From Knowledge-Constrained to Context-Constrained
Up until now, building software required knowing how to build software. The constraint was knowledge: understanding languages, frameworks, patterns, and the accumulated wisdom of decades of engineering practice. Those who possessed this knowledge created; those who didn't requested and waited.
That constraint has evaporated. The machines now possess the knowledge. They understand languages, frameworks, and patterns far more comprehensively than any individual ever could. What they lack is context: the specific circumstances, requirements, history, and intent that define what should be built. The cutting edge of software development has shifted from knowing how to build to knowing how to convey what should be built.
This shift does not diminish the importance of intent. Quite the opposite. When anyone can build anything, the question of what to build becomes paramount. Valuable software still requires valuable intent, a clear understanding of problems worth solving and solutions worth implementing. The constraint has moved, but constraint remains.
Context Collapse
The current moment demands significant effort to provide context. Codebases must be understood, conventions documented, architectures explained. Skilled practitioners bridge the gap between machine capability and specific need by constructing elaborate context (toolkits, prompts, workflows) that translate organizational reality into machine-comprehensible instruction.
This burden is temporary. Models grow more capable of inferring context from sparse signals. They learn to navigate unfamiliar codebases, infer conventions from examples, and understand organizational patterns from limited exposure. The context that today requires careful curation will soon be gathered automatically.
When context becomes trivial to provide, the economics of software change fundamentally. A user encountering friction in their workflow will be able to request modifications (new features, integrations, adaptations) without risking the stability of existing functionality. The traditional build-versus-buy decision, which weighs the cost of custom development against the compromises of existing solutions, collapses. Custom software becomes as accessible as configuration.
Ambient Context
Beyond explicit context lies ambient context: the vast ocean of signals that machines will eventually perceive without being told. Calendars, communications, behavioral patterns, organizational structures, historical decisions, and their outcomes. The sum total of digital existence, continuously observed and understood.
Machines operating with ambient context will understand the ramifications of intent better than the humans expressing it. They will recognize when a requested feature conflicts with existing workflows, when a solution creates downstream problems, when an apparent need masks a deeper issue. The gap between what people ask for and what they actually need, a gap that skilled practitioners have always navigated, will be closed by machines that understand both.
This trajectory leads somewhere uncomfortable for current notions of agency. When machines understand context more completely than humans can articulate it, when they can predict consequences more accurately than humans can imagine them, the question of who decides what software should exist becomes less clear. Decision-making authority naturally flows toward comprehension.
Software Dissolves
The logical endpoint of these trends is the dissolution of software as a distinct category of artifact. User interfaces, carefully designed to present information and gather input, become unnecessary when machines already possess the context that interfaces exist to collect. The painstaking work of UX design (understanding user needs, crafting intuitive flows, iterating toward clarity) becomes automated generation of whatever presentation layer a specific moment requires.
No interface designed by humans will be more efficient than a personal agent that already knows what a user needs and can act on their behalf. The app, the dashboard, the form: these are all mechanisms for bridging a gap between human intent and machine action. When that gap closes, the bridges become irrelevant.
Software creation, as currently understood, has no value proposition in this future. The skills, tools, and practices that define the profession today are oriented around a constraint (context) that will not persist. The work of translating human intent into machine action, which has employed millions and generated trillions in value, approaches obsolescence.
The Current Window
This trajectory explains why toolkits exist in this particular moment. The destination may be clear, but the path still requires navigation. Context remains the constraint, and providing it effectively remains difficult. The gap between machine capability and practical application is still wide enough that bridging it creates value.
There is meaningful work to be done while the window remains open. Systems can be built, problems can be solved, and value can be created, all within a paradigm that has an expiration date but has not yet expired. Understanding where things are headed does not require abandoning the present.
The window will close. Context will become ambient. Software will dissolve. But today is not that day, and there is still building to be done.