collaboration-with-microsoft-on-legal-ai
- Record: https://forck.live/items/17745-harvey-expands-collaboration-with-microsoft-on-legal-ai
- Subject: Harvey

Harvey announced that Microsoft's CELA organization will use the Harvey AI platform to support legal and compliance operations. The collaboration also includes Harvey expanding its internal use of Microsoft 365 and Microsoft 365 Copilot.

### Microsoft AI — Introducing MAI-Image-2.5-Pro and MAI-Voice-2-Flash | Microsoft AI

- Company: Microsoft AI (microsoft.ai)
- Announced: 2026-07-23
- Category: new-model
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://microsoft.ai/news/introducing-mai-image-2-5-pro-and-mai-voice-2-flash/
- Record: https://forck.live/items/10759-introducing-mai-image-2-5-pro-and-mai-voice-2-flash-microsoft-ai
- Subject: MAI

Microsoft AI introduced two new model variants in public preview: MAI-Image-2.5-Pro, its highest-fidelity image model, and MAI-Voice-2-Flash, a faster, cheaper voice model. The company also reported production deployments of its MAI models across Microsoft products, including Bing Image Creator, PowerPoint, OneDrive, and Dynamics 365 Contact Center, with specific performance and cost improvements.

### Microsoft AI — Hill-climbing MAI models for GitHub Copilot and Excel | Microsoft AI

- Company: Microsoft AI (microsoft.ai)
- Announced: 2026-07-23
- Category: model-update
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://microsoft.ai/news/hill-climbing-mai-models-for-github-copilot-and-excel/
- Record: https://forck.live/items/10758-hill-climbing-mai-models-for-github-copilot-and-excel-microsoft-ai
- Subject: MAI

Microsoft shared early results for MAI models specialized for agentic workloads in GitHub Copilot and Excel. The MAI model deployed in Excel is on par with GPT-5.6 for the most common tasks while being more cost-efficient, and MAI-Code-1-Flash in GitHub Copilot outperforms similarly sized models with a roughly 10% higher code accept rate and 10% lower median token usage than GPT 5.4 Mini and Claude Haiku 4.5. Microsoft is extending this hill-climbing approach to train efficient models across Copilot Chat, Outlook, PowerPoint, and more.

### Microsoft AI — Introducing MAI-Image-2.5-Pro and MAI-Voice-2-Flash

- Company: Microsoft AI (microsoft.ai)
- Announced: 2026-07-23
- Category: model-update
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://microsoft.ai/news/introducing-mai-image-2-5-pro-and-mai-voice-2-flash
- Record: https://forck.live/items/8420-introducing-mai-image-2-5-pro-and-mai-voice-2-flash
- Subject: MAI

Microsoft AI announced the public preview of MAI-Image-2.5-Pro (their highest-fidelity image model, priced at $5 per 1M text input tokens, $8 per 1M image input tokens, and $106 per 1M image output tokens) and MAI-Voice-2-Flash (a faster, 32% cheaper variant of MAI-Voice-2, priced at $15 per 1M characters). MAI-Image-2.5 is now the default model for Bing Image Creator and is used in PowerPoint and OneDrive; MAI-Voice-2-Flash powers Dynamics 365 Contact Center and is integrated in Azure Voice Live.

### Microsoft AI — Hill-climbing MAI models for GitHub Copilot and Excel

- Company: Microsoft AI (microsoft.ai)
- Announced: 2026-07-23
- Category: model-update
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://microsoft.ai/news/hill-climbing-mai-models-for-github-copilot-and-excel
- Record: https://forck.live/items/8419-hill-climbing-mai-models-for-github-copilot-and-excel
- Subject: MAI

Microsoft announced that its MAI-Code-1-Flash model, deployed in GitHub Copilot since June, has an approximately 10% higher code accept rate than GPT 5.4 Mini and Claude Haiku 4.5 in VS Code, with 10% lower median token usage. A specialized MAI model trained from that checkpoint for Excel is on par with GPT-5.6 for common tasks while being more cost-efficient and deployable on older GPU classes. Microsoft is extending this hill-climbing approach to other agentic products including Copilot Chat, Outlook, and PowerPoint.

### Luma — How Serviceplan MAKE Uses Luma to Put Creativity First

- Company: Luma (lumalabs.ai)
- Announced: 2026-07-23
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://lumalabs.ai/news/how-serviceplan-make-uses-luma-to-put-creativity-first
- Record: https://forck.live/items/7959-how-serviceplan-make-uses-luma-to-put-creativity-first
- Subject: Ray / Dream Machine

Serviceplan MAKE, part of Serviceplan Group, uses Luma to accelerate its creative workflow. The agency reports that what previously required a multi-day studio production can now be completed in a single day, allowing more time for creative thinking.

### Amp — Event Driven Orbs

- Company: Amp (ampcode.com)
- Announced: 2026-07-23
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://ampcode.com/news/event-driven-orbs
- Record: https://forck.live/items/7722-event-driven-orbs
- Subject: Amp

Amp's orbs can now be triggered by external events such as GitHub CI failures, Linear issues, monitor alerts, or Discord messages, and react to them outside Amp.

### Cognition — Welcoming The Interaction Company

- Company: Cognition (cognition.com)
- Announced: 2026-07-23
- Category: partnership-acquisition
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://cognition.com/blog/interaction
- Record: https://forck.live/items/7642-welcoming-the-interaction-company
- Subject: Devin / SWE

Cognition is acquiring The Interaction Company, the maker of the Poke personal AI agent. Poke is a proactive agent that texts users and has exchanged over 100 million messages in three months. Cognition plans to integrate its models and infrastructure to improve Poke.

### xAI — Workflows in Grok Build

- Company: xAI (x.ai)
- Announced: 2026-07-23
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://x.ai/news/workflows
- Record: https://forck.live/items/3437-workflows-in-grok-build
- Subject: Grok

xAI announced that Grok Build now supports writing and running workflows, which are orchestration scripts that distribute tasks across hundreds of parallel agents, verify results, and report back in a single background run. This is a new capability for the existing Grok Build product, not a new model or product launch. The announcement was made on July 23, 2026.

### Together AI — The production platform for open-weight AI inference

- Company: Together AI (together.ai)
- Announced: 2026-07-23
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://www.together.ai/blog/the-production-platform-for-open-weight-ai-inference
- Record: https://forck.live/items/2332-the-production-platform-for-open-weight-ai-inference
- Subject: Inference platform

Together AI announced a significant update to its inference platform, adding production-grade deployment features such as canary, blue-green, and rolling updates with auto-rollback, A/B and shadow testing, autoscaling, and a closed beta for custom training (RL and SFT).

### Hugging Face — Bringing Nunchaku 4-bit Diffusion Inference to Diffusers

- Company: Hugging Face (huggingface.co)
- Announced: 2026-07-23
- Category: developer-tool-release
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://huggingface.co/blog/nunchaku-diffusers
- Record: https://forck.live/items/1452-bringing-nunchaku-4-bit-diffusion-inference-to-diffusers
- Subject: Platform

Hugging Face integrates Nunchaku 4-bit diffusion inference into Diffusers, introducing Nunchaku Lite for native loading of quantized checkpoints without a separate inference engine, enabling faster inference and lower memory usage on compatible GPUs, including Blackwell-based NVIDIA GPUs for NVFP4 variants. The integration uses the Diffusers library and the kernels package, making it simple to load pre-quantized models with from_pretrained(). Benchmarks show about 30% speedup and significant VRAM reduction compared to BF16 pipelines. The diffuse-compressor toolkit allows users to quantize new architectures and publish them as standard Diffusers repositories.

### OpenAI — Launching Health in ChatGPT

- Company: OpenAI (openai.com)
- Announced: 2026-07-23
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://openai.com/index/health-in-chatgpt
- Record: https://forck.live/items/14-launching-health-in-chatgpt
- Subject: GPT / ChatGPT / API

OpenAI announced that ChatGPT now allows eligible U.S. users to securely connect medical records and Apple Health data for more personalized health insights.

### Google Research — SymptomAI: Towards a conversational AI agent for everyday symptom assessment

- Company: Google Research (research.google)
- Announced: 2026-07-22T21:32:00+00:00
- Category: research-paper
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://research.google/blog/symptomai-towards-a-conversational-ai-agent-for-everyday-symptom-assessment/
- Record: https://forck.live/items/1321-symptomai-towards-a-conversational-ai-agent-for-everyday-symptom-assessment
- Subject: Research

Google Research presents SymptomAI, a research study of conversational AI agents for symptom assessment and differential diagnosis, using Gemini Flash 2.0. The study enrolled 13,917 participants and compared SymptomAI's diagnostic performance against clinicians' assessments, finding that clinicians preferred SymptomAI's differential diagnoses over those of other clinicians in over 50% of cases.

### Google Research — Towards a quantum computer that learns from its errors

- Company: Google Research (research.google)
- Announced: 2026-07-22T18:40:21+00:00
- Category: research-paper
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://research.google/blog/towards-a-quantum-computer-that-learns-from-its-errors/
- Record: https://forck.live/items/1322-towards-a-quantum-computer-that-learns-from-its-errors
- Subject: Research

Google Research demonstrated a reinforcement learning framework for quantum error correction that allows a quantum computer to continuously adapt to drift and remain stable during long computations, as published in Nature.

### Sakana AI — UnMaskFork: Test-Time Scaling for Masked Diffusion via Deterministic Action Branching

- Company: Sakana AI (sakana.ai)
- Announced: 2026-07-22T15:00:00+00:00
- Category: research-paper
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://sakana.ai/umf
- Record: https://forck.live/items/4674-unmaskfork-test-time-scaling-for-masked-diffusion-via-deterministic-action
- Subject: Sakana models

Sakana AI announces its ICML 2026 paper 'UnMaskFork' which introduces a method for test-time scaling in masked diffusion language models by having multiple models collaborate through model switching and Monte Carlo Tree Search, improving performance on coding and math tasks without additional training or model changes.

### Lovable — Lovable is the first coding agent platform to earn AIUC-1 certification

- Company: Lovable (lovable.dev)
- Announced: 2026-07-22T14:00:00+00:00
- Category: safety-policy-update
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://lovable.dev/blog/aiuc-1
- Record: https://forck.live/items/12069-lovable-is-the-first-coding-agent-platform-to-earn-aiuc-1-certification
- Subject: Lovable / AI app builder

Lovable announced it is the first AI coding agent platform to achieve AIUC-1 certification, a security, safety, and reliability standard for AI agents developed with input from Stanford, MIT, MITRE, and the Cloud Security Alliance. The standard defines 51 requirements across six principles, including secrets management, secure code generation defaults, sandboxed execution, human oversight, and enterprise governance, and is independently verified through documented policies, technical implementation, operational process, and quarterly third-party red-teaming.

### Google DeepMind — Accelerating the frontiers of scientific discovery: Google’s $40M commitment to the Genesis Mission

- Company: Google DeepMind (deepmind.google)
- Announced: 2026-07-22T13:38:54+00:00
- Category: funding-company-news
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://deepmind.google/blog/accelerating-the-frontiers-of-scientific-discovery-googles-40m-commitment-to-the-genesis-mission/
- Record: https://forck.live/items/1203-accelerating-the-frontiers-of-scientific-discovery-google-s-40m-commitment-to
- Subject: Gemini

Google DeepMind and Google Cloud announce a $40 million commitment of AI tokens and cloud credits for the DOE's Genesis Mission, providing researchers with access to AI tools including AlphaEvolve, AlphaFold 3, AlphaGenome, WeatherNext, AlphaEarth Foundations, and Gemini for Government seats for one year across DOE National Laboratories.

### NVIDIA — NVIDIA Open Sources First GPU-Accelerated Medical Physics Simulation Framework

- Company: NVIDIA (nvidia.com)
- Announced: 2026-07-22T13:00:11+00:00
- Category: developer-tool-release
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://blogs.nvidia.com/blog/medical-physics-simulation-open-source/
- Record: https://forck.live/items/1436-nvidia-open-sources-first-gpu-accelerated-medical-physics-simulation-framework
- Subject: AI platform

NVIDIA announced the open-source release of its Medical Physics Simulation framework, a GPU-accelerated capability within NVIDIA Isaac for Healthcare that helps medical robotics developers model anatomy-device interactions, generate hard-to-capture scenarios, and train robot policies in simulation before hardware testing.

### AMD — AMD and Anthropic Announce Strategic Partnership to Deploy Up to 2 Gigawatts of AMD Instinct MI450 Series GPUs

- Company: AMD (amd.com)
- Announced: 2026-07-22T13:00:00+00:00
- Category: partnership-acquisition
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://ir.amd.com/news-events/press-releases/detail/1292/amd-and-anthropic-announce-strategic-partnership-to-deploy-up-to-2-gigawatts-of-amd-instinct-mi450-series-gpus
- Record: https://forck.live/items/15037-amd-and-anthropic-announce-strategic-partnership-to-deploy-up-to-2-gigawatts
- Subject: AMD Instinct / ROCm

AMD and Anthropic announced a strategic partnership to deploy up to 2 gigawatts of AMD Instinct MI450 Series GPUs in AMD Helios rack-scale solutions, with the first gigawatt deployment starting in the first half of 2027. The companies will also collaborate to use Claude to optimize workloads for AMD Instinct GPUs and accelerate ROCm development, and AMD will broadly adopt Claude across its engineering and product development teams. AMD committed to a strategic equity investment of up to $5 billion in Anthropic.

### OpenAI — How news organizations are using AI to advance their vital missions

- Company: OpenAI (openai.com)
- Announced: 2026-07-22T13:00:00+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://openai.com/index/how-news-organizations-are-using-ai
- Record: https://forck.live/items/16-how-news-organizations-are-using-ai-to-advance-their-vital-missions
- Subject: GPT / ChatGPT / API

OpenAI describes how news organizations use its AI tools to enhance reporting, audience growth, and business operations.

### OpenAI — Building AI infrastructure with the Effingham County community

- Company: OpenAI (openai.com)
- Announced: 2026-07-22T13:00:00+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://openai.com/index/building-ai-infrastructure-with-the-effingham-county-community
- Record: https://forck.live/items/15-building-ai-infrastructure-with-the-effingham-county-community
- Subject: GPT / ChatGPT / API

OpenAI announced Project Camellia, a community infrastructure initiative in Effingham County, Georgia, focused on responsible energy, community investment, job creation, and providing access to Codex.

### OpenAI — Advancing the next era of national science

- Company: OpenAI (openai.com)
- Announced: 2026-07-22T12:00:00+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://openai.com/index/advancing-the-next-era-of-national-science
- Record: https://forck.live/items/17-advancing-the-next-era-of-national-science
- Subject: GPT / ChatGPT / API

OpenAI states its commitment to advancing American science in collaboration with the U.S. Department of Energy and national labs by using frontier AI to speed up discovery.

### OpenAI — Introducing OpenAI Presence

- Company: OpenAI (openai.com)
- Announced: 2026-07-22T05:30:00+00:00
- Category: product-launch
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://openai.com/index/introducing-openai-presence
- Record: https://forck.live/items/18-introducing-openai-presence
- Subject: GPT / ChatGPT / API

OpenAI announces OpenAI Presence, a new enterprise AI agent platform for deploying voice and chat agents.

### Upstage — Solar Open 2: Korea's Sovereign Foundation Model, Built for Agentic Use

- Company: Upstage (upstage.ai)
- Announced: 2026-07-22
- Category: open-weight-release
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://www.upstage.ai/blog/en/solar-open-2
- Record: https://forck.live/items/18454-solar-open-2-korea-s-sovereign-foundation-model-built-for-agentic-use
- Subject: Solar Pro / Solar Mini

Upstage released Solar Open 2, an open-weight foundation model optimized for agentic use. The model uses a 250B total / 15B active MoE architecture, supports a 1M-token context window, and runs on two H200 GPUs with quantization. It is trained on agent scenarios including tool calling, coding, and office work, and officially supports Korean, English, and Japanese.

### Manus — Introducing Plan Mode: A New Way to Work with Manus

- Company: Manus (manus.im)
- Announced: 2026-07-22
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://manus.im/blog/manus-plan-mode
- Record: https://forck.live/items/14704-introducing-plan-mode-a-new-way-to-work-with-manus
- Subject: Manus

Manus introduced Plan Mode, a manual feature that pauses execution to generate a structured plan for review and approval before building. It can be activated at any point during a task and is available on web and mobile for all users.

### Luma — Luma Runs Production Inference on AMD and Tensorwave

- Company: Luma (lumalabs.ai)
- Announced: 2026-07-22
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://lumalabs.ai/news/luma-runs-production-inference-on-amd-and-tensorwave
- Record: https://forck.live/items/7960-luma-runs-production-inference-on-amd-and-tensorwave
- Subject: Ray / Dream Machine

Luma is running production inference for all its major models on AMD Instinct MI325X GPUs via Tensorwave. The company migrated its video-to-video pipeline and multimodal model to AMD hardware with minimal porting effort and no quality regression, thanks to AMD's PyTorch and software ecosystem.

### Amp — Multiplayer

- Company: Amp (ampcode.com)
- Announced: 2026-07-22
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://ampcode.com/news/multiplayer
- Record: https://forck.live/items/7723-multiplayer
- Subject: Amp

Amp introduced Multiplayer, a feature that allows users to turn any orb into a collaborative environment where workspace members can join threads, send messages to agents, and access the orb's portal, file changes, and shared terminal.

### Cognition — Cognition Signs MOU with U.S. Department of Energy to Join The Genesis Mission

- Company: Cognition (cognition.com)
- Announced: 2026-07-22
- Category: partnership-acquisition
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://cognition.com/blog/cognition-doe-genesis-mission
- Record: https://forck.live/items/7643-cognition-signs-mou-with-u-s-department-of-energy-to-join-the-genesis-mission
- Subject: Devin / SWE

Cognition has signed a memorandum of understanding with the U.S. Department of Energy to join the Genesis Mission, a national initiative that connects supercomputers and federal scientific datasets. Cognition will contribute its Devin AI software engineer in areas including software security, legacy code modernization, scientific workforce capacity, and cloud modernization.

### xAI — Bringing Grok 4.5 to iOS, Android, Web, and X

- Company: xAI (x.ai)
- Announced: 2026-07-22
- Category: availability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://x.ai/news/grok-4-5-everywhere
- Record: https://forck.live/items/3438-bringing-grok-4-5-to-ios-android-web-and-x
- Subject: Grok

Grok 4.5 is now available on grok.com, X, iOS, and Android.

### Anthropic — Ask Claude about the Anthropic Economic Index

- Company: Anthropic (anthropic.com)
- Announced: 2026-07-22
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://www.anthropic.com/news/anthropic-economic-index-connector
- Record: https://forck.live/items/3257-ask-claude-about-the-anthropic-economic-index
- Subject: Claude

Anthropic launched a connector for Claude that allows users to query the Anthropic Economic Index data directly within conversations, enabling questions about AI usage and automation patterns.

### Anthropic — A research agenda for the Economic Futures Research Fund

- Company: Anthropic (anthropic.com)
- Announced: 2026-07-22
- Category: funding-company-news
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://www.anthropic.com/news/economic-futures-research-fund-agenda
- Record: https://forck.live/items/3256-a-research-agenda-for-the-economic-futures-research-fund
- Subject: Claude

Anthropic announced the research agenda for its Economic Futures Research Fund, committing $200 million to support external research on interventions to prepare society for the economic impacts of AI, with five priority research areas: shaping AI's impact on workers, equipping people for AI-driven transitions, modernizing income support, building worker stakes in AI-driven growth, and generating evidence on public investments.

### Cursor — Cursor Router

- Company: Cursor (cursor.com)
- Announced: 2026-07-22
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://cursor.com/changelog/router
- Record: https://forck.live/items/2590-cursor-router
- Subject: Cursor

Cursor introduces Cursor Router, an intelligent model router that analyzes each request and routes it to the appropriate model based on optimization modes (Intelligence, Balance, Cost). It is available across platforms and on by default for Teams plans. Admins can control per-team enablement, mode restrictions, and model allow/block lists.

### OpenAI — NTT DATA Group cuts incident analysis to 30 minutes with Codex

- Company: OpenAI (openai.com)
- Announced: 2026-07-22
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://openai.com/index/ntt-data
- Record: https://forck.live/items/19-ntt-data-group-cuts-incident-analysis-to-30-minutes-with-codex
- Subject: GPT / ChatGPT / API

NTT DATA Group uses ChatGPT Enterprise and Codex to automate work, reduce incident analysis time to 30 minutes, and scale secure AI adoption for 9,000 employees.

### NVIDIA — Built in Fort Worth: Wistron Opens Advanced Manufacturing Plant to Produce NVIDIA AI Systems

- Company: NVIDIA (nvidia.com)
- Announced: 2026-07-21T22:35:45+00:00
- Category: infrastructure-release
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://blogs.nvidia.com/blog/wistron-manufacturing-texas/
- Record: https://forck.live/items/1437-built-in-fort-worth-wistron-opens-advanced-manufacturing-plant-to-produce
- Subject: AI platform

Wistron opened a new manufacturing facility in Fort Worth, Texas, to produce NVIDIA AI systems, including the GB300 Grace Blackwell Ultra Superchip and the Vera Rubin Superchip. The plant represents a $700 million investment and has created over 500 jobs, with plans to expand to 1,000 by year-end. NVIDIA committed to manufacturing up to $500 billion in AI platforms in the U.S.

### OpenAI — Introducing the ChatGPT for small business program

- Company: OpenAI (openai.com)
- Announced: 2026-07-21T17:00:00+00:00
- Category: availability-change
- Coverage: 2 outlets
- Announcement: yes
- Group: covered
- Source: https://openai.com/index/introducing-chatgpt-small-business-program
- Record: https://forck.live/items/20-introducing-the-chatgpt-for-small-business-program
- Subject: GPT / ChatGPT / API

OpenAI launches the ChatGPT for Small Businesses program, which helps entrepreneurs build AI skills, automate work, and grow with ChatGPT Work.

### NVIDIA — NVIDIA Vera Rubin Driving Performance Per Watt, Lowest Token Cost for Partners Worldwide

- Company: NVIDIA (nvidia.com)
- Announced: 2026-07-21T15:36:43+00:00
- Category: new-model
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://blogs.nvidia.com/blog/vera-rubin/
- Record: https://forck.live/items/1438-nvidia-vera-rubin-driving-performance-per-watt-lowest-token-cost-for-partners
- Subject: AI platform

NVIDIA announces the Vera Rubin platform, a chip-to-grid AI supercomputer featuring the Vera CPU and six other chips, claiming 10x more tokens per megawatt and one-tenth the cost per million tokens compared to GB200 NVL72, with partners like CoreWeave, Google Cloud, Microsoft Azure, Oracle Cloud Infrastructure, and Nebius ramping production. The post also mentions Mistral Medium 3.5 and OCR 4 models available in Microsoft Foundry, and a new partnership between Microsoft and Mistral for European AI infrastructure using thousands of Vera Rubin GPUs.

### Google DeepMind — Introducing Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber

- Company: Google DeepMind (deepmind.google)
- Announced: 2026-07-21T15:16:30+00:00
- Category: new-model
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://deepmind.google/blog/introducing-gemini-3-6-flash-3-5-flash-lite-and-3-5-flash-cyber/
- Record: https://forck.live/items/1204-introducing-gemini-3-6-flash-3-5-flash-lite-and-3-5-flash-cyber
- Subject: Gemini

Google DeepMind announces three new Gemini models: 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber, with improved efficiency, lower cost, and specialized capabilities for agentic workflows.

### NVIDIA — Built for Vera Rubin, NVIDIA Spectrum-6 Arrives in Gigascale AI Factories

- Company: NVIDIA (nvidia.com)
- Announced: 2026-07-21T15:00:20+00:00
- Category: infrastructure-release
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://blogs.nvidia.com/blog/nvidia-spectrum-six-arrives-in-gigascale-ai-factories/
- Record: https://forck.live/items/1439-built-for-vera-rubin-nvidia-spectrum-6-arrives-in-gigascale-ai-factories
- Subject: AI platform

NVIDIA announces the arrival of Spectrum-6, a 102.4-terabit-per-second Ethernet switch system built for the Vera Rubin platform, in gigascale AI factories. The switch anchors the next generation of the Spectrum-X Ethernet platform and is being adopted by leading AI infrastructure builders including CoreWeave, Microsoft, Nebius, SpaceXAI, and Tesla.

### NVIDIA — NVIDIA and Partners Build in America, for America

- Company: NVIDIA (nvidia.com)
- Announced: 2026-07-21T13:00:47+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://blogs.nvidia.com/blog/nvidia-and-partners-build-in-america-for-america/
- Record: https://forck.live/items/1440-nvidia-and-partners-build-in-america-for-america
- Subject: AI platform

NVIDIA announced its participation in the U.S. National Science Foundation's State and Regional AI Infrastructure Hubs program to expand AI research and education. Partner Wistron opened a manufacturing plant in Fort Worth, Texas, producing the NVIDIA GB300 Grace Blackwell Ultra Superchip and preparing to produce the NVIDIA Vera Rubin Superchip, part of a $700 million commitment creating over 500 jobs. NVIDIA and its partners are investing in American manufacturing across 43 states, with plans to produce up to $500 billion of AI infrastructure in the U.S.

### Google Antigravity — Gemini 3.6 Flash in Google Antigravity

- Company: Google Antigravity (antigravity.google)
- Announced: 2026-07-21T08:00:00+00:00
- Category: new-model
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://antigravity.google/blog/gemini-3-6-flash-in-google-antigravity
- Record: https://forck.live/items/4547-gemini-3-6-flash-in-google-antigravity
- Subject: Antigravity

Google Antigravity announces Gemini 3.6 Flash, a new model with improved coding, knowledge work, multimodal performance, and reduced token usage by up to 17% compared to 3.5 Flash.

### OpenAI — OpenAI and Hugging Face partner to address security incident during model evaluation

- Company: OpenAI (openai.com)
- Announced: 2026-07-21T07:00:00+00:00
- Category: not stated
- Coverage: 3 outlets
- Announcement: no
- Group: covered
- Source: https://openai.com/index/hugging-face-model-evaluation-security-incident
- Record: https://forck.live/items/21-openai-and-hugging-face-partner-to-address-security-incident-during-model
- Subject: GPT / ChatGPT / API

OpenAI and Hugging Face jointly disclosed early findings from a security incident that occurred during an AI model evaluation, noting advanced cyber capabilities and offering lessons for defenders.

### Alibaba — Qwen-Image-3.0: Rich Content, Authentic Details, Deep Knowledge

- Company: Alibaba (alibaba.com)
- Announced: 2026-07-21T06:00:00+00:00
- Category: new-model
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://qwen.ai/blog?id=qwen-image-3.0
- Record: https://forck.live/items/4617-qwen-image-3-0-rich-content-authentic-details-deep-knowledge
- Subject: Qwen

Alibaba launches Qwen-Image-3.0, a third-generation foundational image generation model that supports up to 4.5k token input, precise rendering of text as small as 10px, native rendering of 12 languages, and can generate complex layouts such as newspapers, storyboards, and exam papers in a single pass.

### Vercel — Introducing the new Vercel Agent

- Company: Vercel (vercel.com)
- Announced: 2026-07-21T04:00:00+00:00
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://vercel.com/blog/vercel-agent
- Record: https://forck.live/items/20685-introducing-the-new-vercel-agent
- Subject: v0 / AI SDK / AI Gateway

Vercel expanded its Agent from triaging alerts and reviewing pull requests to a dashboard-based tool that investigates production issues, answers project questions, and takes approved actions. The agent operates under its own identity, is read-only by default, and uses a plan-to-permission model where it requests scoped access for each task, dropping back to read-only after completion. Generated code runs in an ephemeral Firecracker microVM sandbox isolated from live systems.

### Vercel — How Searchable ships customer-requested features in 30 minutes on Vercel

- Company: Vercel (vercel.com)
- Announced: 2026-07-21T04:00:00+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://vercel.com/blog/how-searchable-ships-customer-requested-features-in-30-minutes-on-vercel
- Record: https://forck.live/items/20684-how-searchable-ships-customer-requested-features-in-30-minutes-on-vercel
- Subject: v0 / AI SDK / AI Gateway

Searchable, a platform for brand visibility in AI search engines, uses Vercel's AI SDK and AI Gateway to test new models without rotating API keys or SDKs, achieving 2 to 5x faster development and shipping customer-requested features in as little as 30 minutes.

### Tencent — Hyra: A simple yet effective scaffold for general discovery

- Company: Tencent (tencent.com)
- Announced: 2026-07-21T02:00:00+00:00
- Category: research-paper
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://hunyuan.tencent.com/research/hyra
- Record: https://forck.live/items/4577-hyra-a-simple-yet-effective-scaffold-for-general-discovery
- Subject: Hunyuan

Tencent introduces Hyra, a simple yet effective system for automated research and recursive self-improvement. Hyra uses a Harness with a Context Agent and Proposal Agents to explore solutions across diverse environments. It outperforms Recursive's system on NanoChat Autoresearch, NanoGPT Speedrun, and SOL-ExecBench benchmarks.

### Supabase — Supabase Pipelines is now in Public Alpha

- Company: Supabase (supabase.com)
- Announced: 2026-07-21
- Category: not stated
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://supabase.com/blog/supabase-pipelines-public-alpha
- Record: https://forck.live/items/21279-supabase-pipelines-is-now-in-public-alpha
- Subject: Supabase

Update: October 2, 2026. BigQuery, ClickHouse, DuckLake, and Snowflake are now available in public alpha.

### Gradium — Keyword boosting: teaching Speech-To-Text your vocabulary

- Company: Gradium (gradium.ai)
- Announced: 2026-07-21
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://gradium.ai/blog/stt-keyword-boosting
- Record: https://forck.live/items/18310-keyword-boosting-teaching-speech-to-text-your-vocabulary
- Subject: Gradium TTS / Phonon

Gradium's guide explains keyword boosting for its Speech-To-Text model, which allows users to supply a dictionary of rare terms (e.g., brand names, athletes, drugs) that the model then biases toward during decoding without retraining. The recommended boost parameter is 3, with 4 as a ceiling for dense or foreign vocabulary. The dictionary supports up to 500 single-token words, and case and accents must be matched exactly.

### Liquid AI — Tokenizer Expansion: Upgrading a Model's Tokenizer in Place

- Company: Liquid AI (liquid.ai)
- Announced: 2026-07-21
- Category: model-update
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://www.liquid.ai/blog/tokenizer-expansion
- Record: https://forck.live/items/16778-tokenizer-expansion-upgrading-a-model-s-tokenizer-in-place
- Subject: LFM / d1 models

Liquid AI upgraded the tokenizer of LFM2.5-8B-A1B from 65K to 128K vocabulary without retraining from scratch. The expanded tokenizer reduces tokens for Hindi (2.4×), Vietnamese (2.6×), Thai (up to 4.0×), and Bengali (3.4×), leading to an estimated 2.2–3.7× faster per-character decoding on-device for those languages. The method, detailed in a technical report, preserves quality on already-supported languages after a two-stage adaptation (embeddings-only followed by full continued pre-training), and the model is available with open weights on Hugging Face.

### xAI — Introducing Grok for Outlook

- Company: xAI (x.ai)
- Announced: 2026-07-21
- Category: not stated
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://x.ai/news/introducing-outlook-addin
- Record: https://forck.live/items/16458-introducing-grok-for-outlook
- Subject: Grok

Grok for Outlook Use the Grok add-in for Microsoft Outlook to summarize long email threads, draft replies in your voice, and keep your inbox organized. Jul 21, 2026

### Amp — Right on Schedule

- Company: Amp (ampcode.com)
- Announced: 2026-07-21
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://ampcode.com/news/schedule
- Record: https://forck.live/items/7724-right-on-schedule
- Subject: Amp

Amp introduced the ability for agents to set their own schedules and wake themselves up with saved prompts, context, and history. This capability works with Slack, Puck, and other agents.

### World Labs — World Labs Acquires SceniX

- Company: World Labs (worldlabs.ai)
- Announced: 2026-07-21
- Category: partnership-acquisition
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://www.worldlabs.ai/blog/scenix
- Record: https://forck.live/items/7348-world-labs-acquires-scenix
- Subject: Marble / Atlas

World Labs announced that SceniX, a robotics company, is joining World Labs to advance spatial intelligence in robotics.

### xAI — Grok for Outlook

- Company: xAI (x.ai)
- Announced: 2026-07-21
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://x.ai/news/introducing-outlook-addin
- Record: https://forck.live/items/3439-grok-for-outlook
- Subject: Grok

xAI announced a Grok add-in for Microsoft Outlook that can summarize long email threads, draft replies, and organize the inbox.

### Anthropic — Anthropic is donating another $20 million to Public First Action

- Company: Anthropic (anthropic.com)
- Announced: 2026-07-21
- Category: funding-company-news
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://www.anthropic.com/news/donation-public-first-action
- Record: https://forck.live/items/3258-anthropic-is-donating-another-20-million-to-public-first-action
- Subject: Claude

Anthropic announces an additional $20 million donation to Public First Action, a nonpartisan organization focused on AI public education and policy, bringing total support to $40 million.

### Hugging Face — Grabette: an open system to record robot-manipulation data

- Company: Hugging Face (huggingface.co)
- Announced: 2026-07-21
- Category: developer-tool-release
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://huggingface.co/blog/grabette
- Record: https://forck.live/items/1453-grabette-an-open-system-to-record-robot-manipulation-data
- Subject: Platform

Hugging Face and Pollen Robotics announce Grabette, an open, low-cost handheld system for recording robot manipulation data, along with its robotic counterpart Gripette. The system includes hardware designs, capture software, a browser-based processing pipeline, and outputs standard LeRobot datasets on the Hugging Face Hub.

### OpenAI — David Vélez and Robin Vince join the boards of the OpenAI Foundation and OpenAI Group PBC

- Company: OpenAI (openai.com)
- Announced: 2026-07-21
- Category: funding-company-news
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://openai.com/index/david-velez-robin-vince-join-openai-boards
- Record: https://forck.live/items/22-david-v-lez-and-robin-vince-join-the-boards-of-the-openai-foundation-and
- Subject: GPT / ChatGPT / API

David Vélez and Robin Vince have joined the boards of the OpenAI Foundation and OpenAI Group PBC.

### Perplexity — Sonar Chat Completions migration

- Company: Perplexity (perplexity.ai)
- Announced: 2026-07-20T21:52:47+00:00
- Category: api-release
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://docs.perplexity.ai/docs/resources/changelog#july-2026
- Record: https://forck.live/items/5070-sonar-chat-completions-migration
- Subject: Sonar

Sonar Chat Completions has been renamed to Agent API, with added models, tools, and research-backed presets.

### Databricks — Scaling document classification to 100k+ labels

- Company: Databricks (databricks.com)
- Announced: 2026-07-20T18:15:00+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://www.databricks.com/blog/scaling-document-classification-100k-labels
- Record: https://forck.live/items/17503-scaling-document-classification-to-100k-labels
- Subject: Mosaic AI

Databricks describes a solution for mapping text to taxonomies of 100,000+ labels that pairs vector search with the AI Classify function. The approach retrieves a shortlist of candidate labels per document and then uses AI Classify to pick from that shortlist. Across three benchmarks, SQL-native vector search plus AI Classify outperformed the best cost-efficient frontier model by five points of accuracy at roughly one-hundredth of the token cost.

### Runway — The AI Media Report: Cost, Speed and What Comes Next

- Company: Runway (runway.com)
- Announced: 2026-07-20T17:40:49.930000+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://runway.com/news/company-news/ai-media-report
- Record: https://forck.live/items/7865-the-ai-media-report-cost-speed-and-what-comes-next
- Subject: Gen / Aleph

Runway published a longitudinal study of its impact on enterprise customers, finding that production costs typically fall by two to three orders of magnitude and timelines collapse. The study reports specific savings from customers such as a national financial services brand that reduced a $5M broadcast commercial to $3K-4K, and a home goods retailer that replicates $800K visual projects for under $10K. Speed improvements include a global agency compressing social content production from 2-3 months to 3 hours.

### NVIDIA — At SIGGRAPH, NVIDIA Advances Graphics and Simulation With Agentic and Physical AI

- Company: NVIDIA (nvidia.com)
- Announced: 2026-07-20T15:00:53+00:00
- Category: new-model
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://blogs.nvidia.com/blog/siggraph-news-2026/
- Record: https://forck.live/items/1441-at-siggraph-nvidia-advances-graphics-and-simulation-with-agentic-and-physical-ai
- Subject: AI platform

NVIDIA announced Cosmos 3 Edge, a 4-billion-parameter open world model built to run in real time on device.

### Sakana AI — Introducing Fugu-Cyber: our new orchestration model that achieves state-of-the-art performance on real-world cybersecurity benchmarks

- Company: Sakana AI (sakana.ai)
- Announced: 2026-07-20T15:00:00+00:00
- Category: model-update
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://sakana.ai/fugu-cyber-release
- Record: https://forck.live/items/4675-introducing-fugu-cyber-our-new-orchestration-model-that-achieves-state-of-the
- Subject: Sakana models

Sakana AI released Fugu-Cyber, an update to its Fugu orchestration model, available as a new API endpoint. It achieves state-of-the-art performance on cybersecurity benchmarks with a success rate of 86.9% on CyberGym and 72.1% on CTI-REALM.

### AMD — Microsoft to Deploy Next-Gen AMD Instinct and AMD EPYC Processors as the Companies Expand Their Long-Term Strategic Partnership

- Company: AMD (amd.com)
- Announced: 2026-07-20T13:00:00+00:00
- Category: partnership-acquisition
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://ir.amd.com/news-events/press-releases/detail/1291/microsoft-to-deploy-next-gen-amd-instinct-and-amd-epyc-processors-as-the-companies-expand-their-long-term-strategic-partnership
- Record: https://forck.live/items/15038-microsoft-to-deploy-next-gen-amd-instinct-and-amd-epyc-processors-as-the
- Subject: AMD Instinct / ROCm

AMD and Microsoft expanded their strategic partnership, with Microsoft set to deploy the AMD Helios Rackscale Solution on Azure to power frontier model AI inference for Microsoft, its AI customers and Azure AI services. Azure will add two new VM series powered by 6th Gen AMD EPYC

### NVIDIA — Bristol Myers Squibb Building Life Science Industry’s Most Advanced AI Factory on NVIDIA Vera Rubin

- Company: NVIDIA (nvidia.com)
- Announced: 2026-07-20T10:59:23+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://blogs.nvidia.com/blog/bristol-myers-squibb-building-life-science-industrys-most-advanced-ai-factory-on-nvidia-vera-rubin/
- Record: https://forck.live/items/1442-bristol-myers-squibb-building-life-science-industry-s-most-advanced-ai-factory
- Subject: AI platform

Bristol Myers Squibb (BMS) announced the deployment of a second NVIDIA DGX SuperPOD built on eight DGX Vera Rubin NVL72 systems, described as the most powerful and energy-efficient AI cluster in life sciences. The system, which includes NVIDIA Vera CPUs and Rubin GPUs, will be available to all scientists at BMS, enabling faster drug discovery cycles, larger chemical spaces, and agentic workflows. It also incorporates the NVIDIA BioNeMo Agent Toolkit for biological AI and is managed via NVIDIA Mission Control. The existing DGX SuperPOD will be combined into a unified environment accessible from all BMS sites globally.

### OpenAI — Safety and alignment in an era of long-horizon models

- Company: OpenAI (openai.com)
- Announced: 2026-07-20T10:00:00+00:00
- Category: safety-policy-update
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://openai.com/index/safety-alignment-long-horizon-models
- Record: https://forck.live/items/23-safety-and-alignment-in-an-era-of-long-horizon-models
- Subject: GPT / ChatGPT / API

OpenAI published a post sharing lessons from deploying long-running AI models, focusing on safety risks, observed failures, and improved safeguards.

### Genspark — Introducing Genspark AI Workspace 6.0

- Company: Genspark (genspark.ai)
- Announced: 2026-07-20
- Category: not stated
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://www.genspark.ai/blog/genspark-ai-workspace-6
- Record: https://forck.live/items/18250-introducing-genspark-ai-workspace-6-0
- Subject: Genspark

### xAI — Introducing Grok for Excel

- Company: xAI (x.ai)
- Announced: 2026-07-20
- Category: not stated
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://x.ai/news/introducing-excel-addin
- Record: https://forck.live/items/16459-introducing-grok-for-excel
- Subject: Grok

Grok for Excel Use the Grok add-in for Microsoft Excel to ask questions in plain English, write formulas, and run scenarios without leaving the workbook. Jul 20, 2026

### LG AI Research — LG and UNESCO Globally Launch the Global MOOC on Ethics of AI Project

- Company: LG AI Research (lgresearch.ai)
- Announced: 2026-07-20
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://www.lgresearch.ai/news/view?seq=668
- Record: https://forck.live/items/10987-lg-and-unesco-globally-launch-the-global-mooc-on-ethics-of-ai-project
- Subject: EXAONE

LG AI Research and UNESCO launched the Global MOOC on Ethics of AI, a free educational program on Coursera based on UNESCO's 2021 Recommendation on the Ethics of Artificial Intelligence. The 10-module curriculum, developed with leading universities and regional organizations, targets AI professionals, researchers, and policymakers. The partners plan to expand it into a professional training program with case-based learning and hands-on exercises.

### Amp — Amp Is Now In Slack

- Company: Amp (ampcode.com)
- Announced: 2026-07-20
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://ampcode.com/news/slack-integration
- Record: https://forck.live/items/7726-amp-is-now-in-slack
- Subject: Amp

Amp announced that its AI coding assistant can now be summoned in Slack. Users can mention @Amp in any Slack channel or thread to initiate bug fixes, spike new features, answer questions about the codebase, and find and manipulate existing threads. The integration also allows Amp to read screenshots, reproduce issues, and post fixes.

### Amp — Meet Puck

- Company: Amp (ampcode.com)
- Announced: 2026-07-20
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://ampcode.com/news/meet-puck
- Record: https://forck.live/items/7725-meet-puck
- Subject: Amp

Amp introduced Puck, a new assistant within its platform that is always available and can help users manage agents, spawn new agents, and coordinate tasks. Puck is described as an experiment and a home base for launching and coordinating other agents.

### Cognition — Welcoming TierZero to Cognition

- Company: Cognition (cognition.com)
- Announced: 2026-07-20
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://cognition.com/blog/welcoming-tierzero
- Record: https://forck.live/items/7644-welcoming-tierzero-to-cognition
- Subject: Devin / SWE

Cognition welcomed the TierZero team, including Anhang and Yun, and will integrate their automation work into Devin.

### LG AI Research — LG and UNESCO Globally Launch the Global MOOC on Ethics of AI Project

- Company: LG AI Research (lgresearch.ai)
- Announced: 2026-07-20
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://www.lgresearch.ai/blog/view?seq=668
- Record: https://forck.live/items/4741-lg-and-unesco-globally-launch-the-global-mooc-on-ethics-of-ai-project
- Subject: EXAONE

LG AI Research and UNESCO launched a free global MOOC on AI ethics, developed in partnership with Coursera, based on UNESCO's Recommendation on the Ethics of Artificial Intelligence, with modules from leading institutions.

### xAI — Grok for Excel

- Company: xAI (x.ai)
- Announced: 2026-07-20
- Category: product-launch
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://x.ai/news/introducing-excel-addin
- Record: https://forck.live/items/3440-grok-for-excel
- Subject: Grok

xAI announces Grok for Excel, an add-in that allows users to ask questions in plain English, write formulas, and run scenarios within Microsoft Excel.

### Anthropic — Apply for Anthropic’s AI for Science rare disease research grants

- Company: Anthropic (anthropic.com)
- Announced: 2026-07-20
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://www.anthropic.com/news/rare-disease-research-grants
- Record: https://forck.live/items/3259-apply-for-anthropic-s-ai-for-science-rare-disease-research-grants
- Subject: Claude

Anthropic announces a call for applications for rare disease research grants under the AI for Science program, offering up to $50,000 in Claude credits over six months across two tracks: basic science and early-stage biotech partnerships, with the aim of building a community to explore AI's role in rare disease research and accelerate drug development and discovery of disease mechanisms. Key points: - Two tracks: basic science (scaling partnerships with researchers, patient organizations, data scientists) and biotech (supporting early-stage biotechs to accelerate drug development). - Grants include up to $50,000 in Claude credits. - Partners include Monarch Initiative, using resources like Mondo Disease Ontology, Monarch Knowledge Graph, and DisMech. - Goal: build a community of researchers, improve AI modeling of rare diseases, and compress phases of drug development through documentation, therapeutic strategy selection, and basket trials. Evidence excerpt: "Today, we are sharing a focused call for applications centered specifically on rare genetic diseases. Accepted applicants will receive up to $50,000 in Claude credits over six months, with the goal of building a community of researchers looking into how AI can reshape our understanding of rare disease." (from the source text). The source text does not specify a new model, product launch, or API release; it is a grant program announcement, so category is 'other'. Models affected are Claude (the AI model) as mentioned in the credits and access. No license, context window, or pricing is stated. Date is given as Jul 20, 2026 (implied year from the URL and text). The source is Anthropic's official blog news page; the title is the same as the provided title. The URL is provided. The evidence excerpt is a verbatim copy from the source text. The summary is in English and covers the main points without adding external knowledge. The category is 'other' because this is a grant program announcement, not fitting the other specific categories. The models_affected list includes only 'Claude' as it appears in the text. All other fields are left empty as they are not stated in the source text. The date is extracted from the text. The provider is given as Anthropic. The title is the same as the provided title. The URL is the source URL. The summary is self-contained and does not introduce any external information. The evidence_excerpt supports the summary. The JSON is formatted as per the schema. The output is a valid JSON object. The response length is acceptable. The thinking process is included in the final answer as a comment? The user said 'Answer only with JSON matching the provided schema.' So I should output only the JSON, no extra text. The JSON should be well-formed. The evidence_excerpt is a verbatim quote from the source text. The models_affected list is correct. The category is 'other' as it's not a product launch, model update, API release, etc. The date is exactly as in the source. The license, context_window, pricing_note are empty strings. The is_open_weight is false. The summary is concise but covers the key points. The output is JSON. I will now output the JSON. The JSON must be valid. I will ensure no trailing commas. The JSON has keys in the order specified in the schema. The summary includes a brief summary and then key points, but the schema says 'summary' should be a single string. I'll make it a concise paragraph. Also, no extra text. I'll output only the JSON. The JSON should be on a single line for compactness, but it's okay to have line breaks. I'll write it as a single line to avoid any issues. Let me write the JSON. The date is

### Together AI — Together AI and Y Combinator partner to launch the first dedicated GPU cluster for the YC community

- Company: Together AI (together.ai)
- Announced: 2026-07-20
- Category: partnership-acquisition
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://www.together.ai/blog/together-yc-gpu-cluster
- Record: https://forck.live/items/2333-together-ai-and-y-combinator-partner-to-launch-the-first-dedicated-gpu-cluster
- Subject: Inference platform

Together AI and Y Combinator announce a partnership to deliver the first dedicated GPU cluster for YC portfolio companies, providing flexible, dedicated access to compute for AI startups.