Software companies are constantly trying to add more and more AI features to their platforms, and AI companies are constantly releasing new models and features. It can be hard to keep up with it all, so we’ve written this roundup to share several notable updates around AI that software developers should know about.
Google announces new and updated tools for building AI agents
The company announced the Agent Development Kit (ADK), an open-source framework that covers the end-to-end process of building and deploying agents and multi-agent systems.
Google also announced the Agent2Agent (A2A) protocol, an open protocol that allows AI agents to communicate with each other, exchange information in a secure way, and coordinate actions on top of enterprise applications. It was developed with support and contributions from over 50 other companies, including Atlassian, Box, MongoDB, Salesforce, ServiceNow.
In Agentspace, organizations will now be able to give their employees access to the platform’s unified enterprise search, analysis, and synthesis capabilities from within Chrome’s search box. Other new capabilities include a no-code Agent Designer for building custom agents, and access to two new agents built by Google: Deep Research and Idea Generation.
GitHub Copilot adds agent mode, MCP support in latest release
A new version of GitHub Copilot has begun rolling out to users, adding a number of new capabilities, such as the addition of agent mode to VS Code.
With agent mode, Copilot can iterate across an entire project, suggesting terminal commands, analyzing run-time errors, and more. “With simple prompts, agent mode takes Copilot beyond answering a question, instead completing all necessary subtasks across automatically identified or generated files to ensure your primary goal is achieved,” the company’s CEO Thomas Dohmke wrote in a blog post.
As an example, GitHub published a demo video in which a developer needs to update a website for runners that allows them to sort races by name, distance, and time. Upon receiving the request, Copilot analyzes the site to determine what needs to be changed, then begins by updating the backend and the UI, then generating unit tests for both, before handing it back over to the user to run those tests.
Microsoft updates Copilot to be a more personalized AI companion
According to Microsoft, the latest updates are designed to bring it from just an “AI companion” to “your AI companion.” It will now be able to remember what you talk about to learn your likes and dislikes, as well as details about your life, such as your dog’s name or work projects you’re assigned to.
Other new capabilities added to Copilot include:
- Deep Research, for conducting multi-step research tasks
- Actions, for completing tasks on your behalf, such as booking event tickets or making dinner reservations
- Pages, which combines notes, content, and other research into a single canvas
- Vision, which lets Copilot see what you see and respond in real time
“Copilot will understand you in the context of your life, and show up, on your terms, in the right way at the right time. This is far richer, more dynamic, supportive and emergent than any software we’ve seen before. It’s a new kind of relationship with technology, a new era,” Microsoft wrote in a blog post.
OpenAI expands ChatGPT’s memory capabilities
In an effort to get ChatGPT to know more about you, OpenAI has announced that it can now reference all past chats that have occurred. This update “points at something we are excited about: ai systems that get to know you over your life, and become extremely useful and personalized,” Sam Altman said on X.
This feature began rolling out yesterday for Pro users, and Plus users will soon get it as well. Users will be able to opt out of this, and can always use the temporary chat feature to have a conversation that won’t be saved to ChatGPT’s memory.
Amazon Nova Sonic is announced
This is the latest model to be added to the Nova family, and it is a speech-to-speech model for conversational AI.
According to Amazon, typically, speech understanding and generation are covered by two separate models, but Nova Sonic unifies both into one model. “The result is an adaptive speech response that dynamically adjusts its delivery based on prosody, such as pace and timbre, of input speech,” Amazon wrote in a blog post.
Gemini 2.5 Pro now in public preview
Due to the adoption of Gemini 2.5 Pro and feedback from users, Google has decided to move it to public preview in the Gemini API in Google AI Studio. It will also be accessible through Vertex AI shortly, Google said.
The pricing for Gemini 2.5 Pro is $1.25/1 million tokens for text, image, audio, and video inputs in a 200k context window or less, and $2.50/1 million tokens text greater than a 200k context window. The output pricing is $10/1 million tokens for context windows under 200k and $15.00/1 million tokens for context windows greater than that. The experimental version will remain free, but will have lower rate limits.
Meta releases first Llama 4 models
The announcement covered three models that are part of the Llama 4 family:
- Llama 4 Behemoth (in preview), a teacher model for distillation that offers 288B active parameter, 16 experts, and 2T total parameters
- Llama 4 Maverick, a native multimodal model with 1M context length, offering 17B active parameters, 128 experts, and 400B total parameters
- Llama 4 Scout, optimized for inference with a 10M context length, that has 17B active parameters, 16 experts, and 109B total parameters
Cloudflare announces remote MCP server to reduce barriers to creating AI agents
Cloudflare is making it easier for developers to create AI agents with several new updates that were announced earlier this week.
First, the company announced a remote Model Context Protocol (MCP) server. MCP is an open standard that allows AI agents to interact with external services, which is important in enabling them to complete tasks on a user’s behalf.
According to Cloudflare, MCP has thus far been limited to running locally, which has prevented its mainstream adoption. Now, users will be able to build and deploy remote MCP servers through Cloudflare, which will allow agents to connect and interact with services without needing to rely on a locally hosted server.
Progress packs Sitefinity 15.3 with AI updates
The latest version of the content management system offers AI capabilities for users like summarizing content, improving writing, personalizing messages, and generating tag suggestions. Users will also be able to search for images using natural language descriptions.
Other AI capabilities include AI translation, content optimization, and the ability to export marketing data in the Parquet format.
“AI is revolutionizing what is possible in creating digital content and experiences, and any organizations not using these capabilities will be at a significant competitive disadvantage,” said Loren Jarrett, EVP & GM of digital experience of Progress. “Sitefinity is at the forefront of innovation, empowering marketers, developers and content teams with the most advanced tools so they can deliver digital experiences that will enable them to compete and win in an AI-driven world.”
Read last week’s announcements here: April 4, 2025: AI updates from the past week — Claude for Education, new website for exploring Amazon Nova models, and Solo.io’s MCP Gateway
