Coding as a problem is largely "solved." That's not my claim — it's from Boris Cherny, Head of Claude Code at Anthropic, in a fantastic interview on Lenny's podcast. Some standout takeaways: Even as a senior leader, Boris still codes daily — but now 100% with AI. He's not writing code by hand anymore (though a human is still in the loop for quality review). He believes that within 18-24 months, it won't even be relevant to understand what's happening under the hood. His analogy: the printing press. Before it, only a select few could read and write — employed by nobility. The press democratized that skill. The transition wasn't painless, but it was unstoppable. My take: We've been talking about "Tech eating Ops" for a while, and the general thrust is right. But I think the better frame is this: Ops is evolving into Technologists, empowered by AI coding tools. Operations experts have deep domain expertise and have spent their careers synthesizing across complex, interconnected systems. That makes them uniquely positioned to build and manage an AI-powered workforce. This won't happen automatically, though. It takes individual initiative and strong leadership to guide people into this moment. And one more thing: I'm betting on generalist models over specialist ones. Specialist models carry too much of the baggage we saw with RPA done wrong — overly complex architecture, cascading failures that are hard to trace, and a total cost of ownership that quietly kills the ROI. The printing press didn't create more scribes. It created more thinkers. That's where we're headed. What do you think — are you betting on generalists or specialists? 🔗 https://lnkd.in/gVYWi6pk
Rukesh Reddy
2mo
I vote for team generalist over team specialist! With big doses of customization per use case
No code platforms fell short of enabling it, I‘d say the cost of deployment and the constraints the platforms imposed weren’t good trade offs at the time. But now things are different. Interfaces are becoming more irrelevant, processes are increasingly automated and AI companies forward deploy eng to work with domain owners on configuration and evaluation. It‘s matter of time that ops becomes tech, and tech more like devops.
Mark Hedley
2mo
I’ve just come out of that conversation. My view, if you have the expertise to understand the output, it’s a super power. If you don’t, start writing your resume.
The printing press analogy is powerful, Leon, but the part that resonates most is your reframe of Ops into Technologists. Domain experts who understand the system deeply are better positioned to direct AI than engineers parachuted into unfamiliar domains. The RPA comparison to specialist models is a warning more teams should hear.
Yes, I think generalist models but deployed with more specific instructions and tools - current agentic coding workflows seem to work best with pipelines of agents that do specific things, but all using general coding models from OpenAI, Google or Anthropic. I'm also very curious to see whether steering teams of AI agents will look more like Operations or like Engineering - I genuinely don't know, it could absolutely go either way. I think the Operations discipline and experience at scaled leadership might be incredibly powerful, and much of that is _already_ in the face of lots of automation. It's going to be really interesting to find out. Thanks for sharing the podcast link.