From PHP to team lead of agents: rethinking judgment, review, and data with Google's Andi Gutmans (Part 1)
Andi Gutmans, head of Agentic Data Cloud at Google and co-creator of PHP, joins Leaders of Code to talk about why agentic development feels less like a break from the past and more like the next chapter of the same story. This is part one of a two-part conversation.
Welcome to Leaders of Code, a segment on the Stack Overflow Podcast where senior engineering leaders come together to discuss their work, team-building strategies, and the biggest challenges they are facing. I'm Eira May, the B2B editor at Stack Overflow, and I'm joined by my colleague Peter O' Connor, the director of platform engineering here at Stack.
Today, we have a guest who is truly exciting for me to talk with: Andi Gutmans, the head of Agentic Data Cloud at Google and one of the creators behind PHP 3. Peter, welcome back. It's going really well. Peter, thank you for joining the show. Andi, welcome to the podcast. Thank you so much for joining us. I wanted to start off by asking, as someone who helped build tools that made web development accessible to a generation of developers, how does the shift to agent-scale feel from your perspective?
Does it feel like a huge change, or more of a natural progression? I think it feels more like a natural progression. When we worked on PHP, we democratized web development, making it accessible to anyone with some technical acumen. Now, with agents, we're enabling even more practitioners to achieve remarkable outcomes. A good example is my sister, who is a lawyer but has built fantastic websites using Lovable.
It's like PHP, but even better on almost every dimension. Another key aspect is the trust and security aspect, which can now be driven by agents, unlike in the past when non-computer science graduates often created insecure websites. Can you share some principles and guardrails you consider when guiding the use of AI and these tools for developers?
Are there specific patterns or best practices you recommend? Sure, I think the same principles apply to AI-assisted development as they do to traditional development. Delivering trust, ensuring security, and maintaining ease of use are crucial. The main difference now is that agents can perform tasks autonomously, so we need to adapt our guardrails accordingly.
However, the fundamental considerations remain largely the same. We must ensure the right data is available to activate these agents, make informed decisions about human oversight versus agent autonomy, and balance human and agent judgment effectively. How do you think about the balance of human versus agent judgment in this new paradigm?
I believe the balance of human and agent judgment is crucial. While agents can handle many tasks autonomously, there are still instances where human oversight is necessary. For example, determining when an agent should be in the loop versus when a human should be involved, or when an agent should operate independently. These decisions depend on the specific context and goals of the project.
It's essential to strike the right balance between leveraging the power of agents and maintaining human oversight where it matters most.
Written by urgent.news from Stack Overflow Blog's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
