Daily AI News
YixScout tracks AI tools, AI apps, AI news, funding, launches, and industry signals so global builders can follow the market quickly.
OpenAI rolls out a more scalable memory system for ChatGPT
OpenAI's June 4 release says ChatGPT is getting a stronger memory synthesis system designed to improve freshness, continuity, and relevance across long-term user histories.
Cursor adds Canvas Design Mode and a context usage report
Cursor's June 4 changelog says agents can now be guided by direct annotations inside canvases, while a new context explorer shows how tokens are spent across prompts, tools, rules, and skills.
Perplexity brings Personal Computer to Windows for local agent work
Perplexity's June 3 post says Personal Computer is coming to Windows so its multi-model agent can work across local files, Microsoft apps, and the web from one surface.
Microsoft Build pushes Work IQ APIs and open controls for production agents
Microsoft's June 3 Build post says Work IQ APIs will go generally available on June 16, alongside ASSERT and the Agent Control Specification for governing agent behavior across frameworks.
OpenAI adds plugins, annotations, and shareable Sites to Codex
OpenAI says Codex now supports role-specific plugins, in-place annotations, and preview Sites for interactive apps and websites that teams can share by URL inside a workspace.
Cognition launches Devin Desktop as the next generation of Windsurf
Cognition says Devin Desktop turns the Agent Command Center into the default IDE surface, adds shared Spaces for agent context, and supports ACP-compatible agents alongside Devin.
GitHub refreshes Copilot CLI with reviews, scheduling, and voice input
GitHub's June 2 changelog says Copilot CLI now adds a new terminal interface, a rubber-duck review agent, scheduled prompts, and voice input for agent sessions.
NVIDIA pushes secure autonomous AI engineers with the NemoClaw blueprint
NVIDIA says more than a dozen industrial software partners are using its open NemoClaw blueprint to automate simulation setup, debugging, post-processing, and reporting for engineering workflows.
OpenAI brings frontier models and Codex to AWS for production workloads
OpenAI says its frontier models and Codex are now generally available on AWS, giving enterprises a way to adopt OpenAI capabilities through existing security, governance, billing, and deployment workflows.
Runway joins NVIDIA to launch the Cosmos Coalition for open world models
Runway says it is a founding member of the Cosmos Coalition, a new effort with NVIDIA and other labs to build and open-source frontier world models for physical AI.
Anthropic launches Claude Opus 4.8 and expands Claude Code workflows
Anthropic says Claude Opus 4.8 improves coding and agent benchmarks, while Claude Code gains dynamic workflows so it can take on much larger software tasks.
ElevenLabs launches Dubbing v2 to preserve voice performance across 90+ languages
ElevenLabs says Dubbing v2 conditions directly on the original performance so tone, pacing, and emotion can transfer across more than 90 languages for creators and studios.
Mistral relaunches Vibe as one agent for long-running work and coding
Mistral says Vibe now combines work and code into one agent across web, mobile, editor, and terminal surfaces, with Work Mode for long tasks and Code Mode for remote coding jobs.
Mistral opens Search Toolkit for production retrieval pipelines
Mistral says Search Toolkit is now in public preview as an open framework that unifies ingestion, retrieval, and evaluation for enterprise search and RAG systems.
xAI launches Grok Build as a terminal coding agent in early beta
xAI says Grok Build is now available in early beta for SuperGrok and X Premium Plus subscribers, bringing plan mode, review flows, and parallel subagents into the terminal.
Replit teaches Agent to suggest Power Mode and adds Browserbase connectors
Replit's May 22 changelog says Agent can now recommend switching to Power Mode when users get stuck, while Browserbase and Quiver AI join the connector catalog.
Google launches Gemini 3.5 Flash for faster agentic coding and task execution
Google says Gemini 3.5 Flash is generally available across Search, the Gemini app, AI Studio, Android Studio, and Antigravity, with stronger coding and long-horizon agent performance.
Google unveils Antigravity 2.0 as its new agent-first development platform
Google's I/O 2026 developer update says Antigravity 2.0 adds a desktop app, parallel subagents, scheduled automations, Managed Agents, and tighter links with AI Studio, Android, and Firebase.
Google DeepMind details Decoupled DiLoCo for more resilient model training
Google DeepMind's April 23 research post says Decoupled DiLoCo can train large models across distant data centers with lower bandwidth demands and stronger resilience to hardware disruption.