Software Should Adapt to You
Why accommodation is the next software paradigm.
Accommodating software is software that changes its context, tools, workflows, permissions, and interface around the person or organization using it, instead of requiring the user to conform to a fixed application.
Executive summary
Most software asks the user to adapt. We learn the interface, adopt the application's vocabulary, translate our work into its schema, and move context between systems when the software cannot.
Generative AI creates the possibility of reversing that relationship. Software can increasingly assemble memory, context, capabilities, rules, and interfaces around the way a person or organization already operates.
This is more than personalization. A personalized application still has a mostly fixed shape. An accommodating system can have a different operating environment for each customer: different knowledge, tools, permissions, workflows, and even different interfaces.
The implication is that the durable product boundary may move away from the application and toward the AI harness: the system surrounding the model that determines what it knows, what it can do, and what it is allowed to do.
The inversion
For decades, the user has been the integration layer.
A CRM has its model of a customer. A calendar has its model of time. A project-management system has its model of work. A finance product has its model of money. The user is responsible for reconciling the differences.
We call software flexible when it gives us enough settings to make that translation less painful. But the basic contract remains the same: understand the application, then reorganize your work to fit inside it.
Accommodating software proposes the opposite contract. The system learns enough about the user's world to meet the user closer to where the work already happens.
A useful distinction
Fixed → configurable → personalized → accommodating
Fixed
The same workflow and interface for everyone.
Configurable
The user changes settings inside a predetermined system.
Personalized
The system changes content or recommendations for the user.
Accommodating
The system can change context, tools, workflow, permissions, and interface around the user.
The central argument
A custom prompt changes what an AI says. A custom harness changes what the system knows, can do, and is allowed to do.
Models are becoming powerful general-purpose components. The differentiating system increasingly lives around them: durable memory, structured knowledge, tools, integrations, workflow state, permissions, approval rules, audit history, interfaces, and feedback loops. Those pieces can be assembled differently for different people and organizations.
White paper storyboard
The full research edition will expand each argument with evidence, examples, diagrams, and source notes.
The argument, section by section
Software taught us to behave like software
For decades, the person has been the integration layer: learning interfaces, translating real work into rigid data models, and moving context between applications.
Personalization was not adaptation
Configuration changes settings. Personalization changes presentation or recommendations. Accommodation changes how the system itself operates around the user.
The model is not the product
A useful AI system also needs memory, context, tools, permissions, workflow state, audit history, interfaces, and feedback loops. The model is one component inside a larger system.
The harness becomes the new unit of software
A custom prompt changes what an AI says. A custom harness changes what the system knows, what it can do, and what it is allowed to do.
Memory makes customization compound
When the system can retain people, preferences, decisions, exceptions, and history, its usefulness can increase over time in a way that a stateless application cannot reproduce cheaply.
Interfaces become temporary
Conversation, dashboards, tables, forms, voice, and automation become task-specific views over the same durable context and capabilities rather than separate silos.
Trust becomes architecture
Adaptive systems need explicit boundaries: identity, permissions, approvals, provenance, audit trails, revocation, and inspectable memory.
What happens to SaaS?
Standard systems remain valuable where standardization matters. What changes is the layer people interact with: intelligence can increasingly mediate the applications underneath.
Why memory matters
Models can be replaced. Accumulated context cannot be replaced cheaply.
Day 1
The system knows what you explicitly connect and tell it.
Month 1
It begins to reflect vocabulary, people, preferences, recurring work, and exceptions.
Month 6
It contains operating history: decisions, commitments, relationships, and patterns.
Year 2
The system represents a world that cannot be recreated by copying a prompt.
Direct answers
Frequently asked questions
- What is accommodating software?
- Accommodating software is software that changes its context, tools, workflows, permissions, and interface around the person or organization using it, instead of requiring the user to conform to a fixed application.
- How is accommodating software different from personalization?
- Personalization usually changes content, recommendations, or presentation inside a mostly fixed product. Accommodating software can change the operating environment itself: which context is available, which tools exist, how work flows, what actions require approval, and which interface is appropriate.
- What is an AI harness?
- An AI harness is the runtime around a model that supplies memory, context, tools, permissions, workflows, execution rules, and interfaces. The harness determines what the AI knows, can do, and is allowed to do.
- Does this mean SaaS disappears?
- No. Standard software remains useful for standardized systems of record, infrastructure, payments, identity, compliance, and other shared primitives. The change is that an adaptive intelligence layer can increasingly mediate how people interact with those systems.
EvaNeT point of view
We are building EvaNeT as an intelligence layer that can be shaped around the individual.
Persistent memory, structured context, connected tools, explicit permissions, governed action, and multiple interfaces are not separate features. Together, they are the machinery required for software to accommodate the person using it.