Detailed Analysis
A Reddit user's complaint about Anthropic's "Cowork" feature highlights a fragmentation problem that's emerged as Claude has expanded across platforms. The poster describes having Cowork available on both their phone and the web app, but discovers that project state isn't synchronized with their existing desktop app Cowork projects. This creates three separate silos of work rather than a unified experience, forcing the user to ask whether migration between local and cloud-based state is even possible, and how this all relates to Anthropic's existing "Projects" feature within the standard Claude chat interface. The confusion suggests that as Anthropic has rapidly shipped new surfaces for interacting with Claude, the underlying data architecture hasn't caught up to provide a coherent, cross-device experience.
This kind of platform sprawl is a familiar growing pain for fast-moving AI companies. Anthropic has been aggressively expanding Claude's presence beyond the core chatbot interface — adding desktop apps, mobile apps, browser-based coding environments, and now apparently a "Cowork" feature that appears to function as some kind of collaborative or agentic workspace distinct from the standard Projects feature. Each of these surfaces likely originated from different engineering teams or sprints, built with different backend assumptions about where state lives (local device storage versus cloud-synced accounts). Users who adopt these tools across multiple devices are the first to hit the seams, and this Reddit post is essentially a bug report happening in public — a common pattern where product gaps surface through community frustration before official acknowledgment or fixes.
The user's closing jab — "it's looking like these guys are vibecoding their actual ideas now" — is a pointed and somewhat ironic critique. "Vibecoding" is a term that emerged from the AI coding community (with Anthropic's own tools like Claude Code being central to that trend) to describe building software quickly and intuitively without rigorous planning, often by leaning heavily on AI assistance rather than careful architecture. The accusation here is that Anthropic itself may be shipping features in this same fast-and-loose style internally — rolling out Cowork across platforms without solving basic synchronization and data continuity issues first. It's a notable turn of phrase because it applies a critique born from AI-assisted development culture back onto the company most associated with popularizing that culture.
More broadly, this incident reflects a tension inherent in the current AI product landscape: companies are racing to ship agentic and collaborative features across every possible surface (desktop, web, mobile) to capture user attention and usage data, but the infrastructure work of unifying state, sessions, and project history across those surfaces is harder and less flashy than launching the features themselves. As AI assistants increasingly become persistent collaborators on multi-step projects — coding, research, writing — rather than one-off chat sessions, users will expect continuity across devices in the same way they expect it from cloud-native productivity tools like Google Docs or Notion. Anthropic's Cowork stumble is a small but telling signal that the industry's rapid feature velocity is currently outpacing the less visible, but equally critical, work of building seamless, unified user experiences.
Read original article →