Artificial General Software

For decades, the quest for artificial general intelligence (AGI)—systems capable of human-level cognitive performance across any task—has captured the imagination of scientists and technologists. While debates around AGI’s feasibility and timelines continue, a parallel and perhaps equally transformative revolution quietly emerges: the era of artificial general software. As language models rapidly expand in capability and scope, we approach an inflection point where specialized software applications begin to yield to AI agents capable of executing virtually any computational task through natural, intuitive interaction.

Beyond Specialization

Traditional software development thrives on specialization. Each application carefully encodes domain knowledge into focused workflows: project management tools embed methodologies of task organization, design software crystallizes visual manipulation principles, and financial platforms encapsulate accounting logic. Specialization has dramatically advanced productivity, enabling software to permeate every aspect of human endeavor—but at the cost of fragmented functionality, steep learning curves, and artificial boundaries between related tasks.

Consider today’s typical knowledge worker navigating between Slack for communication, Asana for project management, Google Docs for writing, Figma for design, and Notion for knowledge management. Each transition represents a cognitive context switch, imposing recurring overhead. This fragmentation constrains our ability to work fluidly across tasks and domains, limiting productivity and creativity.

Rich Sutton’s “bitter lesson” in machine learning suggests that general, scalable computational methods inevitably outperform specialized, hand-crafted solutions. Although originally aimed at machine learning algorithms, this insight naturally applies to software architecture itself. Just as AGI promises to supersede narrow AI, general software now threatens to render specialized applications increasingly obsolete.

The Architecture Shift: From Interfaces to Intentions

Traditionally, SaaS applications rely on carefully designed interfaces that decompose tasks into discrete interactions—buttons, forms, menus—that require users to master each application’s specific interaction model. Even the most advanced graphical interfaces require us to translate our intentions into the software’s predefined language, imposing cognitive burdens and restricting flexibility.

AI-driven software reverses this paradigm. Rather than users adapting to fixed interfaces, AI agents adapt dynamically to user intentions expressed in natural language. This shift is profound, dissolving traditional interfaces into conversational exchanges. A user no longer needs to master Photoshop’s complex menus to remove a background from an image; instead, they simply say, “Remove the background and make it transparent.” Rather than configuring project management tools manually, users instruct, “Create a new project for our Q1 marketing campaign with standard milestones and assign team responsibilities.”

This transformation renders conventional UI design philosophies—consistency, predictability, visual hierarchy—secondary, even counterproductive. The primary challenge shifts from interface design to developing intelligent agents capable of accurately interpreting and executing user intent.

Economic Implications: Commoditizing Interfaces

This architectural shift triggers profound economic consequences. Traditional SaaS businesses build competitive moats around user familiarity, switching costs, and network effects—factors rooted firmly in specialized interfaces and functionalities. Yet, when interactions become natural language and capabilities derive from powerful foundation models, these traditional defenses erode rapidly.

Companies like Adobe, Atlassian, and Microsoft currently secure market positions through complex ecosystems of interrelated tools and steep user learning curves. High switching costs persist because changing tools means disrupting established workflows. But in a landscape dominated by general software, these barriers diminish dramatically. Competitive differentiation shifts from specialized features and interfaces to the quality of an AI agent’s comprehension, adaptability, and execution.

This shift mirrors earlier disruptions: cloud computing commoditized infrastructure, APIs democratized service integrations, and now artificial general software threatens to commoditize the application interface itself. As a result, value moves up the stack—from providing specific functionality toward orchestrating complex actions across diverse domains.

Navigating the Transition Period

The journey toward artificial general software will likely echo previous technological transitions. Initially, AI agents will appear as intelligent assistants embedded within traditional applications. Gradually, these assistants will subsume more functionality, eventually relegating traditional interfaces to secondary status. Ultimately, specialized interfaces may disappear altogether, replaced entirely by direct interaction with increasingly capable AI agents.

Early glimpses of this evolution already exist. GitHub Copilot transitions from code completion assistant to full-fledged pair programmer. ChatGPT evolves from conversational AI into a general-purpose computational interface. These precursors suggest a future where the boundaries between user and system interactions become increasingly fluid.

This transition raises fundamental questions about the future of software engineering itself. When applications become dynamically interpreted expressions of user intent rather than static implementations of functionality, what remains for software developers? The answer lies in shifting focus from implementing specific features to enhancing the comprehension, reliability, and safety of general-purpose AI agents.

In practice, developers will increasingly concentrate on refining agents’ domain understanding, execution accuracy, and robustness—paralleling the shift from feature engineering to architecture design and prompt engineering we have already witnessed in machine learning.

Addressing the Human Element

With general-purpose software comes new challenges in human-computer interaction. Traditional interfaces explicitly communicate capabilities and limitations through visible controls and workflows. Natural-language interactions, while fluid and intuitive, introduce ambiguity around system capabilities and boundaries.

As discussed in The Thinking Placebo, users must now form accurate mental models of what these sophisticated AI agents can and cannot do. When a user instructs the AI to “improve this document,” what precisely should they expect? This ambiguity demands new approaches for user education, expectation management, and maintaining user agency amid opaque computational processes.

Potential solutions include:

  • Clearly defining and communicating capability boundaries to users.
  • Progressive disclosure of complexity, enabling deeper exploration when needed.
  • Feedback mechanisms allowing users and systems to iteratively refine interactions.
  • Safety measures guarding against unintended consequences from ambiguous instructions.

Embracing a New Era

The rise of artificial general software signals a fundamental transformation in our relationship with computing. Just as AGI promises a new era of intelligence, general-purpose software heralds a shift in the core architecture of how humans interact with computational systems. The dissolution of specialized applications does not signify the end of software but rather its evolution toward greater flexibility, nuance, and responsiveness.

This evolution directly intersects ethical considerations raised in The Human Problem. As software grows more capable and general, alignment with human values and intentions becomes critical. How do we ensure general software acts in accordance with shared human values? How do we handle disagreements over those values?

Every major technological advance brings both anticipation and apprehension. The dissolution of traditional software boundaries may seem radical today, but it represents the logical culmination of computing’s longstanding pursuit of natural human-computer interaction. Artificial general software does not simplify computing; rather, it deepens its complexity to match the richness of genuine human intention.

Looking Ahead: Key Areas for Attention

As we navigate this transformative shift, several key areas demand particular attention:

  1. Interface Evolution: How will visual interfaces evolve as natural language becomes primary? Interfaces may persist, but increasingly as dynamic visualizations aiding user understanding rather than as primary interaction mechanisms.

  2. Capability Assessment: How can we measure and communicate fluid, context-dependent AI system capabilities? Traditional feature-completeness metrics lose relevance as capabilities adapt dynamically to user intent.

  3. Security and Privacy: How do we preserve security and privacy in systems designed to fluidly execute tasks across multiple domains? Traditional application boundaries and isolation strategies may need rethinking.

  4. Economic Disruption: How will the software industry adapt to dissolving application boundaries? New business models may emerge around providing foundational capabilities and orchestrating complex actions and integrations.

  5. Educational Impact: How should software education evolve to prepare future developers and users for general-purpose software? Curricula must shift from specialized tools toward broader principles of human-AI collaboration and adaptability.

The advent of artificial general software represents more than just technological innovation; it constitutes a fundamental reshaping of human-computer relations. As we move forward, we must embrace both the tremendous opportunities and significant challenges ahead. The future belongs not to ever-more-specialized tools, but to systems fluid enough to adapt seamlessly and intuitively to the full spectrum of human intent and purpose.