Building Cities: Why Trust is the New Infrastructure Backbone
By Nikhil Raghavan · Reporting from San Francisco ·
Paragrin argues that true urban efficiency requires inverting the data model to prioritize local trust and safety over raw data collection.
The latest episode of "Sequoia Capital," featuring Peregrine’s Nick Noone and Ben Rudolph, highlighted a tension that has defined technology for decades: how to build scalable infrastructure that promises efficiency without creating centralized points of failure—or worse, centralized points of control.
On the podcast, Paragrin argued that building truly "awesome" cities requires an infrastructure backbone focused fundamentally on safety and security, while simultaneously preserving individual privacy. Their approach centers on delivering solutions that are apolitical and governed by a deep respect for local context. Nick Noone described their model as inverting the traditional data business: instead of maximizing raw data collection or creating network effects from volume, Paragrin focuses on "joining disparate information" to improve precision and accuracy within existing governmental systems. Ben Rudolph underscored this problem space with his global experience, noting that many critical issues are simply "downstream of data problems," particularly in underresourced communities where organizations like UNHCR struggle with data scattered across difficult-to-manage spreadsheets.
The Engineering of Institutional Trust The most striking element of their thesis is the operationalization of trust itself as a core product feature. Noone detailed that accessing public safety institutions required them to lead with humility—asking, "We don't know that much... Can we come in, ask you some questions?" This emphasis on time-intensive, low-ego engagement contrasts sharply with the typical Silicon Valley playbook, which often assumes technological superiority can bypass institutional inertia. They view their Forward Deployed Engineers (FDEs) not as cost centers, but as essential "R&D and growth" engines that must co-own the problem with the customer.
The Governance Dilemma of AI The core technical challenge they face—and what makes their platform differentiation so significant—is defining data ownership in an age of powerful AI. Paragrin's entire architecture is built on the principle that each customer/institution owns its own data. They are not building a centralized data lake; they are providing governance capabilities and permission controls to allow secure internal use and selective sharing. This focus means their business incentive is structurally opposed to the historical model of maximizing data extraction for external profit.
However, this commitment raises profound questions about scale versus sovereignty. The strongest opposing case against this "governance-first" approach is that true systemic improvement—especially in public safety—is inherently centralized and requires massive, aggregated datasets (like those built by competitors like Axon). Critics would argue that any system designed to operate at the speed of modern AI must centralize data streams for effective training and pattern recognition. The argument that decentralized, permissioned solutions can achieve the same scale as a single, powerful corporate repository is technically ambitious, perhaps even utopian.
From Data Collection to Contextual Intelligence The historical pattern in civic technology has been predictable: a new technological capability emerges (e.g., body cams, facial recognition), which creates an immediate pressure for data collection. The market then develops a business model that monetizes the volume of that raw data. Paragrin attempts to break this cycle by making the solution dependent on the quality and governance of pre-existing, siloed data. They are building what amounts to an "institutional memory layer" for cities—a governance wrapper, rather than a data vacuum cleaner.
The critical difference between their model and the historical norm is that they focus on enabling law enforcement agencies to achieve their mission using existing resources (like integrating weather patterns with 911 calls to predict rip currents) rather than simply adding more sensors or cameras. They are selling better process, not just better hardware.
Ultimately, while the promise of decentralized AI governance sounds ideal, the reality is that public institutions operate on political and budgetary timelines vastly different from technological ones. The most durable technology platform cannot be purely governed by principles; it must first survive the politics of procurement and departmental budget cycles.