Welcome to the latest edition of The Neural Link!
In this edition, we explore emerging investment trends in AI research, creative tools, and enterprise platforms. A major funding round for world models signals where the next wave of AI development may be heading, while new AI agents and model releases continue to reshape how organisations build and deploy AI.
Here are some of the latest developments shaping the AI landscape.
Latest AI releases
This month saw fewer flagship model launches from major AI labs. Instead, updates focused on expanding capabilities within existing platforms, including enterprise AI agents, improved reasoning models, and new tools for building interactive AI experiences.
|
Company |
Model / Launch |
What’s new |
|
Microsoft |
Copilot Wave 3 |
Adds enterprise AI agents that can plan and execute multi-step tasks across Microsoft 365. |
|
Anthropic |
Claude Sonnet 4.6 |
Introduces hybrid reasoning and a one million token context window for analysing large codebases and documents. |
|
Runway |
Real-time video characters |
Enables developers to create interactive AI video avatars that respond in real time. |
|
Zoom |
AI agent platform expansion |
Introduces no-code tools for building AI agents to automate workflows across meetings and documents. |
|
|
Gemini 3.1 Pro |
A new reasoning-focused model rolled out across consumer, developer, and enterprise tools. |
Meta AI chief raises $1B to build world models
Meta Chief AI Scientist Yann LeCun’s start-up AMI Labs has raised $1.03 billion to develop what researchers call world models. These systems aim to simulate how the real world behaves, allowing AI to predict how environments and objects may evolve. Unlike traditional large language models that focus on generating text, world models are designed to help AI understand cause and effect in dynamic environments. Researchers believe this capability could enable AI systems to plan actions, anticipate outcomes, and interact more effectively with complex real-world environments.
Why it matters:
The funding highlights growing interest in research that moves beyond language-based AI systems. Many researchers argue that the next generation of AI will need deeper environmental understanding to support applications such as robotics, autonomous systems, and immersive simulations. If world models prove successful, they could represent a major step toward more capable AI systems that can reason about the physical world rather than simply generating responses to text prompts.
Runway has introduced a new API that allows developers to create real-time AI video characters capable of responding dynamically to user prompts. Built on the company’s generative video research and world model technology, the system enables characters to move, speak, and react in real time rather than generating static video clips. Developers can use the platform to build interactive avatars for applications ranging from entertainment and gaming to virtual assistants and training simulations.
Why it matters:
Real-time generative video represents a significant shift in how AI-generated media may be used. Instead of producing pre-rendered clips, these systems enable interactive digital characters that respond instantly to users. This capability could open new opportunities across industries, including media production, customer service, education, and gaming, where dynamic AI-powered characters may become a new form of digital interface.
Anthropic launches Claude Sonnet 4.6 
Anthropic has released Claude Sonnet 4.6, the latest update to its Claude family of AI models. The new version introduces hybrid reasoning capabilities alongside a one million token context window, allowing the model to process significantly larger amounts of information at once. This expanded context allows developers to analyse entire codebases, lengthy research documents, or large datasets within a single prompt, making the model more useful for complex tasks.
Why it matters:
Context length is becoming one of the key competitive areas in frontier AI development. Models that can analyse larger volumes of information in a single prompt are better suited for enterprise use cases such as software development, research analysis, and document processing. As model capabilities expand, organisations are increasingly exploring how these tools can support complex workflows and decision-making.
Zoom expands its enterprise AI agents 
Zoom has expanded its enterprise AI platform with new tools that allow organisations to create AI agents capable of automating workflows across meetings, documents, and collaboration tools. The new platform includes no-code tools that enable businesses to build agents that can summarise meetings, generate documents, and coordinate follow-up tasks automatically. These agents can also connect to external systems and enterprise data sources to support broader workflow automation.
Why it matters:
Enterprise software providers are rapidly embedding AI into their platforms to improve productivity and automate routine work. AI agents capable of performing tasks across multiple systems could significantly reshape workplace collaboration and organisational workflows. As these tools mature, they may shift how organisations manage meetings, documents, and operational processes.
UK authors introduce “Human Authored” label 
The UK’s Society of Authors has launched a “Human Authored” label designed to help readers identify books written by people rather than generated by AI systems. The initiative comes amid growing concern among writers and publishers about how generative AI models are trained. Many authors argue that their work has been used to train AI systems without permission or compensation.
Why it matters:
The label highlights the growing tension between AI developers and the creative industries. As generative AI tools become more widely used, debates around copyright, training data, and attribution are becoming increasingly prominent. These discussions are likely to influence how governments regulate AI development and how companies use of copyrighted material in training datasets.
Other news in AI
-
Google launched Gemini 3.1 Pro, a stronger reasoning model designed to handle more complex tasks across Google’s consumer, developer, and enterprise AI tools.
-
Microsoft expanded its Foundry platform, introducing updates to tools for building and deploying AI agents.
- Salesforce launched Agentforce Contact Center, combining CRM data, AI agents, and telephony to automate customer support workflows.
-
Hugging Face brought the ggml and llama.cpp maintainers in-house, strengthening the ecosystem supporting local and on-device AI model deployment.
-
Meta acquired Moltbook, a Reddit-like platform designed for AI agents to generate content, interact, and share information.
-
Black Forest Labs unveiled Self-Flow, a new technique that improves the efficiency of training multimodal AI models by up to 2.8×.
-
NVIDIA released Nemotron 3 Super, an open 120B parameter model designed to improve performance and efficiency for long-running AI agent workflows.
This month’s stories highlight how AI is evolving across research, enterprise software, and policy. Major research investments are accelerating, new tools are entering everyday workflows, and the debate around how AI is built and governed continues to intensify.
We’ll continue tracking the developments that matter.
If you’re thinking about what this means for your organisation, let’s talk. And if you haven’t already, subscribe to The Neural Link for a monthly view of the trends shaping AI and automation, delivered straight to your inbox.
