Liquid Software
A New Age of Computing & The Unbundling of Software into Logic, Data, and Interface
There was once a dream called the ‘personal computer.’ A machine that belonged to you—where you could create, modify, break and rebuild. Gradually over the past 30 years, the technology ecosystem has lost sight of this dream. As software shifted to the cloud and business models centered on distribution at scale, we lost that intimacy. The “personal” was stripped from the PC. In its place: apps and platforms, designed for control, consistency, and compliance.
And yet, breakthroughs in AI are re-opening the door to something we lost: flexibility, agency, and a glimpse into truly personal computing. As models become increasingly capable, it is clear by now that we’re witnessing the erosion of the traditional software stack. AI agents don’t need buttons and tabs; they need access and intent. A user no longer wants to “open the CRM” — they want to “see which leads are most likely to close.” The interface is abstracting. The app is dissolving.
We’re entering the era of liquid software.
Liquid software is what happens when AI breaks software into its base elements—logic, data, and interface—and recomposes them dynamically through reasoning. It’s the unbundling of the app into reusable capabilities, orchestrated by agents, and personalized on demand.
This is already happening. The precursors are here:
LLMs now act as glue across APIs, tools, and data stores
RAG turns static apps into dynamic knowledge surfaces.
Autonomous agents are executing workflows with no UI at all
Plugin and action ecosystems (OpenAI, LangChain) are shifting toward an OS for composable tools.
MCP is emerging as a unifying framework to pass memory, tools, and state between agents and applications, enabling a fluid, inter-operative software mesh.
The app, once monolithic and shrink-wrapped, is melting into a series of permissioned API calls, vectorized memory, and real-time reasoning.
Ramp’s vision for autonomous money. The Browser Company rewriting the interface of the browser with Dia. Devin writing, debugging and deploying code across multiple systems without a formal UI. All powered by a new paradigm in computing — liquid software that is dynamic, proactive and in many cases invisible.
This moment wasn’t possible in the Web 2.0 or mobile eras. We didn’t have models that could reason extensively, APIs that were open enough, or orchestration frameworks that made trustable autonomy viable. Now we do. Just like the browser abstracted the OS, and the app store abstracted the web, AI is abstracting the app. The walls are coming down. Liquid software is what flows through the cracks.
What does this unlock?
Truly personal computing. Software that adapts to the user, not the median workflow. Agents that understand your context, preferences, and goals—and act on your behalf
Explosive interoperability. Context-aware agents reason across tools without rigid integrations. The future filesystem looks more like a live graph than a directory tree.
New primitives for product design. Apps become capabilities. Interfaces become optional. The moat becomes reasoning, memory, and fine-grained control over composition.
This is a foundational shift in computing. The winners won’t be the ones who build better apps—but the ones who dissolve the app entirely.
Liquid software brings us closer to what the personal computer was always dreamt to be: not a cage, but a canvas