Google Antigravity Remote Control
A browser-first interface for remotely controlling AI coding agents from any device.
ddNews Β· a daily AI briefing for Dirk (generative video/ddStudio, Lit
A browser-first interface for remotely controlling AI coding agents from any device.

An AI agent from DeepMind alumni that outperforms OpenAI/Anthropic in replicating scientific research.
A new unknown model is reportedly beating GPT-5.6 Sol and Claude Fable 5 in coding benchmarks.
Reports suggest both OpenAI and Anthropic are delaying Fable 5.1 and GPT Astra to outmaneuver each other.
The tool now supports the general availability of AI-generated music, voiceovers, and sound effects.
Key DeepMind researchers have left to form "Discovery Loop," impacting Alphabet's market cap.
A mysterious new model (possibly from Z.ai) is reportedly outperforming Fable and GPT-5.6 in early tests.
NORINCO introduces a system allowing humans to mirror movements to humanoid robots for high-risk missions.

New approach treats folders of unrelated PDFs as a single document with nested outlines for better retrieval.
Media reports suggest the company's market debut could challenge SpaceX's record-breaking valuation.
Media reports suggest Anthropic's market debut could break the SpaceX IPO record.
A new robot foundation model that can learn physical tasks from a single demonstration.
Anthropic's assistant can now read and reply to emails and manage files without per-action user approval.
Of 13 AI enforcement cases, all focus on deceptive marketing rather than autonomous agent behavior.
Google's open-weight series reaches a major adoption milestone.
Over 300 companies in Beijing are moving humanoid robots from exhibition stages into actual industrial testing.
Strategic restructuring hits Siri and Vision Pro teams to pivot toward new AI advancements.
Ojai robotaxis now feature Gemini-powered voice controls for cabin management and trip details.

Research from Princeton/UCSD shows structured action sequences (skills) drive agent success more than raw knowledge.
Coinbase reports a surge in AI agents using stablecoins for autonomous transactions.
Potential deal with the South Korean NPU startup focused on inference optimization.
Implementation of a multi-agent approach to handle code reviews at enterprise scale.

A new approach to AI world models that incorporates human beliefs and intentions rather than just physics.
ACE Robotics predicts a fundamental breakthrough in embodied AI brains by the end of 2027.

A proprietary language model is being tested to replace hand-built recommendation logic.
The Claude creator is reportedly seeking over $100B in funding with a target valuation of $2 trillion.
The AI data center builder is eyeing public markets following its acquisition of Anyscale.
For the first time, Anthropic has surpassed OpenAI in quarterly revenue and reached positive operating income.
Using AMD's MORI-IO connector to split prefill and decode delivers 2.5x higher goodput on MI300X nodes.
Google's next hardware iteration focuses heavily on Gemini Intelligence and AI-powered video tools.
The company now handles 40% of AI queries via open-source models to reduce costs, targeting 60-70%.
Lumentum reports 109% revenue growth as optical components become the primary data center constraint.
The company has hired a former Google chip executive to reduce dependency on external hardware.
Despite missing overall earnings, Kuaishou reports massive growth for its Kling AI video tool.
The new model shows 50% better coding performance without increasing model size.
First production systems featuring 88-core CPUs and HBM4 GPUs have been delivered.
API costs for the frontier model have been reduced by over 20% to counter pressure from Anthropic and DeepSeek.
Developer costs for the frontier GPT-5.6 Sol model have been reduced by more than 20%.
Prices for the frontier model were slashed by over 20% on August 21 to counter pressure from Anthropic and DeepSeek.
The frontier model's API costs are slashed by over 20% to counter pressure from Anthropic and DeepSeek.
API costs for the frontier Sol model have been reduced by over 20%.

Continued strategic investment in data center development to scale AI infrastructure.
The lab hired Google chip veteran Amir Salek to lead its push into custom AI silicon.
The open-weight model family has seen massive adoption with 100,000 variants created.
A four-month trial shows progress in industrial tasks like installing nuts and sorting panels.
New API pricing and access for AI video generation targeting business and marketing production.
The strongest model is now available via Claude Security for vetted cyber-defense use.
Robots are being trialed for hotel, household, and library tasks at the World Humanoid Robot Games.
Potential collaboration or deal with the Korean AI chip designer.
Potential collaboration with the Korean AI chip designer.
Enterprises can now build software using any LLM within a single governed platform.
A new framework for how AI agents interact with government services.
Funding for a major AI chip deal that could benefit Anthropic.

Anthropic's strongest model is now powering a security scanner for CWE-classified codebase vulnerabilities.
Dedicated collaborative channels where teams can oversee and review AI coding agents in real-time.

An experimental multimodal model for agents that rivals Opus 4.8 in benchmarks.
The AdaPT proof-of-concept shows humanoid robots playing professional-style tennis.
All Claude-generated text now includes watermarks to comply with EU AI Act regulations.

New methodology for turning Codex from an interactive chat into a programmable automation component.
A new tool for creating virtual influencers and animated K-Pop personas.
Researchers found a way to bypass safety guardrails in Grok and Gemini using encrypted prompts.

A technical shift toward retrieving individual table rows with headers rather than full pages to improve enterprise document intelligence.
Sundar Pichai confirms massive scale for AI Overviews.
Ramp data indicates a shift in enterprise spending, though Anthropic still holds a lead in total company adoption.
New multi-model coordination tools to move agents from conversational assistants to autonomous planning systems.
OpenAI and AWS bring these models to Amazon Bedrock with local data processing.

A survey indicates 75% of Americans now oppose the construction of AI data centers in their vicinity.

High-profile filmmakers are under fire for promoting the Higgsfield AI platform.

Washington is reportedly urging partner nations to choose between US and Chinese AI ecosystems.
The model can acquire new physical manipulation tasks from a 3-12 second clip without retraining.
Anthropic provides 20 free courses focusing on Claude Code and agent workflows.
The lab is hiring silicon experts to reduce its reliance on external hardware providers.
A new memory-centric, multimodal agentic platform combining video, music, and speech generation.
New security warnings highlight that current identity frameworks are insufficient for agents acting on behalf of humans.
Organizations are increasingly running open-weight models like DeepSeek and Qwen on local servers.
A unified platform launched in Malaysia to connect enterprises to a wide array of LLMs.
A developer platform allowing AI agents to trade crypto autonomously via API and wallet hubs.
Partnering with Nvidia to deploy next-gen chips on satellites for space-based AI processing.

The robotaxi firm is building its own silicon to reduce dependency on Nvidia.
Generate Music, Speech, and Sound Effects are now generally available.
New native 4K AI video enhancement is now available for creative workflows.
A new "codex-like" approach to video production moving beyond single-clip generation.
AgentAnywhere allows banks and governments to run AI agents entirely on-prem.
Native 4K AI video enhancement is now available via Black Forest Labs' FLUX technology.

A policy change allowing corporate clients to retain their own data for Mythos models.
The lab is filing paperwork for a public offering that could challenge SpaceX's valuation records.

OpenAI reports 35% overall revenue growth and 50% enterprise growth since the July launch.

A $6 billion deal to absorb software for AI model development and 109 employees.
GPT-5.6 Terra and Luna are now available on Amazon Bedrock in India to keep processing local.

OpenAI's API now allows the direct generation of images with alpha channels (no background).
New reports highlight how university labs in China are using "ingenuity and imitation" to chase OpenAI and Anthropic.
New agent-driven tools for e-commerce and UGC production using Seedance 2.5.
General availability of music, speech, and sound effects generation within the Firefly ecosystem.
Enterprise AI adoption is now heavily dependent on strict agent data residency requirements.
Hackers stole publishing keys of a popular AI tool to leak terabytes of sensitive organizational data.
Up to 30,000 humanoid units have been shipped so far in 2026.

A new tool for publishers to fight traffic losses caused by AI Overviews.

A new model routing service allowing companies to switch between LLMs seamlessly.

An experimental AI app for "vibe-coding" and sharing interactive games.
Analysis of three distinct corpus shapes and the specific architectural costs associated with each.

A new system-wide tool allowing users to "talk to their apps" via AI.
New approach to retrieval quality using graph traversal and bitemporal edges instead of simple query wording.

The new Agent OS integrates with ChatGPT, Claude Code, and Cursor for automated trading.

Dedicated collaborative spaces where teams can build with AI agents in real-time.