Good morning, AI enthusiast.

Microsoft has made a significant move to bolster enterprise AI capabilities, unveiling key innovations at its Build 2025 conference. The introduction of Copilot Tuning and multi-agent orchestration aims to empower organizations by enabling highly customized AI solutions tailored to their unique data and operational processes.

These developments signify a major step towards more specialized and effective AI assistants within businesses. But will this deeper level of AI customization truly transform workflows, and how will companies navigate the complexities of deploying and managing these bespoke intelligent agents?

In today's AI recap:

• Microsoft's new enterprise AI with Copilot Tuning

• VS Code's shift to an open-source AI editor

• Apple's next-gen Siri matching ChatGPT capabilities

Microsoft Supercharges Enterprise AI

The Recap: Microsoft unveiled major advancements for enterprise AI at Build 2025, introducing Copilot Tuning and multi-agent orchestration to help organizations customize AI with their unique data and workflows. These tools promise more specialized and effective AI assistants.

Unpacked:

• Copilot Tuning empowers businesses to train AI models and create specialized agents using their company data through Copilot Studio, all with a low-code approach.

• New multi-agent orchestration capabilities in Copilot Studio enable different AI agents to collaborate on tasks, effectively dividing complex work based on each agent's expertise.

• To further accelerate custom AI development, Microsoft launched the Microsoft 365 Agents Toolkit, providing pro-code tools for developers to build, test, and deploy their own agents within the Microsoft ecosystem.

Bottom line: These developments offer companies significant new ways to tailor AI, embedding assistants that deeply understand and act on specific business needs. This push for customization signals a future where intelligent agents become even more integral and specialized collaborators in the workplace.

VS Code Goes Open Source for AI

The Recap: Microsoft transforms Visual Studio Code into an open-source AI editor, fully embracing community-driven AI development. This initiative includes releasing parts of the GitHub Copilot Chat extension, further boosting collaborative AI tool creation.

Unpacked:

• Microsoft will release the GitHub Copilot Chat extension's code under the MIT license and integrate key components directly into VS Code's core.

• This move aligns AI development with VS Code’s core principles of being open and collaborative, fostering broader innovation.

• Developers gain the ability to directly influence and enhance AI tools, shaping the future of coding within the VS Code ecosystem.

Bottom line: This decision significantly empowers developers, allowing them to build upon and customize AI coding assistants with greater freedom. Expect a surge in AI tool innovation as the community gains deeper access to foundational technologies.

Apple's Next-Gen Siri Matches ChatGPT

The Recap: Apple's upcoming LLM-powered Siri now performs on par with recent ChatGPT versions, a major leap reported for Apple's AI assistant. This signals a significant turnaround after earlier development hurdles.

Unpacked:

• Apple is developing a new "monolithic model" for Siri, entirely built on an LLM engine, to replace its current hybrid architecture for more conversational and context-aware interactions.

• The revamped Siri may soon gain the ability to access the internet, allowing it to gather and synthesize data from multiple online sources, much like other advanced AI chatbots.

• Beyond its own improvements, Apple is exploring integrations with other AI services, expecting to add Google's Gemini to iOS 19 and potentially Perplexity AI for Siri and Safari search.

Bottom line: This upgraded Siri could dramatically enhance user experience on Apple devices, making daily tasks more efficient and interactions more natural. Apple's renewed AI push aims to position Siri as a true contender in the competitive AI assistant race.

The Shortlist

GitHub launched its Copilot coding agent in public preview, enabling developers to assign issues like bug fixes, feature additions, and code maintenance directly to Copilot within the GitHub platform.

Meta AI released Open Molecules 2025 (OMol25), a large Density Functional Theory (DFT) dataset, and its Universal Model for Atoms (UMA) to accelerate molecular and materials discovery.

A study published in Nature Human Behaviour revealed that AI models like GPT-4 can be more persuasive than humans in debates, especially when given personal information, raising concerns about potential mass manipulation.

Signing off,

David, Lucas, Mitchell — The Recap editorial team