Recent Developments in AI - Q1 2025
March'25 - Issue #61
Welcome to the Sophomore tribe. Join others who are receiving high-signal stories from the tech world. In case you missed it, Subscribe to our newsletter.Today’s newsletter is brought to you by TLDR!
TLDR is the free daily newsletter with links and 5 minutes TLDRs of the most interesting stories in startups 🚀, tech 📱, and programming 💻!
Over the past few weeks, AI has seen numerous developments and releases. Let’s explore some key updates and their impact on AI in 2025.
Model Context Protocol (MCP)
The Model Context Protocol (MCP) is rapidly gaining traction as an open standard for AI integrations, serving as a "universal glue" that connects AI models to various tools and data sources. Its community-driven approach has significantly contributed to its widespread adoption. If you want to learn more, Check this X post.
Startups like Cursor, Composio, Polar and many others are actively building their own MCP servers and guides.
One of my friends even built an MCP server using Couchbase and Nebius AI—watch the video here👇
OpenAI Agents SDK
OpenAI recently released a new Agents SDK, this SDK is open-source and integrates with any provider supporting Chat Completions.
Key features include:
Tools (Function Calling): Allows LLMs to invoke external functions or APIs also computer use.
Handoffs: Facilitates passing control between different components or agents.
Guardrails: Ensures safer and more controlled interactions.
Tracing: Provides visibility into the agent's decision-making process.
ManusAI Making Waves
ManusAI is a fully autonomous AI agent launched by Butterfly Effect, a Chinese startup previously known for its Monica AI assistant.
Key Features:
Autonomy: Once given a task, Manus works asynchronously, even if you disconnect, notifying you when it’s done.
Multi-Tasking: It can browse the web, write code, analyze data, and generate deliverables like reports or websites.
Performance: It’s claimed to have set new benchmarks on GAIA, outpacing OpenAI’s o3-powered Deep Research in reasoning and task execution.
Manus is still invite-only, but X chatter suggests over 2 million are on the waitlist.
Early users are leveraging Manus for:
Travel: Crafting detailed itineraries (e.g., Japan trips with flights and guides).
Finance: Analyzing stocks like Tesla with dashboards.
Work: Ranking resumes or coding websites.
More: Summarizing logs, managing social media, or creating lesson plans.
Some more developments and releases:
🔥 Nebius AI launched fine-tuning with awesome fine-tune llms using web console option.
🔥 KitOps got accepted as CNCF Sandbox project - project is trying to establish standardized practices for secure and efficient AI/ML deployments in enterprise environments with it’s Modelkits.
🔥 Gemma 3 - Google’s new LLM, the largest 27B model runs on just one H100 GPU.
🔥 Baidu Inc. has introduced two new AI models - ERNIE X1 matches DeepSeek R1 performance at half the price, and ERNIE 4.5 outperforms GPT-4o at a fraction of cost. Both are free to use.
💬 Which AI development excites you most? Reply and share your thoughts!
Sukriya🙏🏼 See You Again Next Week! Till Then Keep Learning and Hustling