AI agents are currently overhyped, especially in consumer applications.

Booking a vacation through an AI agent is as challenging as doing it manually.

An AI agent is defined as a system that allows a language model to autonomously decide how many times to run processes until it finds a resolution, rather than following a predefined path.

The distinction between workflows and agents is crucial; workflows are pre-orchestrated by code, while agents operate more autonomously and can adapt their actions based on the situation.

As AI models improve, agents are becoming more prevalent and capable, prompting the need for formal definitions to clarify their use.

It's important for developers to view the world through the lens of the AI model, considering what instructions would be most applicable for it to understand and operate effectively.

When implementing AI agents, developers should ensure that the tools provided to the model are well-documented and user-friendly, as this significantly impacts the model's performance.

There's a lot of excitement around AI agents, but many people are unsure of their practical applications, often trying to apply them to problems that simpler solutions could address.

One overhyped aspect of AI agents is their perceived capabilities, while the underhyped aspects include the importance of understanding when and how to use them effectively in production, as well as how they can save people time by automating tasks that take only a minute, significantly changing dynamics and allowing you to perform those tasks many more times than before.

It's difficult to determine where agents are truly needed. There's a sweet spot for using agents in tasks that are valuable and complex, but where the cost of error or monitoring is relatively low.

Coding agents are exciting because they are verifiable. You can write tests for code, and the tests will indicate whether the code is functioning correctly, which is not possible in many other fields.

The next limiting factor for coding agents will be verification. While models are improving at writing code, real-world applications often lack perfect unit tests, making verification challenging.

In 2025, I envision a lot of business adoption of agents, automating repetitive tasks and scaling up operations that were previously too expensive. This could lead to significant efficiency improvements.

Agents for consumers are currently overhyped. Specifying preferences for tasks can be as complex as doing the task yourself, and verifying the agent's actions can be very expensive.

To effectively build AI agents, ensure you have a way to measure results. Many developers build in a vacuum without feedback, which can lead to ineffective solutions.

Starting simple and having measurable results is crucial as you build complexity into your AI agents. This approach allows you to adapt and improve as you learn what works.

If your product's success depends solely on the models getting smarter, you're likely building the wrong thing. Instead, create something that improves as the models advance.