GitLab's 'Act 2' Is a Blueprint for the Agentic Era
GitLab cuts management and bets on 60 autonomous R&D teams to handle the speed of agentic AI.
GitLab is restructuring into 60 autonomous R&D units and flattening its entire management structure. The company is also reducing its country footprint by 30%. This isn’t a standard cost-cutting layoff. It is a structural bet that AI agents are a better coordination mechanism than human middle managers.
The Org Chart Flips
GitLab Act 2 is the clearest signal yet that the software org chart of 2023 is dead. We spent years scaling teams by adding layers of process, project managers, and scrum masters. GitLink is betting those layers are a liability in the agentic era. If an AI coding agent can triage bugs, review code, and surface context, what is the value of a manager whose job is just to relay information? You don’t need a hierarchy. You need 60 independent pods armed with agents, pointed directly at the product.
As Simon Willison notes, the logic is that small teams with direct accountability can move much faster when backed by AI. Management overhead doesn’t just slow things down. It prevents the tight feedback loop that makes agentic work effective.
The Maintenance Trap
“Your AI coding agent, the one you use to write code, needs to reduce your maintenance costs. Not by a little bit, either. You write code twice as quick now? Better hope you’ve halved your maintenance costs. Three times as productive? One third the maintenance costs.”
This warning from James Shore is the most important technical insight of the year for builders. The whole point of “ship more” is ruined if shipping creates an exponentially growing pile of broken shit. GitLab’s reorganization is the structural response to this trap. By keeping teams autonomous, they force ownership. You can’t hand a bad PR to someone else. You write the code, you own the maintenance, you build the agents to handle the fallout. The speed gain must be matched by a debt reduction engine. The unit structure is that engine.
No Zombie Codebase
We are watching the rise of the Zombie Internet — AI-generated content flooding search results, draining meaning from the web. The same disease is infecting software. Wall-to-wall AI PRs. Dead code paths. Features no one asked for. GitLab’s autonomous team model is designed to prevent this. When a team of 8 people owns a complete service stack, they feel the pain of every bad line of code an agent writes. The feedback loop is instant. There is no buffer zone where shitty code can hide. The unit becomes self-cleansing, or it fails.
The Signal
GitLab is the first major platform to publicly redesign its engineering structure for the agentic era. The signal is loud: the org chart is the biggest bottleneck to shipping faster. You cannot slap agents on top of a 2022 management structure and expect it to work. You have to thin the herd, shrink the units, and increase accountability. Management overhead is a liability. Maintenance agility is the only lasting moat.
Ship more by getting smaller. Strip your layers, shrink your teams, and force them to own the full lifecycle. GitLab drew the blueprint. Follow it.