Navigating AI Engineering: Emerging as a Distinct Discipline in 2025

As AI engineering continues to evolve, I'll share my insights on the state of the industry, from the rise of agent-based frameworks to the challenges and opportunities emerging in 2025.

  • 1. The speaker is discussing the current state of AI engineering.
  • 2. They mention that there are different perspectives on what AI engineering entails, with some viewing it as primarily machine learning plus a few prompts, while others see it as mostly software engi
  • 3. The speaker believes that AI engineering will emerge as its own discipline over time, and that this year will see the start of its spread.
  • 4. They mention an O'Reilly book on AI engineering and a friend, Chip, who is giving a keynote at the workshops session.
  • 5. According to Gartner, AI engineering has hit its peak and is expected to decline.
  • 6. The speaker aims to mark the state of the art or industry in their talks at each conference.
  • 7. They previously discussed the rise of the AI engineer, the three types of AI engineers, and the maturing and spreading discipline at past events.
  • 8. The speaker mentions resistance from both sides of the AI engineering spectrum, with some viewing it as mostly machine learning and others seeing it as primarily software engineering.
  • 9. They believe that AI engineering will grow to be more than 10% AI and 90% software engineering in the future.
  • 10. The speaker discusses how people describe themselves, form groups, and create industries, using ML and AI as examples.
  • 11. They mention that the AI Engineer Summit has pivoted to become the Agent Engineering Conference.
  • 12. This decision was not made lightly, but rather because it allows for a focus on agent engineering disciplines.
  • 13. Last year's top talks on YouTube featured agentic things, and this year's speakers have been instructed not to give vendor pitches.
  • 14. The speaker mentions that agents combined with other technologies can make money in 2025.
  • 15. They are skeptical about the claim that 2025 is the year of Agents, but believe that agent technology has potential.
  • 16. The speaker invokes Simon Willison, who crowdsourced 300 definitions of an agent.
  • 17. Different perspectives on what constitutes an agent include having goals, using tools, controlling flow, running long processes, delegating authority, and completing small multi-step tasks.
  • 18. OpenAI released a new definition of agents on the live stream, which is worth paying attention to.
  • 19. The speaker believes that agents are working now because capabilities have been growing and hitting human baselines.
  • 20. Model diversity has also increased, with two Frontier Model Labs emerging as potential challenges to OpenAI.
  • 21. The cost of intelligence has decreased 1,000 times in the last 18 months.
  • 22. Faster inference and multi-agent work are being developed due to better hardware.
  • 23. Agent use cases include building effective agents for coding and support, with potential for growth in areas such as anti-use cases like flight booking agents.
  • 24. OpenAI has reported 400 million users, a 33% growth from three months ago, while ChatGPT has grown to 400 million users in two and a half years.

Source: AI Engineer via YouTube

❓ What do you think? What are your thoughts on the ideas shared in this video? Feel free to share your thoughts in the comments!