Software Slivers
Very narrow capabilities could become sustainable products when agents take over discovery, integration and operation.
Emlyn O'Regan
A software sliver is a small, specialised capability offered independently rather than as a feature inside a larger application.
Today, narrow products can be hard to sell. Finding customers, explaining value, building onboarding, maintaining a UI and integrating with other applications may cost more than the problem justifies.
Agents could lower many of those costs. If the capability is discoverable, understandable, callable and appropriately priced, an agent can select it precisely when needed.
A specialist audio transformation, a document conversion operation or a narrowly focused analysis service might then compete on its own merits.
Disaggregation changes the competitive unit
A large software suite has traditionally benefited from bundling: the customer is already there, the UI exists, and integration is built in. But when an agent can compose capabilities cheaply, a bundled feature must compete more directly with a specialist alternative.
This connects to the different 10% problem. Instead of every customer carrying the same vast feature bundle, agents can select different combinations.
Not everything should be a sliver
Small pieces still have costs: latency, dependencies, billing, permissions, shared state and operational failure. Some functions genuinely work better together. A useful service boundary must account for those costs.
The thesis is not that all applications should be shattered into tiny services. It's that the economically sensible size of a software product may decrease as discovery and composition get cheaper.
See Agents as Software Discoverers and The Disaggregation of Software.