🤖AI Agents Weekly: AI Agents 2024 Recap, smolagents, Agents are not enough, Top LangGraph Agents

AI Agents 2024 Recap, smolagents, Agents are not enough, Top LangGraph Agents

In today’s issue:

• AI Agents 2024 Recap

• Hugging Face launches smolagents

• New research arguing why agents are not enough

• Top 5 LangGraph Agents in Production 2024

• On improving LLM code generation, developments in GUI agents, and more.

---

---

Top Stories

AI Agents 2024 Recap

Based on a comprehensive analysis of AI agent trends in 2024, several key developments emerged in how enterprises and researchers approached autonomous systems. Major companies like Microsoft, Salesforce, and LinkedIn integrated agent technology into their products, primarily using LLMs as orchestration layers to streamline existing API workflows rather than building fully autonomous systems. Meanwhile, foundation models evolved to become more "agent-native" with built-in capabilities for planning, tool use, and coordination, as seen in OpenAI's o1 and Google's Gemini 2.0 Flash.

Interface automation dominated commercial applications, with companies like Kura AI and Microsoft developing agents that could interact with web browsers and GUI elements. The field also saw a shift toward tackling more complex tasks, supported by new frameworks like AutoGen and LangGraph. Benchmark evaluations revealed both progress and limitations - while specialized agents showed promise in narrow domains, general-purpose agents still struggled with complex tasks, achieving only 14.41% success rates compared to 78.24% for humans on end-to-end tasks in WebArena.

Blog