Coinbase cuts 14% of staff to go AI-native

Coinbase is taking the idea of an AI-native company to its most extreme conclusion yet.

CEO Brian Armstrong just sent a companywide email announcing a roughly 14% workforce reduction. But instead of framing it as standard belt-tightening, Armstrong laid out a radical vision: Coinbase is being rebuilt as an intelligence, with humans sitting around the edges to align it.

Two compounding forces drove the decision. First, the crypto industry is currently in a down market, and the company needs a lean cost structure to navigate quarterly volatility. Second, AI has completely upended how their teams build software.

Death of the pure manager

According to Armstrong, Coinbase engineers are now using AI to ship projects in days that previously took entire teams weeks. Even non-technical staff have started pushing code directly to production as internal workflows automate away.

To match that new reality, the company is tearing out organizational red tape. Coinbase is flattening its structure to a maximum of five layers beneath the CEO and COO. Managers will oversee far wider spans of control, with some leading more than 15 direct reports.

Crucially, Coinbase is eliminating pure managers entirely. Every leader must now act as a player-coach who actively contributes code or individual work alongside their team. The company is also experimenting with single-person pods, where one employee leverages agent fleets to handle engineering, design, and product management alone.

Departing US staff receive at least 16 weeks of base pay, two additional weeks per year of service, their next equity vesting tranche, and six months of healthcare coverage. However, system access was revoked immediately to protect customer data.

Why it matters

We think Coinbase just handed the tech industry an unvarnished preview of where enterprise AI is actually heading. For over a year, executives have promised that generative tools would augment staff; Armstrong is acknowledging out loud that massive productivity gains inevitably lead to smaller headcounts.

For tech workers, this is a clear warning shot against middle management. The era of the pure coordinator is rapidly closing. Professionals who survive these restructuring waves will need to be high-output individual contributors capable of orchestrating multi-agent systems.

For the broader industry, this experiment will either validate or caution against the hype of the single-person team. While a single engineer-designer-PM hybrid sounds incredibly fast on paper, it introduces massive single-point failure risks. In a highly regulated crypto exchange where a single bug can cost millions, running teams this lean is a high-stakes gamble that rivals will be watching closely.