The hybrid compromise is dead. What the best companies are doing instead.

The "three days in office, two at home" policy was never a strategy. It was a truce. By early 2026, the companies that are still growing have mostly abandoned it — in both directions.

The best fully-remote companies look nothing like the 2021 version. They hire in six time zones, ship by default in writing, and require zero synchronous meetings for 80% of roles. The hiring funnel is global; the payroll is distributed; the software is opinionated about async (Notion, Linear, Loom, Granola). Offsites are quarterly and intense — three days of shipping together, not trust falls.

The best in-office companies also look nothing like the 2019 version. They're small (usually under 120 people), they expect five days in a specific city, and they're brutally honest about why: speed. The CEO, the head of product, and the lead engineer can walk from lunch to a whiteboard in ninety seconds. The compensation has to be high enough that candidates choose the trade.

What almost nobody runs well anymore: the hybrid that tries to serve both.

The hybrid compromise — "be here Tuesday, Wednesday, Thursday" — lost because it solved for neither audience. Remote candidates who wanted global flexibility went to remote-first competitors. Office-first operators who wanted the speed of co-location got two days of it, then watched everyone disappear on Monday and Friday.

The interesting data point: when a company finally commits — really commits, one direction or the other — retention climbs. Employees who wanted the other thing leave in the first ninety days. The ones who stay are exactly aligned with the bet. Trust, for once, is not the variable. Cadence is.

If you're choosing for your company right now, the honest question is not "what do people want?" It's "what are we actually optimizing for?" If it's speed of decisions, go in-person. If it's quality of talent, go remote. If you try to do both, you'll do neither well.

The companies that ship in 2026 picked one.