From: aidotengineer
Current Status and Evolution
AI engineering is an evolving discipline, with varying perspectives on its core identity [01:57:00]. While some machine learning engineers view it as an extension of ML with a few prompts, software engineers often see it as mostly software engineering with calls to LLM APIs [01:39:00]. Currently, it’s described as approximately 90% software engineering and 10% AI, a proportion expected to grow over time [02:00:00]. The discipline is maturing and spreading across different fields [01:19:00].
Previous “state of the industry” talks have covered:
- The rise of the AI engineer [01:11:00]
- The three types of AI engineers [01:13:00]
- The maturation and spread of AI engineering [01:19:00]
The sentiment around AI engineering varies; an O’Reilly book suggests a positive outlook [00:37:00], while Gartner believes it has peaked and is now in decline [00:48:00].
Evolution of AI Engineering and Tools
The Pivot to Agent Engineering
The AI Engineer Summit has strategically pivoted to become an agent engineering conference [02:44:00]. This focus means saying “no” to other topics like RAG, open models, and GPUs, to concentrate solely on agents [02:53:00]. This decision was driven by audience interest, as top-performing talks last year favored agent-related content [03:18:00]. A new rule for speakers requires a focus on production deployment rather than just vendor pitches [03:36:00]. The belief is that “everything plus agent works,” such as agent plus RAG, agent plus search, or agent plus sentiment analysis [04:00:00].
Defining an Agent
Defining “agent” is a monumental task, with various interpretations [05:33:00].
- Machine Learning (ML) perspective: Focuses on reinforcement learning environments, actions, and achieving goals [05:41:00].
- Software Engineering (SE) perspective: Often simplifies it to a for loop [05:49:00].
Simon Willison has crowdsourced over 300 definitions [06:06:00], which generally revolve around:
- Goals [06:17:00]
- Tools [06:17:00]
- Control flow [06:20:00]
- Long-running processes [06:20:00]
- Delegated authority [06:22:00]
- Small, multi-step task completion [06:23:00]
OpenAI also recently released a new definition of agents that the community should pay attention to [06:52:00].
Why Agents are Working Now
The current success of agents, compared to previous years, is attributed to several factors [07:12:00]:
- Improved Capabilities: AI capabilities, including reasoning and tool use, have significantly grown and are hitting human baselines around 2025 [07:19:00].
- Model Diversity: OpenAI’s market share has decreased from 95% to 50% in two years, leading to a much more diverse landscape of models [07:51:00]. New Frontier Model Labs are emerging as potential challengers [08:02:00].
- Reduced Cost of Intelligence: The cost of GPT-4 level intelligence has decreased by 1,000 times in the last 18 months [08:14:00].
- RL Fine-tuning Options: The availability of reinforcement learning fine-tuning options [08:28:00].
- Focus on Outcomes: Shifting focus from charging for costs to charging for outcomes [08:44:00].
- Multi-agents and Faster Inference: Advancements in multi-agent systems and faster inference due to better hardware [08:49:00].
Despite skepticism from some in the community who are “tired of hearing about agents” [05:00:00], industry leaders like Satya Nadella, Roman, Greg Brockman, and Sam Altman predict that 2025 will be the year of agents [04:32:00]. David Luan, former VP at OpenAI, initially advised against branding with agents but now suggests putting it back on [05:15:00].
State of the AI Frontier in 2025
Agent Use Cases
Agents are seeing product-market fit (PMF) in areas such as coding agents and support agents [09:12:00]. Deep research agents also have PMF [09:15:00].
However, certain “anti-use cases” should be avoided:
- Demoing agents that book flights [09:25:00]
- Booking Instacart orders [09:34:00]
Impact on AI Product Growth
The growth of AI products is tightly linked to reasoning capabilities and the number of agents that can be shipped to users [10:41:00]. For example, OpenAI reported 400 million users, a 33% growth in three months [09:47:00]. ChatGPT usage doubled with the introduction of “01 models” (agentic models) [10:24:00]. It is projected to reach one billion users by the end of 2024, quadrupling its user base from September 2023 [10:28:00]. This signifies that one-eighth of the world’s population will be using ChatGPT by year-end [10:50:00].
The Future Role of AI Engineers
The job of an AI engineer is evolving towards building agents, similar to how ML engineers build models and software engineers build software [11:00:00].