Engineers need to be the CEOs of agents. Are they ready?
Leading up to AI Council 2026, I’m digging into some of the themes we’re excited to cover. This week: the rise of AI coding — and what it actually means to build reliable software with agents.
For those of you who don’t know me, my name is Pete Soderling. I started AI Council (formerly Data Council) with one purpose: to bring together the sharpest engineers & tool builders in the world to talk about what they’re actually experiencing in the field. No BS. No hype.
Each piece tees up a topic we’re planning to go deep on at the conference this year. If something resonates — or you disagree — comment below. As they say in open source: the best code comes from the PR comments.
Speed is the easy part. Trust is the hard part.
Alfred Lin, Partner at Sequoia, recently put a number on something I’ve been watching: the top 5–10% of builders are now 3–5x more productive than they were a year ago. The median builder? Up maybe 10–20% (See his full X post). Same tools, very different outcomes.
What separates them isn’t tool access. It’s how they think about the job.
Lin calls it the shift from developer to fleet commander — the engineers winning right now aren’t writing more code, they’re directing agents the way Ender Wiggin directed squadrons: clear intent, parallel execution, and judgment reserved for the decisions only a human can make. Scott Breitenother, co-founder & CEO of Kilo, shares a similar outlook.
He says that engineers need to act with the same ownership as a CEO of the product they are building.
At Kilo, every developer owns a feature in an end-to-end process, from ideation to usage analytics, and manages a pod of agents to do it. No handoffs. No committee reviews.
I sat down with Scott to dig in further — what does this shift actually demand from engineers technically, how does he think about collaboration culture, and what does he tell engineers who are still figuring it out.
My Q&A with Scott Breitenother, Co-founder & CEO of Kilo
Note: These answers have been edited for clarity.
Pete: How do engineers need to change their mindset for AI coding?
Scott: The shift is from first-chair violinist to conductor of an orchestra. It’s a very big shift. You go from “hands on keyboard” to deep-brain thinking — breaking down projects, setting goals, bringing in the human elements of taste, and letting multiple parallel agents do the execution work. You move from doer to visionary and the output reflects it.
At Kilo, developers have gone from individual contributors to managers of a team of agents. And we’re seeing output that’s roughly 5x what it was before.
Pete: So if developers are becoming orchestrators, what skills do they actually need?
Scott: Everyone will need to be full stack; the separation between frontend and backend developers is going away, but the tools will help you get there. You won’t need to memorize everything — it’s like trigonometry: I don’t remember every formula, but I remember enough to know what to Google when I need it. The same applies to parts of the stack you’ve never touched. You’ll know enough to find what you need, and the agent fills in the rest — like an assisted full stack.
But full stack is just the starting point. At Kilo, our developers aren’t just writing code — they’re publishing launch posts, tracking weekly active users on the features they own, and setting product direction. And so a Kilo developer is not just a product engineer. They’re a mini CEO.
Pete: What does the whole orchestration and coordination layer look like from a tooling perspective?
Scott: In a few years, we’ll look back nostalgically at when people were picking which model they were using — “I was using Opus for this and Codex for that.”
In the future, you’ll have a single pane of glass where you write a request and it routes to the right model for the job — architecting in Opus, checking it in Codex, coding in MiniMax, and writing the copy using Gemini. Your platform kicks tasks off into the cloud, runs them, and you pull them down to interact.
I’ll even double down: in five years from now, you'll have always-on agents chewing through your to-do list — reading customer feedback, scanning logs when there's an issue, automatically triaging and fixing things. Right now, agents are ad hoc, spun up for specific tasks. That's the part that's about to change.
Pete: You’ve endorsed PostHog’s recent article: “Stop collaborating”. Is collaboration going away?
Scott: The PostHog article and our comments are intentionally thought-provoking. But I do think that in a world where AI is automating a lot of tasks — from writing code to a strategy — then you have to think, what’s the next bottleneck?
And the next bottleneck is human-to-human communication. At Kilo, we’ve been obsessed with removing these “velocity killers,” and we do that by removing teams.
Every feature has one developer who owns it end-to-end and manages a team of agents to build it. That developer can then focus a hundred percent on shipping — no switching costs, no waiting on a human review that takes a week so everyone can feel heard.
I think too many human-to-human interactions are just ways to make people comfortable. And I think we all need to get over it.
This doesn’t mean a Kilo developer never talks to a colleague — that happens all the time. But it’s not the first or last thing you do. You only talk to someone when it adds value. The more forward-thinking, small companies are operating this way: one person, full ownership, no dependencies. That’s the future.
Pete: What’s the next stop for vibe coding? What advice do you have for engineers?
Scott: AI is like a mech suit. It magnifies the things humans do inherently well — vision, taste, architecture — that crystal clear sense of what you're actually trying to build. And the suits are just going to keep getting better.
But people need to rise to the occasion. You can't put your head in the sand on this one, because there's a line of demarcation forming. If you're above it, you'll be managing agents — doing big brain work, directing things that are running all the time. If you're below it, you're going to be managed by agents, doing something that, for financial reasons, isn't worth automating quite yet.
When I talk to people starting their careers, I tell them: don't be afraid of this — embrace it. Because you have to be above that line.
Pete: Thanks, Scott! Looking forward to continuing the conversation at AI Council.
Scott: Likewise! Thanks, Pete
Watch the full conversation here
There’s a lot more to dig into on AI coding. A few questions still on my mind for AI Council:
What does the quality gate look like when agents are shipping at this speed?
How do you maintain the deep focus and flow state that great engineering requires, when your job is now to direct and verify rather than build?
What new tools do we need to handle the onslaught of vibe coding — and what’s most promising in that category?
What does the most effective human-AI collaboration actually look like in practice?
Also worth reading on this topic: Nick Schrock on refactoring his own role as a founder and CTO.
I’m excited to hear from Eno Reyes, CEO of Factory, Chris Clark, Co-founder & COO of Open Router, Mark McQuade, Founder and CEO of Acree.ai, and Dima Dzhulgakov, Co-founder & CTO of Fireworks AI and others at AI Council, May 12–14, 2026. Grab your ticket here.
If you have another meaty question about AI coding — drop it in the comments.