At QCon San Francisco 2026, Jordan Cutler, Senior Staff Engineer and Web Platform Tech Lead at Pinterest, presented how the role of a staff engineer changes when every engineer has access to the same AI tools. The session argued that most "AI for engineers" content focuses on individual productivity, while the real question for staff engineers is where their leverage comes from now. According to Cutler, AI did not change what a staff engineer is for, but it changed where the leverage is.
The talk opened with a reflection on Cutler's previous conference talk, where he showed how AI could turn a four-hour task into four minutes. At that point he was shipping roughly as much as his whole team, which he described as multiplying himself rather than his team. That framed the central question of the session: what does a 10x engineer look like when everyone has AI?
To illustrate why this matters, Cutler shared internal Pinterest data on merged pull requests per engineer per week between Q1 2025 and Q3 2026. The top 10% of engineers went from 6.7 to 11.3 merged PRs per week, a 68% increase, while the bottom half went from 1.0 to 1.15, an increase of 14%. Everyone has AI, but not everyone uses it the same way, and the gap keeps growing. The presentation organized the response to that gap into three moves: multiply yourself, multiply your team, and multiply the system.
The first move, multiply yourself, is about going first. Cutler explained that he tries new tools before his team does, and that part of that work is hitting dead ends so others don't have to. He mentioned examples such as an agent harness that could not be steered mid-task and a remote agent runner that required too many follow-ups. Sharing those findings early means the team does not lose a week on the same experiments. Running many agents also revealed a new problem: he kept giving the same corrections in every session, such as avoiding em-dashes, adding a diagram to the PR description, and keeping pull requests small.
The second move, multiply your team, started from a limitation: the folder made him faster, but only he had it. With new people joining the team and a new area to ramp up on, Cutler found himself repeating the same comments in code review, such as asking why a value is null instead of fixing the cause, testing like a user would, and keeping PRs small. The talk emphasized that the value of encoding that judgment grows with every teammate who adopts it, compared to the flat curve of a single engineer working alone.
The third move, multiply the system, is about moving up a level. Cutler used the history of container shipping as an analogy: ports got better at cargo by hand, but the standard box changed the economics entirely, from $5.86 a ton to 16 cents a ton. At Pinterest, migrations were the cargo, one kind of change spread across hundreds of files and dozens of teams. Instead of running each migration by hand, the team built a system where a staff engineer points to a plan and an example, an engineer puts it on a schedule, and an agent running in a sandbox splits the work into PRs team by team. A class component migration ran three to five times faster than by hand, in the background while teams worked on other things.
The talk also covered what went wrong. One scheduled job looped and opened 16 PRs in two minutes, increasing CI wait times from under 30 seconds to 14 minutes. The lesson was not to wait for the perfect guardrail, but to ship a small one. Cutler encouraged attendees to find their own "any" problem: a first responder for any incident, a draft fix for any bug ticket, or an agent that triages any flaky test. Today, 40 engineers have scheduler jobs of their own.
The results were significant. Using the same AI tools as other web teams, Cutler's team now ships about four times as much per engineer, without counting his own PRs.
The presentation concluded with practical steps: write down what you repeat, put it where people already work, and build for every case instead of one. Some parts of the job cannot be written down, like an honest conversation with a teammate who is struggling, and those remain the staff engineer's responsibility. As Cutler put it, a 10x engineer isn't ten times faster, they are the reason ten other people are. As AI agents become part of everyday development workflows, the role of senior engineers may increasingly shift from personal output toward encoding judgment into the tools, defaults, and systems that entire organizations use.