Sam Altman on what agents mean for humanity

Plus: Hugging Face's agent-building certification, how to pick the right agent provider for your company, and more

Author

Michael Cunningham
February 13, 2025

Welcome back to Building AI Agents, your biweekly guide to everything new in the AI agent field!

That ex you miss may be taken, but plenty of B2B agent SaaS verticals aren’t. Time to get after it.

In today’s issue…

  • Sam Altman: millions of agents will automate knowledge work

  • Get certified as an agent builder in Hugging Face’s course

  • Microsoft’s forum on agents in healthcare

  • An agent-building agent that requires zero coding

…and more

🗳️ POLL RESULTS

Thank you to everyone who participated in Monday’s poll on how Building AI Agents’ readers are integrating LLMs into their coding workflows!

Of those whose teams write code, fully 78% report using some kind of LLM assistance—but only a small number have begun using fully-automated agentic systems like Replit Agent.

AI coding assistants are rapidly evolving and being adopted, so it’ll be interesting to see how these numbers change over time.

Fully agentic coder (Replit Agent, Devin, etc.)

7%

Assistant in IDE (GitHub Copilot, Cursor, etc.)

36%

Other (pasting into ChatGPT, etc.)

34%

None—100% manual coding

22%

📰 NEWS

Source: TechCrunch

Major HR software provider Workday is rolling out Workday Agent System of Record, a unified control center for all of a company’s AI agents, as well as a new fleet of agents for a wide range of corporate roles.

On Tuesday, February 25, the tech giant will hold a series of sessions on the application of AI—particularly agentic AI—to empowering healthcare and drug discovery. Highlights include the new version of Microsoft’s AutoGen agent framework and a foundation model specifically designed for powering multi-agent systems.

Powered by Anthropic’s latest Claude 3.5 Sonnet model, the cloud data company’s new suite of agents will enable enterprises to intelligently perform complex, multi-step operations on their data.

If you find Building AI Agents valuable, forward this email to a friend or colleague!

🤝 WITH THE AI REPORT

There’s a reason 400,000 professionals read this daily.

Join The AI Report, trusted by 400,000+ professionals at Google, Microsoft, and OpenAI. Get daily insights, tools, and strategies to master practical AI skills that drive results.

🛠️ USEFUL STUFF

Source: HuggingFace

A free course on understanding, using, and building agents, culminating in a final assignment and a certificate of completion that can be useful in obtaining an AI engineering job.

A tutorial by AWS on combining DeepSeek-R1 and CrewAI’s multi-agent framework to create a powerful research agent.

Dive is an open-source application that allows users to easily create agentic applications that expose a Model Context Protocol frontend, Anthropic’s standard for communication between agents that is expected to be the universal interface of a new agentic internet.

Weights and Biases is collaborating with a star panel of judges from Google DeepMind, NVIDIA, AWS, and more to host an agent building hackathon on February 15-16 in San Francisco, with over $50,000 in prizes available.

💡 ANALYSIS

OpenAI CEO Sam Altman | Source: Wikipedia

A blog post by the OpenAI CEO in which he discusses the rise of agentic AI, the exponential fall in the price of intelligence, and the implications for the future of the economy and human society.

A short list of critical items to look for when selecting an agent SaaS vendor for your enterprise.

This article discusses the similarities between the previous generation of business automation tools such as Salesforce and Microsoft Power Platform and AI agents, arguing that the strategies companies used to successfully deploy the former can be applied to the latter.

The author of this piece points to Hugging Face’s recent position paper on the perils of agents operating with near-zero human supervision, arguing that human input is essential in the planning stage of an agent’s operation.

🧪 RESEARCH

Performance on customer support tasks by LLM | Source: LangChain

A study by LangChain on the effect that various factors have on agents’ success rate, finding that greater complexity degrades performance, but some LLMs hold up better than others.

MetaChain is an LLM-powered framework that allows users to describe an agentic system they want built in natural language, which is then autonomously constructed by an agent-building system—an agent that builds agents.

A detailed overview of the various ways in which LLM technology is impacting scientific research, from simple literature summarization all the way to agents which can (theoretically) automate the entire research process.

Thanks for reading! Until next time, keep learning and building!

What did you think of today's issue?

Login or Subscribe to participate in polls.

If you have any specific feedback, just reply to this email—we’d love to hear from you

Follow us on X (Twitter), LinkedIn, and Instagram