Exploring the Emergence of Agent Engineering: Defining and Redefining the Discipline

Welcome to the AI Engineering Conference, where we're exploring the evolving landscape of agents, capabilities, and reasoning in AI engineering.

1. The speaker acknowledges the challenge of short talks for Summit speakers, including their own.
2. AI engineering is doing well, with an O'Reilly book and a keynote speaker, Chip Huyen.
3. Gartner predicts that engineering has peaked, but the speaker disagrees.
4. The speaker aims to mark the state of the art or industry in each talk.
5. There is resistance from both machine learning (ML) and software engineering viewpoints on what AI engineering entails.
6. The speaker believes AI engineering will emerge as its own discipline, currently being a mix of 90% software engineering and 10% AI.
7. Talks at the conference aim to explore anthropological aspects of AI engineering.
8. Summit has pivoted to an agent engineering conference, focusing on agents instead of other technologies like RAG, open models, GPUs, etc.
9. Last year's top talks focused on agentic things, but this year the goal is to feature production-ready agents.
10. The speaker mentions a new rule against vendor pitches, making it harder to find speakers but ensuring more genuine content.
11. Agent + RAG, Cent, or search works is seen as a simple formula for making money in 2025.
12. There's skepticism about 2025 being the "Year of Agents," although it could be true if enough people believe it.
13. The speaker plans to define the term "agent" during the conference, mentioning various POVs and a crowdsourced list of 300 definitions from Simon Willison.
14. Capabilities have grown and hit human baselines, making agents more viable now than before.
15. Model diversity has increased, with open eye's market share dropping from 95% to 50%.
16. The cost of intelligence has significantly decreased, enabling more widespread use of agents.
17. Faster inference, multi-agent work, and charging for outcomes are other factors contributing to the rise of agents.
18. Attendees should consider building effective agents for product-market fit.
19. Anti-use cases include flight booking and Instacart ordering agents, which people may prefer to handle themselves.
20. Open AI reported 400 million users, growing 33% in three months, while ChatGPT reached that number in two and a half years.
21. The speaker suggests that the growth of AI products is tied to reasoning capabilities and the number of agents provided to users.
22. AI jobs are evolving towards building agents, much like how MLOps professionals build models.
23. The conference aims to welcome attendees and introduce them to the world of agent engineering.
24. The speaker encourages enjoying the show and the opportunity to learn more about agent engineering.

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!