Ona (formerly Gitpod) ads

Most AI coding demos end when the human stops typing. Zach's experiment asks a bigger question: what happens when AI engineers keep shipping after people step away? 375 PRs in 10 days. 87% merged without human action.

Agentic engineering is not about replacing engineers. It is about giving teams more capacity for the work that piles up: ops triage, routine changes, and real development. Genentech shows what happens when AI gets context, safe boundaries, and scheduled work.

The companies that win with AI will not be the ones that ban it. They will be the ones that make it safe enough to use. Monzo shows how regulated teams can give AI real access without losing control of data or governance.

Catch me at Ona's Background Agents Summit on May 6-7! I'm talking through what we've learned building minions, Stripe's background agents primitive. Here's a teaser talking about one of our main lessons: that agent quality is a dev infrastructure problem. Isolated devboxes that spin up quickly with reproducible dev environments were built to make engineers at Stripe productive -- but they turn out to be exactly what make agents productive too! If you're figuring out what it takes to make coding agents actually useful inside an engineering org, come check it out. Register at https://lnkd.in/eQ2fNNGK

I'm speaking at the Ona Background Agents Summit on May 6-7 about what I’ve learned building Spectre here at Harvey, which Gabe Pereyra and I wrote about last month. A major theme I will be touching on is alignment. Before anyone writes code, humans have to get aligned. That's what tools like Linear and Jira have always been for. With agents in the mix, that problem gets bigger. It's not just human-to-human anymore. It's humans and AI all needing to be on the same page. With Spectre we're taking a different approach. Instead of discussing in a thread, handing off to an agent, getting PRs back, then discussing those again on GitHub or in a meeting, just give the agent the full context from where the conversation already happened. It knows what everyone decided. The end result should basically be pre-approved. If you're thinking about how agents fit into team workflows beyond solo coding, come check it out. Register at https://lnkd.in/gP9AKVVP

AI coding agents make more PRs possible. They also make review harder. At the Background Agents Virtual Summit on May 6, we are getting into the questions engineering teams cannot dodge: what still needs human review, where agent review is enough, and how risk changes across internal tools, docs, and production surfaces. Register to learn how teams are adapting review without giving up accountability.