Tag
Encrypted Spaces is a research preview of an architecture for collaborative applications that ensures data is encrypted and operations are cryptographically verifiable, allowing servers to store data without being able to inspect plaintext. It includes a prototype sync engine and is developed with Microsoft Research and Harvard's Berkman Klein Center.
FeifanZ questions whether dynamic workflows in Claude Code are essentially vibe-coding with worse observability, referencing Claude's announcement of the feature as a research preview for complex tasks.
Claude Code's research preview now includes dynamic workflows, allowing Claude to write orchestration scripts and spin up coordinated subagents for complex tasks.
Spotify debuts a new desktop app called Studio by Spotify Labs that uses AI to generate personalized podcasts from users' email, calendar, and documents, directly competing with Google's NotebookLM.
Anthropic launches self-hosted sandboxes (public beta) and MCP tunnels (research preview) in Claude Managed Agents, enabling agents to run within the user's own perimeter with default security controls.
Fast mode for Claude Opus 4.7 is now available in research preview, offering faster responses. It is currently opt-in but will become the default.
Thinking Machines AI announces a research preview of interaction models, a new architecture designed for native, real-time human-AI collaboration across audio, video, and text. By replacing turn-based interfaces with a multi-stream, micro-turn design, the model aims to keep humans actively in the loop while delivering state-of-the-art intelligence and responsiveness.
Anthropic has released a new 'agent view' feature in Claude Code, providing a unified list of all sessions, currently available as a research preview.
An OpenAI research preview explores learning from how people interact with their computers beyond chat, accompanied by a new arxiv paper on the topic.
OpenAI introduces Codex Security, an agentic application security tool now in research preview that identifies complex vulnerabilities with high confidence and actionable fixes while significantly reducing false positives and noise compared to traditional security tools.
OpenAI is releasing GPT-5.3-Codex-Spark, a smaller, ultra-low-latency coding model optimized for real-time collaboration, delivering over 1000 tokens per second on Cerebras hardware. It is available as a research preview to ChatGPT Pro users and marks the first milestone in OpenAI's partnership with Cerebras.
OpenAI releases a research preview of GPT-4.5, its largest and most knowledgeable model to date, built on GPT-4o with scaled pre-training, improved emotional intelligence, and fewer hallucinations. The system card details training methods, safety evaluations, and capability assessments conducted prior to deployment.
OpenAI introduces GPT-4.5, their largest and best chat model yet, available as a research preview to Pro users and developers. The model advances unsupervised learning through scaling compute and data, showing improved factuality, reduced hallucinations, and better understanding of human intent compared to GPT-4o.
OpenAI releases Operator, an AI agent that can autonomously use a web browser to perform tasks like form filling, online shopping, and web interactions. Available as a research preview to Pro users in the U.S., with plans to expand and integrate into ChatGPT, powered by the new Computer-Using Agent (CUA) model.
OpenAI announces an expansion of DALL·E 2 research preview access, sharing safety metrics and learnings from 3 million images created by early users. The company plans to onboard up to 1,000 new users weekly while continuing to refine content policy enforcement and address training data biases.