Introduction: When the App Store Becomes an App Factory
In a move that signals a fundamental shift in how mobile software comes into existence, Google Play has begun allowing users to build private AI-generated Android applications. The announcement, while sparse on technical detail, carries implications that ripple across the entire mobile development ecosystem, the artificial intelligence landscape, and the broader question of who gets to create software in the modern era.
This article examines the development from multiple analytical dimensions: the underlying technology route, commercialization logic, industry impact, competitive dynamics, ethical and security considerations, investment implications, and infrastructure requirements. Because the original announcement lacks critical details—no official documentation, no pricing information, no technical specifications, no publication timeline—much of what follows is necessarily inferential. We frame these conclusions as a structured hypothesis list rather than established fact, and we explicitly flag where confidence is limited.
The core thesis is this: Google's move represents a platform-level defensive positioning in the emerging "AI-native application development" arena. It is not a technological breakthrough at the foundational layer but rather a productized encapsulation of mature large language model code generation capabilities. Its strategic essence is to leverage vertical integration across operating system, app store, frontier models, and custom silicon to defend the gateway to the application economy. The most consequential detail may not be that apps can be generated at all—that capability has been repeatedly demonstrated—but rather the deliberate restriction to "private" apps, which functions simultaneously as a risk containment mechanism and an implicit admission about the maturity and safety of the underlying technology.
The Technology Route: Productized Encapsulation Rather Than Foundational Breakthrough
What Is Actually Under the Hood
The tool in question is best understood as an application-layer productization wrapper rather than a底层 (foundational) technological innovation. Its technical core is almost certainly the code generation capability of the Gemini model family combined with agentic scaffolding. This places it at the engineering or combinatorial innovation level, not the paradigm-shifting level.
The real technical question is not whether AI can generate code. That question has been answered repeatedly and affirmatively by tools such as GitHub Copilot, Cursor, v0, Lovable, and Bolt.new. The far more interesting and difficult question is how generated applications can be safely and stably deployed within the Android runtime environment. The original announcement does not address this at all.
Consider the architecture. AI-generated apps depend on the LLM's code synthesis ability. The current technical foundation for that ability remains the Transformer architecture combined with post-training alignment. There is no paradigm-level innovation here. It sits on the same technological trajectory as the aforementioned tools.
If the feature has been publicly launched, it should be classified as being in the production stage. However, the qualifier "private" strongly suggests a deliberate limitation of functional boundaries: no listing on the store, no commercialization, no public distribution. This is a classic risk containment design, and it indirectly reveals that its maturity has not yet reached a level where open listing is advisable.
The Missing Technical Details
The announcement does not mention which underlying model version is used. It does not specify the language or framework of the generated code—Kotlin with Compose? Flutter? Something else entirely? It does not describe the runtime environment—does the generated app run locally on the device, or does it execute in a cloud sandbox? It does not indicate whether the generated app can invoke Play Services or App Functions.
These are not minor omissions. They are the difference between a toy and a tool.
Hidden Information and Implied Constraints
Several critical insights can be inferred from what is not said.
First, "private" status is the result of technical compromise, not a purely user-value-driven proposition. Public listing would require passing Play review, assuming liability for malicious code, and handling commercial revenue sharing. All of these would amplify the uncontrollability of AI-generated code. Using "private" as an isolation mechanism essentially transfers quality and safety risks to the user.
Second, the boundary of generation capability is deliberately blurred. The difference between generating a to-do list app and generating a complex app with payments, maps, and push notifications is a matter of orders of magnitude. The vague phrase "build apps" conveniently sidesteps the question of the capability ceiling.
Third, it is highly likely that this feature is bound to the Gemini ecosystem. Such features typically force the use of Google's own model stack, mirroring Apple's approach with Apple Intelligence and OpenAI's use of its own models. This is simultaneously a technical choice and an ecosystem lock-in mechanism.
Unanswered Critical Questions
The complexity ceiling of generated apps remains unknown. Does it support third-party SDKs, backend services, or persistent storage? Does the app run locally or depend on the cloud? What is the offline availability? Can the generated code be exported, re-edited, or migrated, or is it a black-box product? Is there version management and iteration capability—can the AI modify code while maintaining consistency?
Confidence Assessment: Medium
The technical direction—LLM code generation—is judged with high confidence. However, almost all specific implementation details rely on industry analogy and inference. The announcement provides no verifiable technical information.
Commercialization Analysis: Platform-Level Strategic Defense
The Real Business Logic
This is a platform-level strategic defense, not a direct monetization move. "Private apps" bypass the core business model of the Play Store: the 15 to 30 percent commission on in-app purchases and subscriptions. Therefore, it generates almost no direct revenue.
The true commercial logic is threefold. First, it reinforces the Android developer moat with AI capabilities. Second, it acquires new users for Gemini. Third, it shortens the funnel from idea to prototype, thereby locking in dominance over the application economy over the long term.
Google Play primarily monetizes through app sales, in-app purchases, and subscription commissions. Apps that are private and unlisted cannot enter this revenue-sharing system. This indicates that the key performance indicator for this feature is not revenue but developer retention and ecosystem activity.
The platform competition logic is equally clear. AI app generation capability has become an operating-system-level standard feature. Apple Intelligence, Microsoft Copilot, and others have established this expectation. Google must follow suit or risk losing developer mindshare. This is defensive investment.
The funnel value is substantial. AI-generated private apps serve as an excellent developer conversion entry point. The path from casual experimentation to profitable listing is dramatically shortened, and over time this can convert into a paid developer supply for Play.
Competitor Comparison Matrix
When compared across multiple dimensions, Google's position reveals distinctive advantages and potential vulnerabilities.
| Comparison Dimension | Google Play (Subject) | Apple | OpenAI | Microsoft | Independent Platforms (Lovable/Bolt/v0) | |----------------------|----------------------|-------|--------|-----------|----------------------------------------| | Generation Target | Native Android app (private) | App Intents / Shortcuts | GPTs / Canvas | Power Apps + Copilot | Primarily web apps | | Underlying Model | Gemini (inferred) | Apple in-house + OpenAI | OpenAI in-house | OpenAI multi-model | Multi-model | | Distribution Channel | Play (private, unlisted) | App Store | No independent store | No independent store | Self-deployed | | Monetization Model | Ecosystem lock-in | Hardware + ecosystem | Subscription + API | Subscription + Azure | Subscription | | Core Moat | OS + store + model vertical integration | OS + chip + privacy | Model leadership + user scale | Enterprise channels | Iteration speed |
Hidden Information and Strategic Implications
The move represents a dimension-reduction attack on independent vibe coding platforms. When an operating system vendor makes AI app building a system-level free capability, independent tools such as Lovable and Bolt face platform engulfment risk.
Google's compute costs can be absorbed internally. Gemini runs on Google's own TPUs and cloud infrastructure. The marginal cost is far lower than independent platforms paying for third-party API access. This provides the confidence to wage price wars or offer free tiers.
There is also a subtle point about intellectual property. If Google provides the generation capability, the platform, and the distribution channel, does it claim partial IP rights over generated apps? The announcement does not say, and this silence is itself informative.
Unanswered Critical Questions
Is there a charge for this feature? What is the free quota? If a generated app is later listed for monetization, does Google set conversion incentives or bundling terms? Does Google assert partial IP rights over apps generated using this feature?
Confidence Assessment: Medium
The Play business model is public common knowledge, and the inferential chain is reasonable. However, the announcement provides no information on pricing, charging models, or strategic statements. The specific commercial design remains entirely unknown.
Industry Impact Analysis: Reshaping the Supply Side of App Development
The Displacement Boundary
If the feature matures, the impact will be concentrated on the low-to-mid-end supply side of app development: templated apps, outsourced small workshops, and junior developers. The impact on high-end complex app development will be limited.
AI generation is suitable for apps with single functions, no complex backend, and no strong compliance requirements. For complex applications involving payments, finance, healthcare, or real-time communication, current capabilities are far from sufficient to replace professional development.
The affected industries include no-code and low-code platforms, the app outsourcing market, template app developers, and some junior Android developer positions.
The employment impact gradient unfolds over time. In the short term, zero to six months, the primary effect is augmentation—developers become more efficient. In the medium term, six to eighteen months, displacement of templated and simple app supply begins. In the long term, eighteen to thirty-six months, if agent capabilities break through, medium-complexity development may be touched.
Industry Impact Assessment Table
| Affected Domain | Displacement Rate | Augmentation Rate | Time Window | Typical Scenarios | |-----------------|-------------------|-------------------|-------------|-------------------| | Personal utility apps | High (>60%) | — | Short term | Expense tracking, to-do lists, habit tracking | | Prototypes / proof of concept | High | High | Short term | Startup MVPs, demos | | Templated commercial apps | Medium | Medium | Medium term | Simple e-commerce, display pages | | Complex enterprise applications | Low (<20%) | Medium | Long term | ERP, finance, real-time systems | | Professional developer roles | Low | High | Medium-long term | Efficiency tools |
Hidden Information and Second-Order Effects
The narrative of "accelerating innovation" may in reality mean "accelerating supply surplus." When the cost of app generation approaches zero, app stores will face a garbage app and homogenization problem. This may actually exacerbate the scarcity of user attention.
The true beneficiary may be Google itself. The more developers there are, and the stronger their incentive to remain in the Play ecosystem, the more secure Google's platform rent becomes.
There is also a content governance cost dimension. If AI-generated private apps proliferate, and some of them contain problematic content, does Google bear any responsibility? The private designation may be designed to limit this exposure, but the reputational risk remains.
Unanswered Critical Questions
What is the quality floor of generated apps? Will it produce large numbers of apps that work but are fragile? Is the actual impact on independent developers and small teams displacement or increment? Will it increase content governance costs for the Play Store?
Confidence Assessment: Medium
The direction of industry impact is supported by industry analogies—the existing impact of low-code and Copilot. However, specific data for this case is lacking. The displacement rate figures are inferred values.
Competitive Landscape Analysis: Defensive Positioning in the AI-Native Development Race
Google's Structural Advantage
This is Google's defensive positioning in the AI-native development track. Its structural advantage lies in full-stack vertical integration: Android OS, Play Store, Gemini models, and TPU compute. This is a combination that independent tools and even Apple or OpenAI do not fully possess.
The exclusive resources are significant. Globally, only Google simultaneously owns a mobile operating system, an app store, frontier models, and custom chips. Apple lacks frontier models and cloud compute at scale. OpenAI and Microsoft lack a mobile OS and app store.
The defensive motivation is clear. Facing competition from the OpenAI ecosystem (GPTs, Operator) and vibe coding platforms for developer mindshare, Google needs to offer comparable capabilities on its own platform to prevent developer flight.
The Historical Weakness
Google's historical shortcoming is productization speed and experience polish. Early Bard and Gemini performance, the image generation incident, and product line confusion all demonstrate a "strong technology, weak product" inertia. Whether this feature will be genuinely usable remains the biggest variable.
Ecosystem Barrier Assessment
The developer scale advantage is obvious. Android's global developer base is enormous, and reach cost is low. The distribution closed loop is a unique advantage: from generation to Play listing is a natural pipeline. Model leadership is neutral: Gemini and GPT are in the first tier, but neither maintains stable dominance. Product experience is a potential disadvantage given historical volatility.
Hidden Information and the Real Battlefield
The real battlefield is not generating apps. It is generating what can be natively scheduled by the operating system. If Google combines generation capability with Android system-level agents such as App Functions and system-level intent invocation, it could form a deeper moat than generating a standalone app. The announcement does not touch this layer at all.
There is also an indirect countermeasure against OpenAI. OpenAI's GPTs and application ecosystem lack mobile distribution. Google can use OS-level capabilities to hedge.
Unanswered Critical Questions
What is the product line relationship with Google's own Firebase Studio and AI Studio? Is there functional overlap or positioning conflict? Does this feature integrate with Android system-level AI such as Gemini system integration? Does Apple already have an equivalent capability, and how large is the time gap?
Confidence Assessment: Medium
The macro judgment of the competitive landscape—Google's vertical integration advantage—has relatively high confidence. However, the announcement provides no competitor comparison or official positioning information. The specific competitive situation is inferred.
Ethical and Security Analysis: The Private App Paradox
The Central Contradiction
"Private apps" represent the biggest security paradox in this case. It lowers the development threshold while simultaneously bypassing the most critical security gate of the app store: listing review. The combination of AI generation, private operation, and no review theoretically opens a gray space for malicious code, data theft, and privacy violations.
This may be the core risk control consideration behind Google's decision to limit the feature to private status. It is also the point most likely to attract regulatory intervention.
Security Risk Assessment Table
| Risk Category | Risk Level | Assessment Basis | Mitigation Effectiveness | |---------------|------------|------------------|--------------------------| | Malicious code generation | Medium-high | AI can be induced to generate dangerous code; private apps have no review | Unknown (depends on sandbox/permission controls) | | Data leakage | Medium-high | Generated apps may request sensitive permissions; data flow unclear | Unknown | | Prompt injection | Medium | Generation process may be hijacked by malicious descriptions | Unknown | | Content compliance | Medium | How to intercept apps that generate violating content | Unknown | | Copyright / IP | Medium | Generated code may replicate others' app logic | Unknown | | Alignment quality | Medium | Gemini safety filtering capability determines the floor | Partially effective |
Core Reasoning
The absence of review is the central risk. Play Store review is the primary barrier protecting the Android ecosystem from malware. Private apps bypass this mechanism, transferring security responsibility to users and OS sandbox mechanisms.
The double-edged sword of generation capability is also significant. LLM code generation is highly sensitive to functional descriptions. Attackers can craft prompts to generate phishing or data theft apps.
Privacy sensitivity compounds the problem. If AI-generated apps access contacts, location, or photos, users often cannot assess their safety.
Hidden Information and Regulatory Implications
"Private" may be a temporary regulatory arbitrage scheme. By not entering public distribution, it avoids app store review obligations and some content compliance requirements. However, if the feature is abused, regulatory pressure will rise rapidly.
The reputational risk exposure for Google is substantial. Once a public incident occurs involving an AI-generated malicious app stealing user data, Google will face dual pressure from public opinion and regulators.
Regulatory implications vary by jurisdiction. Under the EU AI Act, general-purpose AI generation tools fall under the limited risk tier, but if they involve app distribution, transparency obligations may be triggered. In China, if the feature enters the market, it would face algorithm filing, generative AI service management measures, content security, and other strong compliance requirements. Whether private apps are exempt remains unclear. In the United States, there is currently no specific regulation for app generation tools, but the FTC maintains attention on data security and consumer protection.
Unanswered Critical Questions
How is the permission model for generated apps designed? Can they access sensitive user data? Is there runtime sandbox isolation? Does code execute locally or in the cloud? Is there safety filtering for generated content? How are violating code blocks intercepted? Do Play developer policies and liability terms apply?
Confidence Assessment: Medium
The risk framework is general reasoning—AI generation plus no review equals risk. The logic holds. However, the announcement provides no information on security mechanisms, permission design, or review strategy. The specific risk levels are inferred values.
Investment and Valuation Analysis: Marginal Impact on Alphabet
Direct Impact on Alphabet
The valuation impact on Alphabet is negligible. This is a product feature iteration, not an independent financing or valuation event. The direct revenue contribution is likely minimal, given that private apps bypass the monetization pipeline.
The real investment implication lies in the marginal impact on the secondary market's AI application development track. It may suppress the valuation narrative of independent vibe coding platforms and low-code companies.
Potential Beneficiaries and Losers
At the Alphabet level, the feature does not generate significant incremental revenue, and the valuation impact is close to zero. However, it strengthens the AI plus platform narrative, which is mildly positive for sentiment.
Potentially affected targets include independent AI app generation platforms (Lovable, Bolt, Replit, and others in the primary market) and traditional low-code and no-code companies if their moats are eroded by OS vendors.
Potential beneficiaries include the Google Cloud and TPU ecosystem and Android ecosystem service providers.
Signals to Track
Whether the feature is charged for or bundled with a Gemini subscription will affect the commercialization narrative. Whether there is an enterprise version for internal tool development will affect the enterprise low-code market such as Power Apps.
Confidence Assessment: Low to Medium-Low
There is a lack of valuation-related facts. This is purely directional inference. Given Alphabet's scale, the financial impact of this feature is minimal.
Infrastructure and Compute Analysis: Negligible Incremental Demand
The Compute Picture
The incremental compute demand is negligible. The inference load for generating lightweight apps is far smaller than model training. Google has its own TPUs and global cloud infrastructure, so marginal costs can be absorbed internally.
The real compute significance lies in the scale effect of Gemini inference invocation volume. If the feature becomes widely adopted, it will slightly increase Gemini inference load, but it will have no substantial impact on the overall compute landscape.
Core Reasoning
Generating apps is small-scale code inference. Token consumption is orders of magnitude lower than long-form or multimodal tasks. Google has compute self-sufficiency through TPUs and cloud, with no supply chain exposure risk. Compared to training cluster requirements at the tens-of-thousands-of-GPUs level, this kind of application inference demand is a rounding error.
Confidence Assessment: Medium
The judgment of infrastructure dependency relationships is based on Google's publicly known vertical integration facts. The logic is clear. However, specific compute data for this case is lacking.
Comprehensive Analysis: Strategic Positioning and Key Uncertainties
Integrated Judgment
This is a platform-level defensive layout by Google in the AI-native application development track. Its strategic essence is to use the vertical integration advantages of OS, store, model, and compute to defend the gateway to the application economy. It is not an independent technological breakthrough.
The technical core is a productized encapsulation of mature LLM code generation capability, representing engineering-level or combinatorial-level innovation. Commercially, it does not directly monetize and targets ecosystem lock-in.
The biggest point of interest is not that it can generate apps. It is the technical maturity and security considerations behind the private risk containment design. The announcement itself is severely information-deficient. Phrases such as "democratization," "accelerating innovation," and "reshaping the app economy" are optimistic narratives unsupported by evidence and should be treated with caution.
Top Three Key Risks
| Rank | Risk Description | Probability | Impact | Response Recommendation | |------|------------------|-------------|--------|-------------------------| | 1 | Private apps bypass review, leading to malicious code or data theft incidents | Medium | High | Track its sandbox and permission model design; monitor for first security incident | | 2 | Fragile generated app quality leading to user trust erosion and reputational backlash | Medium-high | Medium | Observe actual user experience feedback and retention data | | 3 | Regulatory determination as disguised evasion of app review or content compliance | Medium | Medium-high | Monitor EU and China regulatory statements |
Top Three Core Opportunities
| Rank | Opportunity Description | Capture Difficulty | Time Window | Action Recommendation | |------|-------------------------|--------------------|-------------|-----------------------| | 1 | Mature OS-level AI development capability giving rise to a new personal software market | High | Medium term | Position around AI-generated app peripherals: templates, hosting, monetization | | 2 | Deepening Google ecosystem lock-in creating Android and Gemini developer dividends | Medium | Short term | Early access to the Gemini developer toolchain | | 3 | Independent vibe coding platforms under pressure creating differentiation opportunities | Medium | Short term | Watch independent platforms transitioning to vertical or enterprise scenarios |
Signals to Track
In the short term, zero to six months, watch for official Google feature documentation, pricing, underlying model disclosure, and initial user experience feedback. Also watch for whether an AI-generated malicious app security incident occurs.
In the medium term, six to eighteen months, the key question is whether Google opens generated apps for Play listing, transitioning from private to public. This is the critical watershed for capability maturity. Also watch how independent vibe coding platforms respond and their financing dynamics.
In the long term, eighteen to thirty-six months, watch the degree of integration with Android system-level agents such as App Functions and the time gap with Apple's equivalent capabilities.
Article Bias Assessment
Information selection bias is high. The coverage contains only the positive aspects of democratization, accelerating innovation, and reshaping the economy. It completely lacks technical details, pricing, functional boundaries, security mechanisms, and competitor comparisons.
Emotional tendency bias is medium. There is clearly positive phrasing such as democratizes, accelerating innovation, and reshaping, with no balanced perspective.
Stakeholder bias is medium. Crypto Briefing is a crypto news platform reporting on a Google product unrelated to its domain. This suggests hot topic aggregation or general tech news filler, lacking independent editorial depth. There is a possibility of AI-generated or rewritten content. Source credibility is low.
Overall Confidence Assessment: Medium, Leaning Medium-Low
The core judgments of this report—that the technology is engineering-level encapsulation, that the business logic is ecosystem defense, that competition relies on vertical integration, and that security risks center on unreviewed private apps—are built on industry baseline knowledge and logical deduction. The directional judgments are reliable.
However, the first-stage input information is extremely thin: no timeline, no technical details, no official sources, no pricing. Almost all specific figures and mechanisms are inferred values.
This report should be treated as a list of hypotheses to be verified. Once Google official documentation or first-hand reporting is obtained, three key conclusions should be recalibrated: technical maturity, functional boundaries, and security mechanisms.
Timeliness Statement
This analysis is based on knowledge cutoff of early 2025 and existing industry cognition. If the feature receives major subsequent updates, such as opening public listing, disclosing underlying models, or publishing pricing, the conclusions of this report need corresponding revision.
Invalidation conditions: If Google actually uses a non-Gemini model, or if the feature actually supports full commercial distribution, then the three conclusions regarding technical maturity, commercialization logic, and security risk all need reassessment.
Broader Implications: The Future of Software Creation
The Democratization Narrative and Its Limits
The democratization narrative is compelling. It suggests a world where anyone with an idea can build an app without learning to code. This is genuinely transformative for prototyping, personal tools, and low-stakes experimentation.
However, the narrative has clear limits. Professional software development involves far more than code generation. It involves architecture decisions, security considerations, performance optimization, user experience design, testing, deployment, maintenance, and ongoing iteration. AI generation addresses the first step of a long journey.
The gap between a working prototype and a production-grade application remains vast. The announcement does not address this gap, and it is the gap where most of the value and risk reside.
The Platform Power Question
There is a deeper question about platform power. When the operating system vendor also controls the app store, the AI models, and the compute infrastructure, it occupies every layer of the stack. Developers who build on this stack are dependent on Google at every turn.
This concentration of power raises questions about fair competition, innovation incentives, and the terms under which developers operate. The private app designation may be a temporary technical limitation, but it also represents a decision about how much control to cede to users and developers.
The Attention Economy Dimension
If app generation becomes trivially easy, the bottleneck shifts from production to distribution and attention. The app stores of the future may be flooded with AI-generated content, making discovery and quality assessment even harder.
This suggests that curation, recommendation, and trust mechanisms will become more valuable, not less. The platforms that can effectively surface quality amid abundance will win. Google's existing investment in Play editorial and algorithmic curation positions it reasonably well here, but the challenge is unprecedented.
The Security Imperative
The security imperative is unavoidable. If AI-generated apps can access sensitive data, and if they bypass review, the potential for harm is significant. Google's decision to restrict to private apps may be a recognition of this risk.
The long-term solution likely involves a combination of runtime sandboxing, permission controls, automated security scanning, and possibly a new review paradigm designed for AI-generated code. None of these are trivial, and none are described in the announcement.
The Competitive Response
Competitors will respond. Apple has its own AI strategy. Microsoft has Copilot and Power Apps. OpenAI has GPTs and Canvas. Independent platforms will differentiate on model quality, developer experience, and specialized use cases.
The next eighteen months will reveal whether Google's vertical integration advantage translates into a dominant position or whether the historical product weakness reasserts itself. The answer depends on execution, which is precisely what the announcement does not describe.
Conclusion: A Signal Worth Watching, Not a Conclusion Worth Drawing
Google Play's move to allow private AI-generated Android apps is a signal worth watching. It indicates that AI-native development is moving from novelty to platform-level strategy. It suggests that operating system vendors see app generation as a competitive necessity.
But it is not a conclusion worth drawing. The announcement is too thin, the details too sparse, the evidence too limited. The optimistic narratives of democratization and innovation acceleration are plausible but unproven. The risks around security and review bypass are real but unquantified.
What we can say with reasonable confidence is this: the technology is mature enough to productize but not mature enough to trust with public distribution. The business logic is defensive rather than monetizing. The competitive dynamics favor vertical integration but hinge on execution. The security implications are significant and unresolved. The compute impact is negligible.
The next twelve to eighteen months will be telling. Watch for Google to open public listing, which would signal a leap in capability maturity. Watch for the first security incident, which would validate the risk concerns. Watch for competitor responses, which would reveal whether the moat holds. Watch for regulatory attention, which would determine the boundaries of acceptable practice.
Until then, this remains a hypothesis list. The private app may be a cautious first step or a permanent limitation. Only Google knows, and it has not said.
The broader arc, however, is clear. Software creation is being transformed. The question is not whether AI will change how apps are built, but who will control the transformation and under what terms. Google has made its move. The game is far from over.
Disclaimer: This analysis is based on a knowledge cutoff of early 2025 and existing industry cognition. It is not investment advice. Readers should verify all claims independently and consult primary sources. The original announcement lacked critical details, and much of this analysis is inferential. Confidence levels are explicitly noted throughout.