Summary

  • David Ulevitch’s observable record from 2018 to 2026 shifts from operating network-security businesses towards allocating capital to defence, public safety and other regulated infrastructure. Prepared and the American Dynamism practice make that change visible, but they do not by themselves demonstrate procurement success or better public outcomes.
  • Prepared is a useful test because emergency-communications software sits inside a system owned and governed by public institutions. Investor conviction can finance a company; it cannot substitute for contracting, interoperability, cybersecurity, training, continuing service or the lawful decisions of 911 authorities.
  • Ulevitch’s defence writing favours software-first systems, open interfaces, faster testing and incentives for public buyers to consider startups. Those are authored proposals, not evidence that government has accepted them, and the institutional safeguards that slow adoption can also protect continuity and accountability.
  • The strongest reading of Ulevitch’s second act is therefore neither celebration nor dismissal. It is a capital-allocation thesis under an unusually demanding burden of proof: founders execute, agencies procure, elected and accountable institutions legitimate use, and public outcomes must be measured separately from financing.

The new operating surface

After Cisco, Ulevitch moved into an investment landscape where a cheque is only the beginning of the causal chain. The current a16z account identifies him as a general partner leading its American Dynamism practice, with a remit spanning national security, aerospace and public safety as well as enterprise software.

The same page records a disclosed set of investments, but it is an institutional biography and portfolio presentation, not an external audit of their performance (a16z author profile). What it establishes is role and direction: a security operator became a capital allocator whose public work increasingly concerns systems through which governments exercise responsibility.

That is a material change in operating surface. In conventional business software, a buyer can often decide that a tool is useful, integrate it and reverse course if it disappoints. Emergency communications and defence are different. The buyer acts under public law, spends public money, manages sensitive information and must maintain service when failure is most consequential.

A 911 authority cannot evaluate a new data layer only as a customer-experience improvement; it has to consider how that layer behaves with existing call-taking systems, contracts, security procedures, training and neighbouring jurisdictions. The federal procurement guidance for Next Generation 911, or NG911, accordingly places contract negotiation, scope of work, service arrangements and critical security procedures inside the transition itself (National 911 Program procurement guidance).

This is the public-safety capital test. The attractive technical mechanism is necessary but radically limited public evidence. Capital must be patient enough for procurement, disciplined enough not to mistake a funding round for adoption, and institutionally modest enough to recognise that the investor does not own the public decision.

The company must earn the right to operate within a continuity system; the authority must decide whether, how and on what terms to use it; the public outcome belongs to neither investor storytelling nor startup momentum until evidence shows it.

Prepared makes that test concrete. Ulevitch’s 2023 investment note argues that emergency call centres need better software and describes Prepared as a way to bring smartphone-derived video, images, text, location, translation and transcription into the work of 911 operators. The note is explicit advocacy by the investor who announced the investment, so it is strong evidence of his thesis and weak evidence of the results he anticipated (Ulevitch’s Prepared investment announcement).

Prepared’s own announcement says Andreessen Horowitz led its $16 million Series A and describes how the company intended to use the financing; that is the company’s account of funding and product, not independent validation of reach or effect (Prepared’s Series A announcement).

The distinction is more than a citation precaution. It is the core analytical fact. Ulevitch can be credited for publishing an investment rationale and for the disclosed decision to invest. Prepared’s founders, management and employees are responsible for building and operating the product.

Public-safety agencies are responsible for procurement and authorised use. Dispatchers and responders work within the resulting system. Any claim about faster or safer emergency response would require evidence about those institutions and outcomes, not merely evidence that capital arrived.

A transition, with a firm boundary

The earlier history matters only because it identifies what Ulevitch brought to the new role. Cisco announced in 2015 that it had completed the OpenDNS acquisition and that Ulevitch would become a Cisco vice-president; the announcement belongs to Cisco and presents the transaction in the company’s own strategic terms (Cisco acquisition announcement).

Cisco’s author archive now labels him “No Longer with Cisco” and shows a final listed post in March 2018, providing a reasonable boundary for the period examined here (Cisco author archive).

The relevant continuity is not another account of how resolver services were built. It is the movement from responsibility for a security organisation to responsibility for selecting and supporting companies. Operators control a product roadmap and an organisation; investors choose where to place capital and influence, but company teams still execute.

The difference becomes sharper when public institutions are customers, because a third centre of authority enters the picture. Government buyers set requirements and accept or reject systems. The investor is neither operator nor sovereign.

In a 2026 interview, Ulevitch gave his own account of this shift. He told Semafor that work with federal customers at Cisco shaped his interest in government technology; he also described investing in defence technology, recruiting Katherine Boyle and launching American Dynamism as a practice in 2022. Those statements are useful evidence of how he narrates his transition, while remaining self-reported claims inside an independently conducted interview (Semafor interview and reporting).

Semafor’s reporters supplied the more sceptical frame: venture capital had been optimised for capital-light software, whereas defence and industrial technology demand heavy capital, physical execution and a different kind of accountability. That question, rather than the old company story, defines the second act.

Prepared and what an investment can prove

Prepared offers an unusually clean view of both the promise and the attribution trap. The technical idea is intelligible: a modern phone can produce information beyond a voice call, while a public-safety answering point needs to turn incoming information into decisions. Ulevitch’s investment note presents an “advanced data layer” connecting caller media and location to operators and potentially to field responders.

Its language treats information and context as valuable in urgent calls and places software inside a critical public function (a16z’s authored investment rationale). This is the investor’s reason for acting.

Prepared’s own announcement adds the company-side account. It describes a free core product for receiving live video, photos and text, alongside paid features for field sharing and translation, and says the Series A would support hiring across engineering, marketing, sales and product (Prepared’s company announcement).

These details establish what management said it had built and planned to do. The announcement also contains broad claims about reach and benefit. Because the company is raising capital and promoting its product, those claims should remain attached to Prepared; they cannot be converted into verified emergency-service outcomes by removing the words “Prepared says.”

Independent reporting supplies a different, though still limited, layer. In 2024, TechCrunch reported that Prepared had raised a $27 million Series B led by Andreessen Horowitz and described a product through which dispatchers could receive location, text, images and supported video.

It attributed the ambition to reduce call burden and speed response to co-founder and chief executive Michael Chime, and it separately noted risks in AI-generated summaries, translation and speech transcription (TechCrunch’s Prepared report).

That is more independent than a company announcement, but the operational assertions in a financing story can still derive from executives. TechCrunch confirms the financing event and product category; it does not turn management’s desired effects into a controlled assessment of 911 performance.

The mechanism capital is buying

The mechanism is best understood as a reduction in information friction. A caller may have location, images or video that a voice-only workflow cannot readily absorb. A telecommunicator must interpret what arrives, preserve attention and pass usable information onward.

Software may organise, translate or summarise inputs. In that limited sense, security automation is not a recycled claim about network speed. It is the attempt to make emergency information more legible at the point where a public employee must decide what to do.

But information abundance is not automatically situational awareness. More inputs can create cognitive load, inconsistent records, new security exposure and new requirements for storage or access. GAO’s 2024 review of federal 911 centres identified funding priorities, interoperability, cybersecurity and increased data-management responsibilities among the challenges agencies reported during NG911 work (GAO’s NG911 assessment).

The report does not assess Prepared. It does show why adding data is an institutional task rather than a feature demonstration: every new channel has operational consequences outside the startup’s interface.

The investable proposition and the public proposition therefore differ. Investors may ask whether a company can build a valuable product, attract customers and grow. A public authority must ask whether the system fits its legal powers, procurement rules, service obligations, security model and local operating practices.

Those questions overlap, because a product that cannot satisfy the authority will struggle commercially. They are not identical, because commercial success does not settle whether a particular public use is legitimate, equitable or effective.

Four ledgers, not one success story

The cleanest way to read the record is to keep four ledgers apart. The first is capital: who financed the company and on what disclosed occasion. The Prepared and TechCrunch reports establish successive a16z-led rounds, without proving a return. The second is execution: what the company’s team built, sold and supported. Company and press descriptions establish the product category, while outcome claims require stronger evidence.

The third is procurement: what a 911 authority contracted for, under which terms and safeguards. None of the investment announcements can answer that for the public sector as a whole. The fourth is public effect: what changed for callers, telecommunicators, responders or system continuity. The frozen public record considered here does not establish a general causal result.

This separation protects both criticism and credit from exaggeration. It prevents a sceptic from treating all investor involvement as operational failure simply because implementation is difficult. It also prevents an advocate from treating financing, a product release or a customer claim as proof that emergency services improved.

The honest conclusion is narrower and more useful: Ulevitch selected Prepared as an expression of a public-safety software thesis; Prepared’s team pursued that thesis in product and market; public institutions retained the authority and responsibility to decide on use.

The public buyer changes the causal chain

American Dynamism’s institutional language makes government central. The practice says it invests in companies supporting the national interest across defence, public safety, manufacturing, education and other fields, and that government may be customer, competitor or stakeholder (a16z’s American Dynamism page).

This is a16z’s programme framing, designed to describe and promote a strategy. It establishes the categories the firm wants associated with the practice. It does not independently establish that the portfolio serves the national interest or that any agency should buy from it.

Yet the phrasing reveals the strategic difficulty. If government is a stakeholder, public authority affects a company even when it is not the immediate purchaser. If government is the customer, the route to revenue passes through appropriations, procurement and contract management.

If a company’s tool participates in an emergency workflow, the public body cannot outsource responsibility merely by buying software. The state is not a passive market segment. It is a rule-maker, a buyer and a bearer of continuity obligations at once.

Procurement is part of the system

The National 911 Program’s guidance is revealing because it does not describe procurement as paperwork surrounding the real technology. It says PSAPs must understand contract language and identifies negotiation, scope of work, continuing service and critical security procedures as areas requiring attention.

Its stated goal includes protecting the interests of the 911 entity and local community during procurement (911.gov procurement guidance). Contract design is thus one of the mechanisms through which public responsibility becomes operational.

For an investor trained on product velocity, this can look like friction. Sometimes it is. A vague requirement, fragmented authority or uncertain budget can delay a useful tool. But the same process allocates risk and specifies what happens after the demonstration.

Who maintains service? What is included in the scope? How are security procedures managed? What happens when an interface changes? A public system that fails to answer those questions has not removed friction; it has deferred it into an emergency.

This is why an investment thesis cannot be evaluated solely by how accurately it identifies outdated technology. The stronger test is whether the financed organisation can operate through procurement without treating public safeguards as an obstacle to route around.

Institutional competence includes translating product capability into terms a public buyer can lawfully evaluate, supervise and sustain. The buyer, for its part, needs enough technical competence to distinguish a serious service from a persuasive demo. Capital can support both processes indirectly by funding staff and product work, but it cannot make the public decision legitimate on its own.

Transition is uneven by design and constraint

The national context complicates any simple story of a single “911 market.” CRS reported in 2024 that state and local governments exercise authority over 911 funding and operations, producing different levels of NG911 progress.

It explained that upgrades are usually planned and managed by the local government entity operating the PSAP and often sit within broader state strategies. It also warned that state progress does not mean each local centre has completed its own transition (Congressional Research Service report).

That structure has commercial consequences, but it should not be reduced to sales fragmentation. Local authority reflects the fact that emergency communications are embedded in specific institutions and communities. A system may need to interoperate regionally while remaining accountable locally. The difficulty is to gain the benefits of common technical capability without pretending that every jurisdiction has the same budget, infrastructure, procedures or tolerance for risk.

GAO found a similarly uneven picture among federal agencies. As of its June 2024 reference point, seven of eleven agencies operating 911 centres had begun planning or implementation, but GAO characterised progress as limited; four had taken no upgrade steps, in a context without a federal implementation requirement.

Officials cited funding priorities, interoperability, cybersecurity and data management, while views differed on whether one federal agency should coordinate the effort (GAO report). These are findings about government institutions, not an evaluation of a16z or Prepared. Their value here is to discipline causality: nationwide need does not produce uniform procurement, and technical availability does not erase authority, resources or coordination.

Cost is not a fixed market-size number

Funding evidence also resists a smooth venture narrative. CRS explained that 911 systems are financed primarily through state and local surcharge fees and general funds, that fee structures differ, and that many jurisdictions report shortfalls. Its 2024 report cited the earlier federal estimate of $9.5 billion to $12.

7 billion for ten years of nationwide deployment and transition, while noting broader lifecycle costs and uncertainty over federal support (CRS funding analysis). Those figures describe a public transition burden; they are not a revenue forecast for any company.

In April 2026, NTIA published a new estimate of $5.8 billion to $9.27 billion in remaining nationwide implementation costs. It attributed the decline from the 2018 range to state investment already made and a market shift from equipment-heavy systems towards software subscription models.

NTIA also said the study was for planning, coordination and oversight, did not prescribe policy, did not assume a federal grant programme and excluded the administrative costs of one (NTIA’s 2026 NG911 cost study).

The revision matters in two directions. It supports the proposition that software delivery models can change the cost profile of infrastructure. It also demonstrates why a headline cost should not be treated as a stable total addressable market.

Some spending has already occurred; subscription pricing moves costs across time; implementation and administration are not the same; and policy choices determine who pays. An investor may see a large opportunity. A public authority sees a multiyear obligation with recurring service, governance and integration needs. Both can be true without being the same claim.

The shift towards subscription delivery also redistributes risk rather than making it disappear. Lower initial equipment spending may help an authority begin a transition, while recurring charges make service quality, renewal terms and long-run budget capacity more important.

CRS presents Congress with choices between a large one-time appropriation and annual support, noting that each would shape planning differently: a single commitment can offer assurance, whereas annual appropriations can adjust over time but may fluctuate (CRS funding analysis).

For a venture-backed supplier, recurring revenue may look commercially attractive. For a PSAP, the same arrangement is a continuing public obligation whose affordability and exit conditions matter. The contract has to reconcile those perspectives.

This is another reason not to equate a large public need with an unconstrained market. Need may be urgent while purchasing authority remains distributed; an agency may value a capability while lacking a durable funding path; a lower national cost estimate may reflect work already completed rather than a new pool of demand.

Capital can help a supplier absorb development and selling costs before public revenue arrives. It cannot promise appropriations, convert a policy option into a budget or transfer the authority’s continuity risk back to the investor. The public-safety capital test is passed only when the commercial arrangement remains supportable on the buyer’s timescale, not merely financeable on the company’s.

Security automation at the emergency boundary

“Security automation” becomes meaningful here only when tied to the emergency-communications task. Prepared’s reported tools concern the intake and organisation of caller information: location, messages, images, supported video, transcripts, summaries and translation.

TechCrunch described those capabilities while carefully assigning performance claims to Chime and noting that AI systems can introduce errors and uneven transcription quality (TechCrunch report). The relevant question is not whether automation sounds modern. It is which burden the system removes, which new burden it creates and where human judgment remains.

Assistance is not authority

In an emergency centre, a summary can help a telecommunicator find salient details; it can also omit or distort them. Translation can widen access; an error can alter meaning. Location and video can supply context; they also create data that must be received, displayed, protected and governed.

The responsible design posture is therefore assistive: automation organises evidence for authorised professionals, while the public institution defines how it may be used and remains accountable for the service.

That distinction follows from the available evidence rather than from a claim that one particular product has achieved it. Prepared describes its software as a data layer and TechCrunch reports optional AI features. Government assessments identify cybersecurity, interoperability and data management as transition challenges. Together, those sources establish the operating tension. They do not establish a specific error rate, response-time gain or regulatory conclusion, so none should be inferred.

Automation also changes procurement questions. A conventional interface can be assessed as a relatively stable tool. A system producing machine-generated transcripts, summaries or translations raises questions about updates, monitoring and how staff recognise failure.

Ulevitch and Porter Smith’s defence article makes a related argument in another context: they advocate automating testing so software-first systems can move faster, but expressly condition the benefit on maintaining quality-control thresholds (a16z Pentagon essay). The condition is the important part. Faster certification without maintained quality would not satisfy the thesis on its own terms.

Continuity is a property of institutions

Public-sector continuity is sometimes described as a technical matter of redundancy. NG911 sources show a wider meaning. CRS describes IP networks, core routing services, call-taking equipment and geographic information systems as interconnected components; it also discusses the ability to transfer calls and data across centres and the prospective interconnection with FirstNet (CRS NG911 overview).

GAO adds the organisational dependencies: funding, agency priority, interoperability, cybersecurity and data responsibility (GAO assessment).

Continuity therefore lives across technical and institutional layers. A resilient network without trained staff is incomplete. A capable application without a supportable contract is incomplete. A local deployment that cannot exchange information with its surrounding system may be incomplete. A financing round has no direct bearing on those properties unless the company and authority translate resources into sustained operational capacity.

This widens the standard by which public-safety capital should be judged. Capital is useful when it lets a team build, test, support and improve a service over the timescale public buyers require. It is harmful to analysis when its arrival is used as a proxy for readiness.

The investor’s role is upstream: selecting a thesis, accepting commercial risk and supporting an organisation. Continuity is produced downstream by the company’s execution and the public institution’s governance, with dependencies neither side controls alone.

The defence thesis—and its transfer limits

Prepared is not a defence programme, and 911 procurement should not be collapsed into Pentagon acquisition. The comparison is useful only at the level of institutional architecture. In both settings, Ulevitch’s public record focuses on software entering legacy, regulated and high-consequence systems.

In both, the buyer must reconcile faster technical iteration with certification, interoperability, budgets and responsibility. The differences in mission, law and operational consequence remain decisive.

What Ulevitch and Smith actually argue

In their 2023 essay, Ulevitch and Smith propose a Pentagon organised more like a distributed computing system. They argue for more interchangeable hardware, interfaces owned by the Department of Defense, translation between heterogeneous data formats, automated testing, embedded programmability, resilient networks and incentives for procurement officers to consider startup risk.

They also contend that software updates move on a different cadence from multiyear hardware budgeting and certification (their a16z defence essay).

This is a coherent authored thesis, and it clarifies Ulevitch’s view of infrastructure investment. The unit of innovation is not merely an application. It is the interface between software, hardware, data and the institution that buys and certifies them.

Open interfaces are meant to reduce lock-in; translation layers are meant to preserve heterogeneity; automated testing is meant to reconcile faster iteration with quality; changed incentives are meant to alter which suppliers can reach programmes.

But the essay is an investor-policy argument published by a16z. It does not show that the Pentagon accepted the proposals, that a startup won a procurement because of them or that a public outcome followed. Even its own disclosure notes that information may come from third parties and is not guaranteed. The appropriate credit is precise: Ulevitch co-authored a programme for procurement reform. Government decisions remain the government’s.

Speed and safeguards are not simple opposites

The essay’s most productive tension is between software cadence and institutional assurance. Ulevitch and Smith want testing and certification to keep pace with software while maintaining quality thresholds. That is more serious than simply demanding deregulation, because it recognises a public buyer’s need for assurance.

Yet the practical difficulty lies in the condition. Automating a test is valuable only if the test captures the risk that matters; opening an interface is useful only if security and accountability survive; increasing procurement risk appetite is defensible only if officials can explain and supervise the decision.

The 911 evidence reinforces that point. The National 911 Program treats security procedures and service terms as procurement essentials. GAO records cybersecurity and interoperability as implementation challenges. CRS shows that responsibility is distributed across state, local and federal actors. These are not examples of government endorsing Ulevitch’s defence thesis. They are independent reminders that institutional delay can contain both avoidable inertia and necessary care.

The investor’s difficult job is to distinguish the two without claiming the authority to resolve the public choice. Capital can back companies that make compliance, testing and integration easier. It can support longer sales cycles and the unglamorous work of reliability. It can also press for better interfaces and clearer purchasing paths. What it cannot legitimately do is define every safeguard as incumbent protection merely because a portfolio company would benefit from speed.

The Semafor challenge

Semafor framed American Dynamism as a test of whether venture skills developed around light, scalable software transfer to capital-intensive, strategically complicated businesses—or whether ideological enthusiasm is being mistaken for competence. Ulevitch answered that automation, software-enabled manufacturing, large potential outcomes and an acceptance of venture losses can make selected industrial businesses suitable for venture capital.

These are his claims about the asset class, presented within the reporters’ sceptical frame (Semafor interview).

For public safety, the answer cannot rest on industrial scale alone. A company may be venture-backable because the upside appears large, yet still face a public market composed of bounded budgets and accountable jurisdictions.

The very thing that makes venture work—concentrating on a small number of outsized winners—does not map neatly onto a public obligation to provide dependable service everywhere. Investors can tolerate a portfolio loss ratio. A 911 authority cannot treat failed local continuity as the routine cost of finding a winner.

That mismatch does not disqualify venture capital. It defines the discipline required of it. The financing model can absorb technical and commercial risk at the company level. The public institution must contain operational risk at the service level. A credible public-infrastructure company respects both equations rather than asking the customer to adopt the investor’s.

Institutional legitimacy is an operating requirement

Ulevitch’s 2026 Semafor interview moves from economics to authority. Discussing defence AI, he rejected a model in which a technology chief personally approves each government use and invoked elected leaders, law, judicial review and congressional oversight. Semafor’s headline distilled his criticism as a warning against a tech-chief “God complex.”

These are Ulevitch’s views in an interview, not external proof that oversight is sufficient in any particular programme (Semafor’s report and transcript).

The instinct to reject private sovereign power is important. A vendor should not secretly become the chain of command. Public decisions about authorised use must remain with institutions answerable to law and politics.

But institutional legitimacy is more demanding than saying “the government decides.” Which level of government? Under what authority? Through which procurement? With what contract, audit, retention rule, complaint path and remedy? Democratic delegation is an architecture, not an incantation.

Legitimacy must survive implementation

Emergency communications makes this concrete without requiring the defence controversy to be imported wholesale. A PSAP is already a public institution with defined responsibilities. When it adds new media or automated interpretation, legitimacy depends on the authority’s decision and on the operating controls that make the decision real.

Procurement guidance’s attention to scope, service and security is one part of that architecture (911.gov guidance). State and local control over funding and operations is another, as CRS documents (CRS report).

This does not mean every local decision is wise or every process adequate. It means the route to correction remains public. A vendor can supply evidence and design controls; an investor can argue for reform; neither acquires the authority of the buyer. When Ulevitch says elected institutions should decide, his investment practice inherits the obligation to build for institutions capable of making and revising those decisions—not merely to invoke them after controversy.

The distinction also sharpens responsibility for automation. If an agency authorises an assistive summary, the company remains responsible for building and supporting the feature according to contract, while the agency remains responsible for its authorised workflow.

If evidence later shows a problem, accountability should not dissolve into a claim that the algorithm acted or that the customer chose. Legitimacy requires traceable roles. The available sources do not document a particular Prepared failure or success; they show why those roles must be kept visible before either is claimed.

Public acceptance cannot be inferred from procurement

Even a completed purchase would not close the legitimacy question. Procurement proves that an authorised body selected a service under some process. It does not automatically prove broad public acceptance, equitable operation or improved outcomes. Conversely, controversy does not by itself prove that a system lacks public value. Those are empirical and political questions for the responsible institutions and communities.

Ulevitch told Semafor that jurisdictions and voters would make different trade-offs about public-safety technology. The interview concerned surveillance tools rather than Prepared, so it should not be used as evidence about emergency communications.

It nevertheless exposes a general feature of his position: he places legitimacy in public choice rather than unilateral vendor control (Semafor interview). The analytical response is to take that claim seriously enough to demand the institutions behind it. Voters cannot assess what contracts obscure; officials cannot supervise what systems do not log; agencies cannot sustain what budgets do not support.

Institutional legitimacy is therefore not a brake applied after innovation. It is productive infrastructure. Clear authority reduces ambiguity. Defined scope makes performance assessable. Security procedures protect continuity. Auditability makes correction possible. Stable funding supports maintenance. These qualities may lengthen the path from investment to deployment, but they also make a public system durable enough to deserve trust.

Reading the record from 2018 to 2026

Across the period, the observable change is substantial but bounded. In 2018, Cisco’s archive still records Ulevitch publishing as a security executive and now marks him as having left the company (Cisco author archive). By 2023, he was publicly co-authoring a defence-procurement thesis and an investment announcement for an emergency-communications company (Pentagon essay; Prepared announcement).

By 2026, a16z presented him as the leader of American Dynamism, while Semafor examined his case for venture capital in defence, public safety and industrial systems (a16z profile; Semafor interview).

That sequence supports the thesis of a move from network-security operator to infrastructure investor. It does not establish that every element was planned from the outset, that the investments succeeded, or that public institutions adopted his reform agenda. The public record is a record of role, authored argument and disclosed allocation—not private motive or verified outcome.

Prepared is the most revealing case because it narrows the abstraction. The firm’s American Dynamism language can stretch across many national-interest categories. Emergency communications forces a specific test: can venture-backed software enter a public continuity system without confusing product velocity with authority? The company’s financing and product story shows that investors and founders are attempting it. Government sources show why the attempt remains institutionally contingent.

Three judgments follow.

First, Ulevitch’s operating background appears in his investment writing as a preference for software layers, data availability, automation and interfaces. That is an interpretation of his published record, not a claim that his prior company produced Prepared’s product. The credit for Prepared’s execution belongs to its team.

Second, American Dynamism converts government-facing technology from a side interest into an explicit investment category. Its own page and Ulevitch’s profile demonstrate that positioning, while their interested nature prevents them from validating the programme’s effects. Semafor independently confirms that the strategy is significant enough to attract scrutiny and supplies a test that a portfolio page will not: whether venture’s methods transfer to heavy, accountable systems.

Third, the public evidence makes procurement and legitimacy central rather than peripheral. The NG911 transition is uneven, costly and institutionally distributed. The 2026 NTIA estimate suggests progress and a changed delivery model, but it still describes billions of dollars in remaining implementation cost and expressly avoids prescribing policy (NTIA cost study).

GAO documents limited federal progress and persistent operational challenges; CRS documents the funding and authority structure. None endorses Ulevitch, a16z or Prepared. Together they define the environment in which any investor thesis must prove itself.

The fairest assessment is consequently demanding. Ulevitch deserves credit for making a legible bet: advanced software can serve emergency communications and other public infrastructure, and private capital can bear some early risk.

He also deserves to be read literally when he argues for procurement reform and public institutional authority. But those positions create obligations. The investment case must tolerate oversight, the portfolio company must do the execution, the public buyer must retain judgment, and claims of benefit must wait for evidence.

Conclusion

David Ulevitch’s second act is not a sequel to the resolver story. It is an experiment in whether an operator’s preference for software and automation can be converted into patient, institutionally literate capital. Prepared, the Pentagon essay and American Dynamism make the direction observable. The NG911 evidence makes the standard unavoidable.

The standard is not whether money reaches public-safety technology. It is whether capital supports companies that can earn procurement, operate within legitimate authority and strengthen continuity without appropriating credit for the work of founders, public servants or institutions. Between 2018 and 2026, Ulevitch clearly moved towards that test. The evidence establishes the wager. The public outcomes remain for responsible institutions—and future evidence—to establish.

Sources