Podcast: Scott Jenson on Evolving Desktop OS, Local-First, & Agentic

What happened

Scott Jenson spent decades shaping the interfaces people use every day, from early Macintosh software to Google Maps and Chrome, and he has now turned his attention to a question that rarely gets asked directly: why have desktop operating systems stopped evolving in any meaningful way? In a recent episode published by InfoQ, Jenson walks through his thinking on stagnation at the OS level, the shortcomings of mobile and cloud-first design, and what a more capable computing model might look like when local data and autonomous agents enter the picture.

This piece breaks down what Jenson argued, how the underlying concepts work, and what the conversation signals for anyone building software today. It is aimed at developers, product designers, and technical architects who care about the direction of computing platforms and want to understand the structural pressures shaping the tools they rely on.

Scott Jenson, a veteran UX designer with credited work on the Macintosh, Google Maps, and Chrome, sat down with InfoQ for a podcast conversation that covers some of the most underexamined tensions in mainstream computing. The episode focuses on three overlapping themes: the long-term stagnation of desktop operating systems, the limitations that mobile and cloud-centric models impose on users, and what agentic user experiences might demand from the software layer beneath them. Jenson brings a practitioner's lens to these topics, drawing on direct experience shipping products at organizations that helped define the categories he is now critiquing.

How it works

The conversation is notable because it comes from someone with insider experience at the companies most associated with the platforms under scrutiny. Jenson is not an outside critic speculating from a distance. He worked inside the systems he is describing, which gives his observations about structural stagnation a different weight than typical industry commentary. InfoQ, which publishes practitioner-focused content for senior developers and architects, is an appropriate venue for this kind of technically grounded design critique.

Jenson's argument about desktop OS stagnation rests on the observation that the fundamental interaction model has not changed substantively in decades. Users still manage files, launch discrete applications, and move data manually between those applications. The operating system acts as a passive container rather than an active participant. Cloud-centric models moved data off the local machine but did not change the underlying interaction metaphor. Mobile made the container smaller and the application silos tighter, which Jenson characterizes as a step backward for user capability rather than a step forward.

Why it matters

The local-first framing Jenson discusses proposes that data should live on the user's device by default, with synchronization to remote infrastructure as a secondary concern rather than a primary architectural dependency. This approach preserves offline access and gives users direct control over their information. When agentic software is layered onto a local-first architecture, the agent can act on data that is present and accessible without routing every operation through a remote API call. That changes both latency and the nature of what the agent can reliably do on the user's behalf.

The stagnation Jenson describes has compounding consequences for the next generation of software products. If the OS layer does not evolve to support agentic workflows natively, developers will be forced to build brittle workarounds that approximate the capabilities they need. Application silos, which already make cross-app automation difficult, become a harder obstacle when an agent needs to observe context, take actions, and reason across multiple data sources simultaneously. The current model essentially requires agents to interact with software the way a human would, by navigating UIs designed for manual use, rather than through richer programmatic interfaces.

For builders, this matters because design decisions made at the OS and platform level constrain what is possible at the application level years before any product ships. Jenson's critique suggests that teams working on agentic features should pressure-test their assumptions about data access, context persistence, and cross-application coordination early. A product that depends on a capability the underlying platform does not provide is in a fragile position regardless of how well-designed the product itself is.

Who is affected

Developers building consumer or enterprise software on top of existing desktop and mobile platforms will feel the constraints Jenson describes most acutely when they attempt to add agentic or AI-assisted features. Product designers working on complex workflows that span multiple applications will recognize the file and clipboard limitations he identifies. Platform teams at operating system vendors are the actors best positioned to change the underlying situation, and Jenson's conversation implicitly poses questions to them about what investment in the interaction model would look like.

Users who depend on sophisticated desktop workflows, particularly knowledge workers who routinely move information between many applications, are the population most directly limited by the current state of affairs. They have adapted through habit, scripts, and manual effort to compensate for what the platform does not provide.

What to watch next

The local-first software movement has produced concrete specifications and tooling, and tracking that work is a practical starting point for anyone who wants to understand what implementation looks like. Agentic UX is still early enough that few platforms have committed to a native model for it. Watching how operating system vendors handle the intersection of local data access, agent permissions, and user control will be informative over the next several release cycles.

Jenson's full conversation is available through InfoQ for listeners who want the unabridged argument. For builders, the most useful exercise is auditing a current project for places where the design implicitly assumes manual user coordination that an agent would need to perform instead. That audit will surface the concrete platform gaps faster than any general discussion of stagnation can.