🤖 DAILY AI NEWS DIGEST

01Releases · openclaw/openclaw · GitHub 2026-07-19🔍 ExaAI

• ▸ OpenClaw 2026.7.2 introduces remote coding sessions allowing Control UI sessions on cloud workers and terminal sessions for Codex, Claude, OpenCode, and Pi.
• ▸ It adds native automation parity to mobile, including Android Voice Wake and exposes camera, location, and notification features from headless Linux nodes.
• ▸ The release improves channel operation safety by preventing Telegram ingress loss, maintaining Signal control responsiveness, and restricting channel allowlist owner access.
• ▸ Guided Control UI setup enables configuring model providers, onboarding channels via a setup page, and selecting images/models when creating sessions.
• ▸ Packaging enhancements include Linux deb and AppImage bundles with Gateway guidance and seamless Windows installs continuing after winget adds Node.js.

github.com

02@simonw: If you have Claude Code installed you're running software that uses the new (unreleased) version of Bun that's... 2026-07-19𝕏 X (trusted)

140 likes, 8 reposts | If you have Claude Code installed you're running software that uses the new (unreleased) version of Bun that's been rewritten in Rust - here are two commands you can run to see that for yourself https://t.co/7AG2iZTvB2

x.com

03ChatGPT — Release Notes 2026-07-19🔍 ExaAI

• ▸ Launched an updated ChatGPT desktop app for macOS and Windows that integrates Chat, Work, and Codex in a unified interface with features like a global switcher, unified recents, project management, and cross-device syncing of cloud conversations.
• ▸ Increased custom instructions character limit to 5,000 for Pro, Enterprise, Business, and Education users to allow more detailed customization of ChatGPT’s behavior.
• ▸ Introduced a supercharged search feature across chats, projects, images, and documents accessible from the ChatGPT sidebar on web, iOS, and Android.
• ▸ Released ChatGPT Sites in public beta, enabling users to create interactive websites or lightweight apps directly from ChatGPT with preview and editing capabilities before publishing.
• ▸ Announced the deprecation of Atlas browser-based agentic capabilities, transitioning browser data handling and enhanced browser experience features into ChatGPT and Codex, with a shutdown planned for August 9, 2026.

help.openai.com

04Fabric Updates Blog 2026-07-18🔍 ExaAI

• ▸ SQL Server 2025 introduces governed AI connectivity, allowing database-side workloads to securely connect to external AI services for scenarios like retrieval-augmented generation (RAG).
• ▸ AI calls from SQL Server 2025 are enforced within its existing trust and control security model, maintaining governance and access control closer to the data platform.
• ▸ This integration reduces the need for custom middleware and ad-hoc data exports, minimizing data movement and improving security policy consistency and auditability.

community.fabric.microsoft.com

05An important update: Transitioning Gemini CLI to Antigravity CLI · google-gemini gemini-cli · Discussion #27274 · GitHub 2026-07-18🔍 ExaAI

• ▸ Google is transitioning its Gemini CLI tool to a new Antigravity CLI, designed to support complex, multi-agent workflows with improved performance and asynchronous background task handling.
• ▸ Antigravity CLI retains key features from Gemini CLI, including Agent Skills, Hooks, Subagents, and Extensions (now called Antigravity plugins), and shares architecture with the Antigravity 2.0 desktop app.
• ▸ Starting June 18, 2026, Gemini CLI will stop serving Google AI Pro, Ultra, and free tier requests, which will be supported exclusively via Antigravity CLI; enterprise access via Gemini CLI remains unchanged.
• ▸ Gemini CLI remains open source under Apache 2.0 license with ongoing updates for enterprise customers, while Antigravity CLI is a new proprietary tool built in Go.
• ▸ A migration guide and upcoming video walkthroughs are provided to assist users in transitioning from Gemini CLI to Antigravity CLI.

github.com

06TensorSharp : Open Source Local LLM Inference Engine - Show and Tell - Hugging Face Forums 2026-07-18🔍 ExaAI

• ▸ TensorSharp is an open source local LLM inference engine supporting Unsloth (GGUF) models like Gemma4, DiffusionGemma, and Qwen3.6 with multi-modal capabilities (image, vision, audio) and function tool integration.
• ▸ It runs on Windows, MacOS, and Linux, fully leveraging GPUs from Nvidia, Apple, AMD, Intel, and others via Vulkan, CUDA, and Metal APIs.
• ▸ The engine is implemented from the ground up in C#, with pure C# CPU backend and additional CUDA, MLX, and GGML backends, offering performance on par or better than llama.cpp.
• ▸ TensorSharp’s API is fully compatible with OpenAI and Ollama interfaces, enabling easy integration with existing tools and workflows.
• ▸ It incorporates advanced optimizations such as paged KV cache, continuous batching, SSD-based cache for MoE models, and GGUF quantization for improved inference efficiency.

discuss.huggingface.co

07@nvidia: Agents are becoming specialized, using reinforcement learning post-training to develop domain-specific skills ... 2026-07-17𝕏 X (official)

409 likes, 43 reposts | Agents are becoming specialized, using reinforcement learning post-training to develop domain-specific skills 🛠️ Every post-training rollout is an inference call, so a lower cost per token flows straight into a higher Intelligence per Dollar on every run. That is what the htt

x.com

08@OpenAI: GPT-5.6 Sol sets a new state of the art in cybersecurity on “The Last Ones” cyber range. We’re already seeing... 2026-07-17𝕏 X (official)

3484 likes, 313 reposts | GPT-5.6 Sol sets a new state of the art in cybersecurity on “The Last Ones” cyber range. We’re already seeing that capability translate into defensive outcomes: helping teams find, validate, and fix vulnerabilities in real-world code. Put it to work with Codex Security: htt

x.com