United States foreign policymaking under the second Trump administration is frequently described as erratic and incoherent. We’ve seen the launch of trade wars and actual wars, all without consulting allies first. This administration advances foreign policy through trial actions that are adjusted or abandoned, depending on what happens. The administration has a political vision, but many decisions are not easily reconciled with it.Trump’s approach is not, however, idiosyncratic or perverse. It’s an extreme expression of a decades-long broader shift in governance around the world, including in Australia. Trump’s foreign policy is symptomatic of a turn towards governance by prototype, an approach that is traceable to Silicon Valley. But governments aren’t in charge of products or websites. The mistakes of government can cost, at best, huge amounts of taxpayer money, or at worst, human lives.‘Prototyping’ policyGovernance by prototype is a way of exercising power that relies less on comprehensive strategies and long-term plans, and more on rapid rollout and fast feedback. Prototyping, in this sense, means deploying partial and temporary interventions designed to generate information quickly about what “works”.These prototypes are not policies in the traditional sense. They are intentionally incomplete measures, such as experimental diplomatic signals, test communications, or what commentators sometimes call “minimally viable” policies, borrowing from business development lingo. They are developed just enough to be put into use so governments can see how people react. Many examples of prototyping in policy take the form of digital applications or measures introduced by bypassing full parliamentary or congressional votes. In Australia, the COVIDSafe app is one noteworthy example from the past decade. In the United States, Trump’s 2017 travel ban, and its rapid cancellation and replacement a short time afterwards, offer further illustrations of policy prototyping.Prototyping treats governing as an ongoing experiment rather than delivery on a considered, declared course of action. Interventions are launched as probes. Feedback is continuously gathered from voters, consumers, financial markets, allies and even adversaries.Silicon Valley comes to governmentThis turn toward governance by prototype did not arise by accident, and it’s not unique to Donald Trump. It has been shaped by business thinking from the technology sector, particularly the “lean start-up” approach associated with Eric Ries. This method tells organisations not to spend years designing a perfect product before release. Instead, they should launch something basic as quickly as possible. They can then observe how people use it, learn from that and make repeated adjustments. What might once have been seen as failure is re-framed as valuable feedback.This way of thinking moved from Silicon Valley into government, facilitated by decades of contracting out, public-private partnerships, and increasing reliance on consultants. These are all trends that have been happening since the austerity politics of the 1980s. As governments outsource work and expertise, they also absorb private-sector assumptions, including prioritising speed over deliberation, closing deals over community-building and market responsiveness over long-term commitment. Shifting the burdenMethods devised to build digital products more cheaply are informing the exercise of public power, with profound and concerning implications.Australia’s Robodebt scheme illustrates the dangers of this style of governing. Robodebt was rolled out as an experimental system reliant on automated data matching and income averaging to generate welfare debts, on the apparent assumption that problems could be identified and fixed later. Read more: NACC investigation into Robodebt reveals public service corruption, but it will take much more to fix the system As many Australian researchers have shown, the burden of testing the system was effectively shifted onto welfare recipients, who were expected to disprove debts generated by flawed methods. Seen through the lens of governance by prototype, Robodebt sounds a warning about what happens when governments treat core public functions as beta tests. Legality should be continuously defended, not merely managed after the fact.A less consistent worldThe political importance of this shift lies in how responsibility and authority are redistributed. Governance by prototype is presented as practical, flexible and responsive. Yet it can also re-centre power in opaque systems and processes, making decisions difficult to challenge. Traditional, comparatively slow spaces for democratic debate, such as the United States Congress or Australian parliament, may more easily be bypassed.This matters internationally as well as domestically. Respectful alliances depend on reliability, shared expectations and planning together over time. A foreign policy built around probes and rapid pivots corrodes those foundations, even when a move “works” in the narrow sense of generating a desired short-term change. The US’s illegal seizure of Venezuela’s President Nicolas Maduro is a case in point.The issue, then, is not only that US foreign policy under Trump is unpredictable. It is that familiar ways of criticising governmental decisions have lost some of their force. Governing this way means policy is framed as experimental and reversible. Failures aren’t mistakes, but information. Responsibility is spread across platforms, partners and processes. All of this makes appeals to consistency, legality and stability far less effective.Recognising governance by prototype is essential to understanding why some of the most consequential decisions of our time take the forms they do. If democratic electorates do not insist that policy experimentation remains answerable to law, parliamentary oversight, and people’s long-term interests, “prototyping” could effect structural change in how we are governed nationally and globally without our even noticing.Fleur Johns has received funding for related research from the Australian government through an Australian Research Council Future Fellowship and the Australian Research Council's Discovery Projects funding scheme. However, the views expressed herein are those of the author and are not those of the Australian Government or the Australian Research Council.