Large Language Models
Jun 8, 2026

The Gist
OpenAI is shifting focus from ChatGPT chatbots to AI agents (programs that can take actions automatically) and coding tools, signaling a major change in how AI companies are developing their products. Meanwhile, developers are building open-source tools to help AI agents remember conversations and work together more effectively. Cloud company Nebius announced a £1.7 billion investment to expand AI computing infrastructure in the UK.
Today's Stories
- 1
OpenAI plans major ChatGPT overhaul as company shifts focus beyond chatbots
Microsoft-backed OpenAI is preparing to move away from its chatbot focus toward AI agents (programs that can automatically perform tasks) and coding tools. This represents a significant strategic shift for the company behind ChatGPT, moving from conversational AI to AI that can take actions on behalf of users.
ChatGPT users may soon see the service evolve from answering questions to actually performing tasks like booking appointments or managing emails automatically.
- 2
Nebius invests £1.7 billion to expand AI computing infrastructure in UK
Cloud computing company Nebius announced it will invest approximately £1.7 billion to build three new data centers in the UK powered by NVIDIA infrastructure. The company is expanding its London operations to serve both commercial customers and AI research organizations.
UK businesses and researchers will have access to more powerful AI computing resources locally, potentially reducing costs and improving performance for AI applications.
- 3
Developers create open-source tools for AI agent memory and collaboration
Multiple developers released free tools to solve AI agent limitations: Midas provides local memory storage that costs nothing to operate, while Lore allows different AI agents to share context and learn from each other's conversations. These tools run entirely on users' computers without sending data to external servers.
AI assistants could become more helpful by remembering past conversations and sharing knowledge between different AI tools, while keeping all data private on your device.
- 4
Engineers build companion robots with emotional expressions and movement
An open-source project called Olaf demonstrates a companion robot with synchronized voice, head movements, ear motion, and a beating heart display. The robot can switch languages mid-conversation and uses improved text-to-speech technology for more natural interactions.
Home companion robots may become more lifelike and emotionally engaging, potentially serving as helpful assistants for elderly care or children's education.
- 5
New toolkit enables motion tracking for robotics without VR headsets
A developer created a headless toolkit for VIVE Trackers that allows robots and AI systems to use motion tracking data without requiring virtual reality equipment. The tool streams position and rotation data over the internet and can integrate directly with AI workflows.
Robotics research and industrial automation could become more accessible and cost-effective by using existing motion tracking technology in new ways.
What to Watch
OpenAI's shift toward AI agents suggests major AI companies are moving beyond chatbots to build AI that can perform real-world tasks automatically. This trend, combined with new open-source memory and collaboration tools, could lead to more capable AI assistants in the coming months.
Sources
- Market Chatter: Microsoft-Backed OpenAI Plans Major ChatGPT Overhaul as It Expands Focus Beyond Chatbots
- Nebius expands in UK with more NVIDIA-powered infrastructure, more customers, and more cloud capabilities for agentic and enterprise AI
- Headless tool kit for Vive Trackers
- I built a MuJoCo skill for AI agents after using AI to create simulation scenes as a beginner
- Closed out the "expression engine" phase on my open-source companion robot — voice, synced head/ear motion, and a beating-heart display. Looking for feedback
- I got tired of flying blind on my AI agents so I built the monitoring layer myself
- How are you actually controlling AI agents in production?
- My 5-layer memory architecture for long-running creative AI agents
- Midas: 100% local agent memory — no LLM at ingest, $0, nothing leaves the box (MCP + Python SDK)
- I made it so any agent can agent can use any context form another agent. Claude learns from codex, visa versa
Share this with a friend
Send today's roundup to anyone who wants to keep up.
Get daily AI news free with AIToday
200+ AI sources, summarized in 1 minute. Email / LINE / Slack.
Sign up free