Agency within constraint is the intellectually honest position for navigating technological transitions: acknowledging that technology forecloses certain futures (the pre-AI world cannot be restored) while insisting that institutional arrangements determine which specific future, within the constrained range, actually occurs. The position rejects both hard determinism (outcomes are technologically predetermined) and pure voluntarism (technology is neutral clay shaped entirely by human will). Instead, it maps the empirically grounded middle: technology constrains, institutions determine. For AI, this means recognizing that large language models' capabilities limit what is possible (no governance framework can make them incapable of generating fluent text) while insisting that deployment terms, benefit distribution, worker protections, and cognitive-capacity preservation depend on institutional choices being made now.
The position's intellectual foundations lie in Smith's soft determinism, Kuhn's recognition that paradigms constrain without fully determining scientific development, and Sen's capability approach distinguishing formal freedoms from substantive freedoms. Each framework holds constraint and choice simultaneously: paradigms limit what scientists can think while leaving genuine theoretical alternatives; capabilities expand formal options while requiring institutional conditions for their substantive exercise; technologies impose real limits while leaving outcomes genuinely open to institutional determination. The synthesis acknowledges constraint's reality without surrendering to determinism's passivity.
For builders navigating the AI transition, agency within constraint means studying the technology's genuine limits—what it can and cannot do, what it makes easy and what it resists—while focusing effort on the institutional dimensions where choice remains open. Segal's decision to maintain rather than reduce headcount after demonstrating twenty-fold productivity gains exemplifies the position: he acknowledged AI's capability (the constraint) while choosing to invest gains in expanded ambition rather than margin improvement (the agency). The choice was constrained—competitive pressure, investor expectations, quarterly metrics all pushed toward headcount reduction—but genuinely open within constraints.
The position's practical value lies in its capacity to sustain engagement under conditions where both optimism and pessimism produce paralysis. The optimist who believes technology automatically produces progress requires no institutional effort—the arc bends on its own. The pessimist who believes outcomes are predetermined by technological or economic forces sees institutional effort as futile—the river cannot be dammed. Agency within constraint holds both truths: the river is powerful (acknowledging constraint), and the dam determines where the water goes (exercising agency). The holding is uncomfortable but empirically grounded—it is the position the historical evidence of Springfield, Harpers Ferry, the Factory Acts, and every successfully navigated transition supports.
The framework emerged from Smith's confrontation with the empirical puzzle his armory research presented: if technology determined outcomes, why did Springfield and Harpers Ferry diverge? If human will alone determined outcomes, why did both eventually adopt precision manufacturing despite Harpers Ferry's decade-long resistance? The answer—that technology constrained the range while institutions determined the specific path—resolved the puzzle and became the methodological foundation for analyzing technological transitions as sites of genuine choice within real constraint.
Constraint is real, determination is false. AI capabilities impose genuine limits (the pre-AI world is foreclosed) without determining which post-AI world materializes—the range within constraints remains open to institutional choice.
The position enables sustained engagement. Hard determinism produces passivity (outcomes are inevitable); pure voluntarism produces denial (constraints are imaginary); agency within constraint produces the engagement historical evidence shows to be decisive.
Institutional quality is the decisive variable. Technology provides capability and imposes constraint; institutions convert capability into outcomes and determine who bears constraint's costs—the quality of institutions matters more than the power of technology.
The historical evidence is consistent. Every successfully navigated transition exhibits the pattern: technology constrained, institutions determined, and institutional quality separated equitable from exploitative outcomes.
The position is empirical, not philosophical. Agency within constraint is not a diplomatic compromise but the finding that comparative institutional analysis, applied across two centuries of documented transitions, supports—the honest reading of the evidence.