When agentic coding startup Kilo launched Kilo Pass, its subscription plan, in January, the idea was only three days old.
While most AI coding tools have customers pay per user or buy set amounts of monthly usage via a subscription, Kilo is pay-as-you-go. Still, some loyal users were saying they wanted some kind of subscription plan. So CEO Scott Breitenother asked Igor Šćekić, one of the company’s 14 engineers, to go for it.
At Kilo, that meant conceptualizing the product on his own, building it end-to-end, and speedily launching it into production, skipping common steps like wiring framing, drafting a PRD (product requirements document), and even seeking approval from the team or leadership. Šćekić designed the gamification approach that ramps up monthly credit bonuses over time, built it, and worked with a partner in marketing to create a landing page. After going live with Kilo Pass in just three days, he then kept iterating based on customer feedback and pushed a new version every day for five days until landing on a finalized product.
“We essentially did the full product life cycle in a week,” Breitenother told Fortune.
This isn’t a one-off, but rather exactly how Kilo’s engineering team is designed to operate. When Breitenother and his cofounder, Sid Sijbrandij, former CEO of GitLab, launched the company in early 2025, they made the deliberate decision to remove the product management layer and give engineers the license to work autonomously and even ship of their own accord. Responsibilities typically held by product managers—like product discovery and prioritization decisions—sit directly with engineers, whom Breitenother describes as operating like “mini CEOs” of their own products.
The intention is to enable the company to move at true lightning speed. Even with experienced cofounders and $8 million in seed funding from top VCs like General Catalyst, Kilo is operating in a rapidly evolving and highly competitive space, vying with popular tools like Claude Code, Cursor, and GitHub Copilot.
Breitenother sees this way of working as both made possible by AI and a necessity for the company in today’s fierce AI-driven market. Now that AI is capable of writing the code engineers once spent their days on, it allows them to focus on other aspects of the product workflow, particularly those that only humans can address, like setting the vision and making strategic architecture decisions, he said. In the near future, he believes this is how all engineers—and people in all sorts of roles—will work, too.
Killing the ‘velocity killers’
“Adding additional cooks in the kitchen slows things down,” Breitenother said, discussing his decision to eliminate the product management layer.
Coding is faster than ever, but he’s seen that collaboration, handoffs, and processes make projects crawl instead of run. So he sought to eliminate these “velocity killers” and have engineers take on multiple roles in one, describing them as “part strategist, part technical architect, part product manager.”
“If we can kind of make this as single-threaded, with end-to-end ownership by an individual being able to execute, that’s how we move really fast,” he said. “So it’s like this combination of AI enabling the human, but also you’ve got to get those velocity killers out of the way. You can peanut butter as much AI as you want onto an existing process, but if the existing process is slow, you’re not going to see those benefits.”
Breitenother takes the approach so far that he expects new hires to make a couple of code commits on their first day.
“Sometimes they’re like, ‘Oh, ha ha.’ And I’m like, ‘No really,’” he said, adding that people have been conditioned to seek approval, have alignment meetings, and go through “bureaucratic corporate speak.”
“It’s just so exciting to see folks kind of open their eyes as they join Kilo [and see] that we’re serious: ‘You own this product. Go for it.’ And then, once it clicks—I mean, the speed is phenomenal,” he said.
Remon Oldenbeuving, an engineering manager at Kilo, said being the CEO of his product means not only identifying what to build, prioritizing features, and shipping them, but also making sure people know about it and use it. He actually has experience working without a product manager at a past company, yet his autonomy at Kilo is next-level.
“Instead of working on a project alone, you’d be part of a team, and the team had to stay in sync, so there was much more communication before the project was completed,” he said, describing workflows at his previous company. “At Kilo, we focus on shipping as often and as fast as possible. The fastest way to achieve that is to reduce the number of lines that must remain open.”
Asked about the risk of burnout for engineers moving at this pace while juggling various parts of strategy, product management, and technical work, Breitenother says it’s a fair question. The goal, he said, isn’t to run people at maximum velocity all the time, but rather to build systems where humans set direction and AI handles more of the heavy lifting. He believes it’s not about asking them to take on more, but rather work at a higher level of abstraction. And their salaries are still in line with what senior engineers at a startup at this stage would be paid, he said.
Growing an ultra-autonomous company
While Breitenother credits the engineering team’s speed and early successes—including shipping 15-plus features a week, often occupying the top spot on OpenRouter (a platform where developers compare models), and scaling to over 1 million developers in under a year—to this strategy, it’s not limited to engineering. He’s taking a similar approach across the organization.
Companywide, Breitenother said, Kilo has no entry-level talent. Everyone is at the senior level, and many individual contributors were formerly directors before coming to Kilo, enabling the sort of autonomy he’s fostering.
Kilo did recently hire a product manager, but it’s to keep this momentum, not go back to the usual ways. The PM’s job is to work on the core infrastructure and components that span all of Kilo, like the polish and consistency, that ultimately allows each of the engineers to operate on their own.
“Our engineers have really adapted to being their own product manager,” Breitenother said. “But you know, it’s always helpful to chat with someone who’s done this for 20 years.”
The PM is one piece of steadiness in what otherwise seems like a freewheeling system, along with an AI program designed to catch mistakes in the company’s AI-generated code. But overall, Breitenother doesn’t seem very concerned about what would appear to many like an astonishing lack of safeguards. In terms of drawbacks to the approach, he said Kilo hadn’t experienced any problems beyond minor issues like mismatched copy between two pages, and that “minor inconsistency is an acceptable price for speed.”
The real question is, however: Could a larger enterprise bear these kinds of risks? When Breitenother and Sijbrandij first started the company, he said they challenged themselves to think about what a team will look like in 2028. Acknowledging that being a startup is often what enables faster speeds, Breitenother doesn’t think their startup status is what’s driving it. In fact, he believes AI will allow larger firms to soon architect themselves around smaller, more dynamic units that can enable individuals. They won’t have a choice, with competitive pressures forcing them to go beyond adding AI as tools and reimagine their processes.
“I’m going to try my darndest to keep this forever,” he said. “We do think that companies of all sizes will be operating in this model in the future.”
Read more from the latest Fortune AIQ special report, which highlights the one strategy, tool, or approach companies are using to bring the most out of enterprise AI.
