Economics of Infinite Software: Abundance, Automation, and the Post-Scarcity Paradigm [Analysis, 2026]

The foundational premise of the digital economy over the past three decades has rested upon a distinct economic asymmetry: the high fixed cost of software creation paired with the near-zero marginal cost of its distribution. This dynamic allowed technology enterprises to invest millions of dollars and thousands of engineering hours into a single codebase, which could then be duplicated and distributed to millions of users at virtually no additional expense. This fundamental mechanic created unprecedented “winner-take-most” monopolies, defining an era where software methodically consumed traditional industries¹.

Today, a structural inversion of this paradigm is underway. Artificial intelligence, propelled by generative architectures and autonomous agentic systems, is driving the marginal cost of software creation itself toward zero¹. When the capacity to write, test, and deploy functional code is untethered from the scarcity of human engineering labor, the economic moats that have historically protected software companies evaporate. This transition from scarcity to abundance represents a discontinuous disruption. When algorithmic labor allows software to write software, the constraints of the digital economy shift entirely, fundamentally rewriting the logic of value creation, altering the trajectory of venture capital, and redefining the organizational structures of global technology hubs³.

The Carlota Perez Framework and the Sixth Technological Surge

To fully grasp the macroeconomic implications of zero-marginal-cost software, the evolution of artificial intelligence must be situated within the broader context of historical technological revolutions. The Venezuelan economist Carlota Perez provides a robust theoretical model detailing how capitalism undergoes a major technological revolution every forty to sixty years⁶. Historically, these techno-economic paradigms follow a predictable rhythm, driven by waves of innovation and reshaped by financial speculation.

According to Perez’s framework, every major technological surge is divided into two primary acts: an installation phase and a deployment phase, separated by a period of structural crisis and institutional upheaval⁷. The installation phase is characterized by the “irruption” of new technologies, followed by a “frenzy” of speculative financial capital that heavily funds infrastructure while decoupling from actual production capital⁸. Eventually, the market reaches a turning point—often punctuated by a financial crash—before entering the deployment phase. This latter phase, divided into “synergy” and “maturity,” is where production capital re-engages, institutions adapt to the new paradigm, and the technology experiences widespread, stable diffusion throughout the broader economy⁷.

Technological Revolution	Core Innovations and Infrastructure	Key Geographies	Irruption Year
First: Industrial Revolution	Mechanized cotton, wrought iron, canals, water power	Britain	1771
Second: Age of Steam and Railways	Steam engines, machine tools, railways, ports	Britain, spreading to Europe and USA	1829
Third: Age of Steel and Electricity	Cheap steel, heavy chemistry, electrical networks	USA and Germany	1875
Fourth: Age of Oil and Mass Production	Automobiles, petrochemicals, mass consumerism	USA, spreading globally	1908
Fifth: Age of Information and Telecom	Microprocessors, software, the internet, cloud computing	USA, spreading globally	1971

The digital revolution that began with the microprocessor represents the fifth major technological surge¹⁰. Recent macroeconomic indicators suggest that this fifth paradigm has decisively entered its maturity phase. Global IT and software revenues are expanding at just four to five percent annually, suggesting that the markets are heavily saturated, the major platforms have consolidated, and the easy gains of basic enterprise digitization have been harvested⁷.

The emergence of generative artificial intelligence signifies either the final, most radical extension of the information age, or the explosive irruption phase of an entirely new, sixth technological revolution⁷. The global economy is currently exhibiting the classic hallmarks of an installation frenzy. Hyperscaler technology companies are projected to deploy approximately $825 billion in capital expenditure to build underlying artificial intelligence infrastructure¹². However, the AI industry’s reliance on brute-force scaling laws—the assumption that training compute will double every six months—is colliding with physical and economic reality. If current scaling trends continue unabated, training a single large AI model by 2035 would require financial resources exceeding the entire United States economy, a mathematical impossibility akin to scaling the Apollo Program¹³. Furthermore, foundation models are rapidly exhausting publicly available, high-quality human data, with estimates suggesting data scarcity will become a critical bottleneck between 2026 and 2032¹³.

This impending technical ceiling forces a transition from the frenzied installation of infrastructure to the intelligent deployment of reasoning systems. The focus is shifting from “thinking fast” (rapid pattern matching) to “thinking slow” (System 2 deliberate reasoning at inference time), utilizing architectures that pause to evaluate outcomes and solve complex, out-of-sample problems⁵. As capital markets demand clearer evidence of returns on massive infrastructure investments, the industry is pivoting from building models to extracting real-world utility, setting the stage for the complete disruption of the traditional software business model¹².

The Collapse of the SaaS Moat and the Build-versus-Buy Reset

For two decades, the Software-as-a-Service (SaaS) business model has dominated the technology landscape. The economic logic of SaaS relied heavily on the absolute scarcity and high cost of human engineering talent. Because building bespoke internal software was cost-prohibitive for most non-technology enterprises, companies were forced to “buy” rather than “build.” This dynamic pushed enterprises into purchasing access to bloated, one-size-fits-all SaaS products that amortized high development costs across tens of thousands of corporate clients².

To justify continuous price increases, SaaS vendors perpetually added complex features, leading to severe software bloat, vendor lock-in, the accumulation of technical debt from integration complexities, and the widespread procurement of unused “shelfware”¹⁶. The fundamental premise was that proprietary code created a defensible, high-margin economic moat. Today, as advanced models can autonomously generate code, that moat is dry¹⁸.

When coding capacity becomes effectively unlimited, the marginal cost of software variation drops to near zero². This capability fundamentally flips the “build versus buy” calculus for the enterprise. Instead of engaging in a multi-person, multi-year software implementation project, organizations can now utilize agentic development environments to rapidly generate bespoke micro-SaaS applications¹⁶. Rather than paying to access their own customer data through a generic third-party interface, enterprises can deploy autonomous agents to build precisely the interface they need, connected directly to their proprietary databases¹⁷.

This shift transforms business model design from a restricted art constrained by engineering budgets into an unlimited science of instant prototyping³. Because the cost of generating software is near zero, the resulting applications are effectively disposable. When an enterprise workflow evolves, the underlying custom application is no longer meticulously patched, refactored, or migrated across framework versions; it is simply discarded and regenerated from scratch to perfectly match the new process². The traditional agile sprint of one to three weeks is compressed into an end-to-end project sprint lasting only days, birthing an era of rapidly replaceable digital experiences that are economically superior to enterprise SaaS licensing¹⁷.

Service-as-a-Software: The Trillion-Dollar Arbitrage

As the intrinsic value of proprietary code approaches zero, the strategic imperative for technology ventures is shifting from selling software tools to selling completed business outcomes. This paradigm shift, heavily championed by prominent venture capital firms, represents the transition from Software-as-a-Service to “Service-as-a-Software”⁵.

Historically, the total addressable market for software companies was strictly capped by corporate IT budgets. However, a stark economic ratio has defined the limits of SaaS valuations: for every single dollar a business spends on a software license, it spends approximately six dollars on the human labor required to operate that software and execute the underlying service¹⁹. Traditional cloud companies targeted the software profit pool and sold access by the seat; AI-native companies are now targeting the vast services profit pool and selling work by the outcome⁵.

By deploying autonomous reasoning agents that replicate, augment, or entirely replace human workflows, Service-as-a-Software firms absorb both the software tool and the labor simultaneously¹⁴. If a vendor sells a traditional software tool, they are caught in an unwinnable race against open-source foundational models. Conversely, if a vendor sells the finalized outcome—the drafted legal contract, the reconciled financial audit, the resolved customer claim—every incremental improvement in underlying AI capabilities simply makes the service faster, cheaper, and more difficult for legacy incumbents to compete with⁵.

Economic Characteristic	Traditional SaaS Model	Service-as-a-Software Model
Core Value Proposition	Providing digital tools for human operators	Delivering autonomous, finalized business outcomes
Target Budget Pool	Corporate IT and Software Licensing budgets	Broader Operational and Workforce Labor budgets
Standard Pricing Mechanism	Fixed subscription per user (Seat-based pricing)	Variable fee per completed unit of work (Outcome-based)
Typical Gross Margins	80% to 90%	45% to 60% (influenced by heavy inference compute costs)
Strategic Defensibility	High switching costs, proprietary codebases	Domain expertise, regulatory trust, verified outcomes

While this model unlocks access to trillions of dollars in global labor spend, it fundamentally alters the margin profile of the technology sector. Software companies historically enjoyed gross margins approaching 90 percent. In the artificial intelligence era, a nonlinear combination of heavy inference compute costs, the necessity of handling complex corner cases, and the integration of open-source components drag gross margins down to the 45 to 60 percent range²¹. However, because the addressable market spans the entire operational expenditure of the enterprise rather than a fraction of the IT budget, the absolute revenue potential dwarfs the legacy SaaS ecosystem¹⁴.

The 3H Framework and the Automation Asymptote

The extent to which any specific service vertical can be transformed into a high-margin software business is governed by its “Automation Asymptote.” This asymptote represents the natural boundary between machine execution and absolute human necessity. To systematically evaluate which service industries are ripe for AI disruption, strategic frameworks such as the 3H Model assess enterprise workflows against three fundamental constraints: Hands, Hearts, and Handcuffs¹⁸.

The “Hands” constraint measures the necessity of physical presence or manipulation. Workflows requiring on-site inspections or physical object handling cannot be fully consumed by software¹⁸. The “Hearts” constraint evaluates the need for emotional intelligence, empathy, and nuanced psychological negotiation, such as the final stages of closing a complex sale¹⁸. Finally, the “Handcuffs” constraint dictates the regulatory, ethical, and legal barriers to automation, including strict data privacy laws, broker liabilities, and mandated human-in-the-loop sign-offs for critical compliance matters¹⁸.

When applied to various service verticals, the 3H Framework dictates the maximum potential for margin extraction. In real estate brokerage, artificial intelligence can autonomously execute lead generation, comparative market analysis, and document preparation. However, the physical property viewing (Hands) and the psychology of the final negotiation (Hearts) cap the agentic capability at an automation asymptote of roughly 70 to 80 percent, allowing the human to retain a physical presence and trust premium¹⁸. In heavily regulated fields such as financial auditing (Handcuffs), AI agents can ingest massive ledgers and perform intelligent anomaly detection, but the legal mandate for professional human skepticism caps the automation asymptote at 75 to 85 percent¹⁸.

Conversely, in sectors such as insurance brokerage and claims administration, the workflows are heavily dictated by strict contractual parameters, probabilistic mathematical models, and structured documentation. In these environments, zero Hands and zero Hearts are required, allowing the automation asymptote to approach 100 percent, constrained only by minor broker liability regulations¹⁸. This allows AI-native platforms to drive almost pure software-level margins within a legacy service industry.

To capture these varying margin profiles, three distinct architectural pathways are emerging for AI-native service firms: the “Business in a Box,” which equips independent professionals with agentic toolkits to increase take rates; the “Full Stack Autonomous Service Provider,” which bypasses software vendors entirely to employ domain experts who systematically agentify their own workflows; and the “AI Enabled Roll Up,” which acquires low-margin legacy service companies and aggressively automates their back-office operations to engineer compounding private equity returns¹⁸.

Jevons Paradox and Evolving Value of the Developer

A widespread assumption regarding the automation of software engineering is the impending obsolescence of the human developer. However, the application of classical economics—specifically the Jevons Paradox—suggests a counterintuitive reality. Formulated over 160 years ago, the Jevons Paradox dictates that as technological progress increases the efficiency of utilizing a resource, the overall consumption of that resource rises due to falling costs²³. Because artificial intelligence radically reduces the cost of producing software, the aggregate demand for specialized, highly personalized, and internal software is exploding⁴. Lower costs lead to higher demand, resulting in an unprecedented volume of software creation and amplifying systemic complexity⁴.

Extensive empirical research provides a granular view of this transformation. Working papers and randomized controlled trials evaluating the introduction of generative AI tools across thousands of software engineers and customer support agents reveal profound productivity shifts²⁴. In controlled environments involving over 5,000 agents at a Fortune 500 software firm, access to generative conversational assistants increased overall productivity by 14 to 15 percent²⁴.

However, these gains are distributed highly unevenly. Novice and lower-skilled workers experienced massive productivity improvements of up to 34 percent, allowing them to rapidly move down the experience curve. Agents with only two months of tenure were able to perform on par with untreated workers possessing over six months of experience²⁴. The artificial intelligence model effectively disseminates the tacit knowledge and best practices of top performers, raising the baseline quality of the entire workforce²⁴. Conversely, the most experienced and highly skilled workers saw minimal impacts on productivity, with some even experiencing slight declines in quality when overly reliant on the tools²⁴.

When evaluating software development specifically, the productivity impacts scale dramatically with the autonomy of the AI tool. Studies evaluating tools like GitHub Copilot and Cursor across hundreds of thousands of engineers demonstrate that basic autocomplete functions increase task-level productivity by approximately 40 percent. The introduction of interactive, synchronous coding agents pushes this cumulative effect to 140 percent, while the deployment of autonomous, asynchronous agents capable of authoring and submitting independent commits raises productivity gains to 180 percent²⁶.

This massive acceleration in routine execution generates a severe “barbell effect” within the technology labor market. The demand for junior engineering roles focused on repetitive boilerplate writing, CRUD (Create, Read, Update, Delete) generation, and low-complexity implementation is rapidly shrinking, as AI absorbs the scaffolding of software development⁴. Simultaneously, the value of senior systems thinkers is skyrocketing. The core competency of the human developer is pivoting away from writing syntax toward systems architecture, product ownership, and AI orchestration². The most valuable engineers in the post-scarcity software economy are those who understand the logical boundaries of complex systems, who know precisely when not to write code, and who can effectively translate ambiguous business intent into safe, verifiable, machine-driven execution².

Technical Debt, Governance, and the Trust Wall

The democratization of code generation introduces severe architectural and security risks. The primary failure mode of the infinite software era is not an inability of the artificial intelligence to produce functional code; rather, it is the generation of an avalanche of unreviewed logic that human operators fundamentally do not understand². When organizations permit “vibe coding” without robust governance, developers routinely accept massive, multi-hundred-line code diffs without thorough review, prioritizing speed over comprehension³⁰.

This dynamic rapidly compounds into a virulent new form of AI-driven technical debt. Unreviewed, machine-generated code drifts significantly from established architectural standards, becoming an opaque liability³⁰. Furthermore, generative models frequently reproduce the historical vulnerabilities present in their training data, introducing SQL injection vectors, missing authentication checks, and hardcoded secrets directly into enterprise production environments³⁰. Because AI introduces probabilistic, non-deterministic behavior into historically deterministic systems, debugging becomes exceptionally complex⁴. If an AI-generated system fails, and the engineering team lacks a mental model of how the specific function was implemented, the codebase must often be abandoned and entirely rebuilt².

To prevent these catastrophic governance failures, the generation of code must be strictly decoupled from its deployment through the construction of a “Trust Wall”². Standard human peer review is fundamentally incompatible with the speed and volume of autonomous generation. Consequently, organizations are forced to invest heavily in automated verification systems².

This Trust Wall necessitates replacing manual testing paths with property-based testing that continuously checks system invariants. It requires formal verification protocols that provide mathematical proofs that machine-generated code adheres strictly to human-defined specifications². Furthermore, it demands sandboxed execution environments to quarantine and observe side effects before production deployment, backed by real-time circuit breakers that automatically revert systems if anomalous behavior is detected². In this automated paradigm, the negative space matters significantly more than the positive space. Instructing an AI on what it must never do—such as dictating that a financial balance must never drop below zero, or that a system must never expose personally identifiable information in its logs—becomes the primary mechanism of control². Ultimately, the verification and observability teams emerge as the most vital engineering units, shifting the organizational focus from writing code to building the automated reasoning systems that gate its deployment².

Intellectual Property, Copyright, and Differentiation

As software generation shifts from human keyboards to inference clusters, enterprises face a labyrinth of unresolved intellectual property and copyright challenges. Historically, global copyright law—and particularly United States law—has been anchored immutably in the concept of human authorship³². The legal framework protects original works created by human beings who exercise deliberate creative judgment, leaving the status of machine-generated code in a precarious legal gray area³².

The U.S. Copyright Office has repeatedly affirmed that works created entirely by artificial intelligence, lacking meaningful human involvement, are not eligible for copyright protection³³. Landmark decisions, ranging from the denial of copyright for the “monkey selfie” in Naruto v. Slater (2018) to the rejection of AI-generated art in Thaler v. Perlmutter (2023), establish a clear precedent: copyright ownership does not vest in a machine³⁴. When an AI system produces the expressive elements of an output based solely on a user’s prompt, that generated material is devoid of human authorship, is legally unprotected, and falls immediately into the public domain³⁴.

For enterprises accustomed to treating proprietary codebases as highly defensible corporate assets, this presents a severe commercial risk. If a developer uses a generative tool to instantly build a complex application with minimal post-editing, the enterprise cannot claim ownership of that software. Consequently, the organization loses the legal authority to prevent competitors from copying the codebase, to issue infringement takedown notices, or to license the software to third parties³².

To secure copyright protection, a human developer must meaningfully contribute to the creative process. This entails substantial editing, curation, and the creative arrangement of AI-generated components such that the resulting work, as a whole, constitutes an original work of human authorship³³. The U.S. Copyright Office mandates that applicants explicitly identify and disclaim AI-generated content within their registration applications, ensuring that copyright protection extends solely to the human-authored elements, as demonstrated in the mixed-registration decision for the graphic novel Zarya of the Dawn³⁴.

Beyond the inability to protect new assets, the deployment of machine-generated code introduces profound infringement and licensing liabilities. Generative models are trained on massive datasets that include public repositories governed by restrictive open-source licenses, such as the General Public License (GPL) or MIT license³². If an autonomous coding agent inadvertently replicates licensed material, the enterprise deploying the code may unknowingly violate those licenses. This copyleft contamination exposes the organization to severe litigation, including the potential legal requirement to open-source their entire proprietary system³². To mitigate these existential legal threats, enterprises must implement forensic source code analysis, authorship validation workflows, and continuous licensing compliance audits before allowing AI-generated code to enter production³².

Global Innovation Hubs: The Indian IT Reset and European Regulation

The economic shockwaves of zero-marginal-cost software are forcing rapid structural transformations across the world’s primary technological hubs, driving a stark divergence in regional strategies.

Transformation of Indian IT Services

Nowhere is the disruption more immediate than in the Indian IT services sector, an industry historically valued at over $315 billion annually⁴⁰. For decades, giants such as Tata Consultancy Services (TCS), Infosys, Wipro, and HCLTech built their empires on a labor-intensive outsourcing model. Their competitive advantage relied on a “pyramid model,” leveraging immense scale by deploying tens of thousands of entry-level engineers to bill global clients on a linear, time-and-materials basis⁴⁰.

As generative AI automates up to 37 percent of entry-level programming, testing, and documentation tasks, the demand for raw engineering headcount is plummeting⁴⁰. Enterprise clients, acutely aware of AI-driven productivity gains, are demanding steep price cuts, with reports indicating requests for 25 to 30 percent reductions in contract pricing for legacy work⁴⁰. This AI-driven revenue deflation has shattered the traditional hiring pyramid; the top five Indian IT firms recently reversed years of growth by reducing their collective workforce by thousands of employees, with TCS alone implementing mass layoffs exceeding 12,000 personnel in a single year⁴⁰.

To survive this pricing reset, the Indian IT sector is aggressively transitioning away from manpower-led billing toward outcome-based service delivery⁴². Firms are repositioning themselves as orchestrators of business processes, embedding AI agents into client operations and tying their compensation directly to measurable performance outcomes and unit-cost reductions rather than hours worked⁴⁰.

Despite short-term deflationary headwinds, the sector is positioned for a massive growth recovery as global technology spending pivots from building foundational AI infrastructure toward widespread enterprise deployment and system integration¹². This expansion of the total addressable market requires immense support in data preparation, legacy COBOL modernization, AI governance, and cybersecurity¹². Notably, agile mid-tier firms—such as LTIMindtree, Coforge, and Persistent Systems—are capturing a majority of the sector’s incremental organic revenue by rapidly deploying senior talent and offering flexible, AI-native pilot programs, proving that nimbleness is beginning to outweigh massive scale in the post-scarcity landscape⁴⁰.

European Ecosystem and Regulatory Divergence

Simultaneously, the geopolitical balance of software innovation is being heavily influenced by the concentration of specialized talent and the divergence of regulatory frameworks across Europe. The continent has cultivated a highly attractive environment for AI scaling, possessing a talent pool of nearly 200,000 engineers with AI experience, representing a per-capita concentration roughly 30 percent higher than the United States and nearly three times that of China⁴⁵.

London remains the dominant epicenter of this talent, harboring over 20,000 AI experts, anchored by historical legacies like Alan Turing and massive modern employers such as Google DeepMind⁴⁵. However, significant hubs are rapidly maturing in Berlin and Paris, each commanding dense concentrations of top-tier talent⁴⁵. A pivotal cultural shift is accelerating this growth: elite AI practitioners are increasingly abandoning academia and legacy tech conglomerates, drawn instead to the immediate impact and equity upside of agile early-stage startups⁴⁵.

The future trajectory of these European hubs, however, will be sharply defined by political and regulatory environments. The European Union has adopted a strategy of heavy state intervention, aiming to funnel €20 billion annually into AI development by the end of the decade through its Coordinated Plan on Artificial Intelligence⁴⁵. Yet, this investment is counterbalanced by the stringent EU AI Act. This comprehensive legislation prioritizes consumer safety, ethical deployment, and rigorous compliance, imposing severe regulatory burdens on the development and adoption of AI systems⁴⁵.

This regulatory friction poses an existential challenge to resource-constrained startups in hubs like Berlin and Paris, potentially suffocating rapid iteration in favor of bureaucratic compliance. In stark contrast, the United Kingdom has structured its National AI Strategy to foster a looser, pro-innovation regulatory environment designed explicitly to maximize commercial development and strategic leadership⁴⁵. As the economics of infinite software demand high-speed experimentation and immediate deployment, this tension between the European Union’s protective legislation and the United Kingdom’s accelerated commercialization will fundamentally dictate the flow of venture capital and the competitive hierarchy of the global technology sector over the coming decade⁴⁵.

Final Thoughts on Post-Scarcity Economy

The transition to a zero-marginal-cost paradigm for the creation of software marks a definitive rupture in modern economic history. The foundational era in which proprietary, hand-crafted code served as an impregnable economic moat has definitively concluded. As the primary bottleneck of technology creation shifts from the mechanical writing of syntax to the orchestration and governance of intelligent systems, the broader global economy is experiencing widespread, multi-sector upheaval.

The ramifications of this shift extend into every facet of commerce and law. Traditional software-as-a-service models are facing obsolescence, rapidly replaced by outcome-based Service-as-a-Software paradigms that bypass IT budgets entirely to capture trillions of dollars in global labor spending. The human workforce is experiencing a violent bifurcation: routine cognitive execution is being relentlessly automated, while the economic premium placed on high-level systems architecture, complex reasoning, and automated governance has reached unprecedented heights. Simultaneously, the legal foundations of intellectual property are fracturing under the realities of machine generation, forcing enterprises to navigate treacherous risks involving public domain exposure and copyleft licensing contamination.

To thrive in the post-scarcity era of infinite software, organizations must fundamentally abandon the legacy structures of the information age. Success will no longer be determined by the ability to mass-produce code, but by the capacity to verify autonomous outputs, integrate proprietary data into dynamic workflows, navigate shifting geopolitical regulatory regimes, and consistently deliver verifiable business outcomes. As the world transitions out of the frenzy of the installation phase and firmly into the deployment phase of this new technological revolution, the ultimate victors will be those who recognize a profound new reality: when the cost of building software drops to zero, the only remaining constraints on value creation are human imagination and strategic architectural intent.

Works cited

1. Accelerate Monthly – Software as a Disservice – Julian Klymochko, https://julian-34159.medium.com/accelerate-monthly-software-as-a-disservice-aafcad6abbcc
2. When Code Becomes A Liability, Not An Asset | by Travis Williams, https://medium.com/@travis.hunt.williams/when-code-becomes-a-liability-not-an-asset-7226a89de822
3. AI Programming: The Overlooked Engine of the Business Model, https://eu.36kr.com/en/p/3531450735107200
4. The State of Software in the Age of AI (2026–2030) – Aqweeva®, https://aqweeva.com/blog
5. Embracing the Shift to Service-as-a-Software: The Future of AI, https://rosebudpr.io/embracing-the-shift-to-service-as-a-software-the-future-of-ai-applications/
6. Tech Bubble or Golden Age? | Whitepapers – 8VC, https://www.8vc.com/resources/tech-bubble-or-golden-age
7. AI and the Perez Cycle: is a new technological surge underway, https://hermasayi.medium.com/ai-and-the-perez-cycle-is-a-new-technological-surge-underway-edbe476442e4
8. Technological Revolutions and Financial Capital – Wikipedia, https://en.wikipedia.org/wiki/Technological_Revolutions_and_Financial_Capital
9. The Five Technology Cycles, https://technologycycles.org/five-technology-cycles/
10. (PDF) Technological revolutions and techno-economic paradigms, https://www.researchgate.net/publication/46513433_Technological_revolutions_and_techno-economic_paradigms
11. Technological revolutions and techno-economic paradigms | e-TCS, https://e-tcs.org/wp-content/uploads/2012/04/PEREZ-Carlota-Technological-revolutions-and-techno-economic-paradigms1.pdf
12. Indian IT companies may see growth recovery from FY27 as AI spending shifts to enterprise adoption, https://m.economictimes.com/ai/ai-insights/indian-it-companies-may-see-growth-recovery-from-fy27-as-ai-spending-shifts-to-enterprise-adoption/articleshow/133647426.cms
13. The State of AI in 2025 – Leonis Capital, https://www.leoniscap.com/research/the-state-of-ai-in-2025
14. Reveal a Better Future – Revaia, https://www.revaia.com/news/service-as-a-software-how-ai-enables-the-next-saas-frontier/
15. Generative AI’s Act o1: The Reasoning Era Begins | Sequoia Capital, https://sequoiacap.com/article/generative-ais-act-o1
16. Flipping the math: How AI changes Build vs. Buy – CustomerThink, https://customerthink.com/flipping-the-math-how-ai-changes-build-vs-buy/
17. The Fast Fashion Era of SaaS: How Agentic Development Disrupted, https://rangle.io/blog/fast-fashion-era-of-saas-how-agentic-development-disrupted-our-business-model
18. The AI Native Economy: From Software to Services with the 3H Model, https://thedelta.io/blog/the-ai-native-economy-from-software-to-services-with-the-3h-model
19. How we’re using AI in our accounting firm | Flare, https://withflare.co/blog/how-were-using-ai-in-our-accounting-firm
20. Service as a Software – – AxisOfEasy, https://axisofeasy.com/best-of-the-web/service-as-a-software/
21. AI Gross Margins: Why the 90% SaaS Benchmark Is Gone, https://jeffbrokaw.com/blog/ai-gross-margins/
22. Understanding AI startups – Max Versace, https://maxversace.com/understanding-ai-startups
23. Jevons’ Paradox and AI: Why It Means More Developers, Not Fewer, https://www.mrphilgames.com/blog/jevons-paradox-software-development-ai-more-developers
24. NBER WORKING PAPER SERIES GENERATIVE AI AT WORK Erik, https://www.nber.org/system/files/working_papers/w31161/w31161.pdf
25. NBER WORKING PAPER SERIES GENERATIVE AI AT WORK Erik, https://www.nber.org/system/files/working_papers/w31161/revisions/w31161.rev0.pdf
26. The Effects of Generative AI on High-Skilled Work – MIT Economics, https://economics.mit.edu/sites/default/files/inline-files/draft_copilot_experiments.pdf
27. Generative AI at Work – arXiv, https://arxiv.org/pdf/2304.11771
28. Artificial Intelligence in the Firm – Fiona Chen, https://fion.ac/jellyfish.pdf
29. Productivity Effects Across Generations of AI Coding Tools, https://www.nber.org/system/files/working_papers/w35275/w35275.pdf
30. Vibe Coding for Enterprise: Risks & Controls – Tembo.io, https://www.tembo.io/blog/vibe-coding-for-enterprise
31. The Human Bottleneck – Scott Logic Blog, https://blog.scottlogic.com/2026/05/14/the-human-bottleneck.html
32. Ai-generated Code and Copyright: Who owns Ai-written Software?, https://eurekasoft.com/blog/aigenerated-code-and-copyright-who-owns-aiwritten-software
33. AI and Copyright Ownership: Rights in Generated Works | Trestle Law, https://www.trestlelaw.com/blog/ai-and-copyright-ownership-rights-in-generated-works
34. Who Owns Generative AI Copyrights in Major Jurisdictions, https://finkellawgroup.com/2024/10/21/copyright-ownership-of-generative-ai-work-product-in-major-international-jurisdictions/
35. Who Owns the Copyright to AI-Generated Works?, https://copyrightalliance.org/faqs/artificial-intelligence-copyright-ownership/
36. Copyright Office Releases Part 2 of Artificial Intelligence Report, https://www.copyright.gov/newsnet/2025/1060.html
37. Copyright and Artificial Intelligence | U.S. Copyright Office, https://www.copyright.gov/ai/
38. Works Containing Material Generated by Artificial Intelligence, https://www.copyright.gov/ai/ai_policy_guidance.pdf
39. The U.S. Copyright Office is refusing to register copyright in AI, https://www.reddit.com/r/COPYRIGHT/comments/1ujl0da/the_us_copyright_office_is_refusing_to_register/
40. AI reshapes India’s IT services sector contracts as clients demand, https://m.economictimes.com/tech/artificial-intelligence/ai-reshapes-indias-it-services-sector-contracts-as-clients-demand-more-for-less/articleshow/133391316.cms
41. Artificial Intelligence and Future of India’s IT Companies, https://www.epw.in/engage/article/artificial-intelligence-and-future-indias-it
42. IT companies gradually shift toward outcome-based service delivery, https://www.newindianexpress.com/business/2026/May/12/it-companies-gradually-shift-toward-outcome-based-service-delivery-model
43. TCS Just Put a Number on AI-Led IT Services: What Its Q1 Says, https://callitdev.com/ko/blog/tcs-ai-services-inflection-it-outsourcing-buyer-guide-2026
44. Layoffs at Uber; Astrobase eyes fresh fundraise, https://m.economictimes.com/tech/newsletters/tech-top-5/uber-cuts-3300-jobs-inside-happiest-minds-itc-infotech-merger/articleshow/133711742.cms
45. Europe’s AI Engineering Hubs – Sequoia Atlas, https://atlas.sequoiacap.com/a-talented-home-for-ai/
46. Berlin’s AI Startup Scene in 2026: Why the German Capital Is, https://www.contextstudios.ai/blog/berlins-ai-startup-scene-in-2026-why-the-german-capital-is-europes-ai-powerhouse
47. impact of the EU AI Act on startups – appliedAI Institute, https://www.appliedai-institute.de/en/resources/ai-act-impact-survey/

© 2026 Museum of Vibe Coding — Research Division. All rights reserved. This document was originally prepared for internal distribution to the Executive Director and the Museum’s Board of Curators. It was approved for public release on September 5, 2026. Cite as: Museum of Vibe Coding Research Division.