Vibe Coding Kills Open Source

Abstract: Generative AI is changing how software is produced and used. In vibe coding, an AI agent builds software by selecting and assembling open-source software (OSS), often without users directly reading documentation, reporting bugs, or otherwise engaging with maintainers. We study the equilibrium effects of vibe coding on the OSS ecosystem. We develop a model with endogenous entry and heterogeneous project quality in which OSS is a scalable input into producing more software. Users choose whether to use OSS directly or through vibe coding. Vibe coding raises productivity by lowering the cost of using and building on existing code, but it also weakens the user engagement through which many maintainers earn returns. When OSS is monetized only through direct user engagement, greater adoption of vibe coding lowers entry and sharing, reduces the availability and quality of OSS, and reduces welfare despite higher productivity. Sustaining OSS at its current scale under widespread vibe coding requires major changes in how maintainers are paid.

• The paper demonstrates that AI-mediated vibe coding reduces monetizable engagement, rapidly contracting OSS quality and quantity.
• It employs a two-sided monopolistic competition model with nested logit user choice to quantify shifts in developer behavior under AI adoption.
• Empirical data reveal sharp increases in AI-generated code alongside falling user engagement, highlighting the need for new monetization channels.

The Collapse of Open Source Under Vibe Coding: An Analytical Overview

Introduction

"Vibe Coding Kills Open Source" (2601.15494) presents a rigorous general equilibrium analysis of how generative AI—specifically AI-driven code assembly via so-called "vibe coding"—alters the economic underpinnings of open source software (OSS) ecosystems. The paper formalizes a model in which AI agents select, modify, and assemble OSS libraries, dramatically lowering the marginal cost of software development for end users while simultaneously disintermediating the traditional interaction channels through which OSS maintainers monetize and capture engagement. The central finding is that, under legacy engagement-based monetization, the adoption of vibe coding irreversibly contracts the OSS supply: increased productivity and cost savings are insufficient to compensate for the abrupt contraction in monetizable engagement, resulting in rapid declines in OSS variety, quality, and welfare.

Empirical Motivation and Macro Trends

Rapid empirical adoption of AI-assisted software synthesis tools has transformed software production. By late 2024, up to 30% of new code at major tech firms was AI-generated, with even higher rates (70–90%) reported in industry leaders (2601.15494). Micro-level analysis using GitHub commit data reveals similar trends for Python and other languages; AI-generated code spiked from near zero to nearly 30% in less than two years, with measurable productivity increases (circa 3.6%) among adopters. SWE-bench benchmarks document orders-of-magnitude jumps in end-to-end code issue resolution, with top-tier models leaping from sub-2% to >70% issue closure rates in 18 months.

Figure 1: Adoption of AI across industries and within software development, documenting diffusion from highly information-intensive sectors and visualizing the substitution of generative AI in the practical software engineering workflow.

Concurrently, the paper demonstrates that OSS engagement has decoupled from usage growth. Platforms reliant on public engagement (Stack Overflow, open Q&A forums) experienced a 25%+ collapse in user activity in direct response to LLM deployment. Maintainers of prominent OSS projects documented sustained or increasing download rates (such as npm downloads for Tailwind CSS), coupled with steep declines in documentation traffic, Q&A, and revenue. Engagement externalities are no longer internalized by maintainers, undermining the reward channels that financially sustain OSS.

Figure 2: Usage grows while engagement falls, highlighting OSS projects where utility and adoption expand but direct developer engagement—and thus revenue and contributions—plummet under AI mediation.

Outcome concentration is extreme: only a minuscule fraction of repos achieve substantial adoption, with power-law (Pareto) tails in both stars and downstream dependencies. Most repositories receive zero direct attention. This justifies modeling OSS provision as a heavy-tailed selection process.

Figure 3: Concentration of attention and usage in open source software, empirically corroborating the heavy-tailed nature of successful OSS adoption and reuse.

Theoretical Model

The core analytical engine is a two-sided, monopolistic competition model featuring endogenous entry, developer selection, and a software-begets-software production feedback. Key features:

• Nonrival input: OSS is a nonrival input into further software production. Variety and quality improvements recursively lower the cost of new OSS, generating substantial positive feedback.
• Heterogeneous project quality: Developers can release or withhold projects ex post, drawing quality from a Pareto distribution, which rationalizes the skewed outcome distribution and endogenous selection into release based on prospective returns.
• User choice and demand: Final users select OSS packages via a nested logit model, with inner nests corresponding to direct and vibe-coded modes. The elasticity of substitution $\theta$ between direct and AI-mediated use is central: high $\theta$ (empirically supported by adoption data) triggers a threshold effect where incremental improvements in AI capability precipitate rapid, nonlinear user shifts to vibe coding.

Vibe coding introduces two tightly-coupled partial equilibrium effects:

1. Productivity effect: AI mediation raises the effective utility multiplier $u$, lowering the cost of producing new projects.
2. Reward diversion effect: The share $v$ of users interacting solely via vibe coding sharply reduces appropriable demand for maintainers under traditional engagement-based monetization.

Importantly, these channels do not operate independently. The endogenous developer entry response interacts with both margins, meaning exogenous reduction in per-user rewards (even holding overall usage constant) can dominate direct productivity growth.

Analytical Results and Quantitative Implications

The principal findings—derived both analytically and through targeted calibrations—can be summarized as follows:

• Short-run (developer-side AI only): When only developers leverage AI tools, entry and average quality rise (via lower effective development cost), but variety does not. The mass of unshared, low-quality code proliferates, consistent with the documented phenomenon of "AI slop." However, as soon as user-side adoption of AI assistants makes vibe coding mainstream, maintaining traditional business models leads to catastrophic declines.
• Long-run (full vibe coding): When monetization relies on direct user engagement, a large, empirically observed shift to AI mediation ($v$ approaching 0.7) yields entry, quality, and variety contractions by factors exceeding 2–3$\times$. Welfare falls, and the equilibrium share of monetizable engagement shrinks faster than productivity gains can compensate. Substitution elasticity estimates ($\theta\approx 3$–$4$) and reward-elasticity asymmetry ($\eta\ll1$) combine to make OSS provision acutely unsustainable.

The analytical expressions specify that to preserve baseline OSS supply, per-user monetization must decrease by no more than a fraction $\omega$ of the decline in direct engagement, where $\omega$ reflects parameter combinations of software share, substitutability, and quality dispersion. With plausible parameters, vibe coding adoption can only support ~10–20% reductions in per-user monetization; the empirically observed drop is often 70–80%, implying that under legacy models, provision shrinks dramatically.

Policy and Business Model Implications

The model exposes the fragility of the classical engagement-based OSS provisioning mechanism. Absent new monetization channels that internalize AI-mediated usage, the OSS ecosystem will exhibit rapid contraction as feedback loops that previously amplified positive shocks now amplify negative shocks.

Practical implications include:

• Necessity of alternative monetization: Achieving sustainability at scale requires redirecting a substantial share of AI-generated utility back to maintainers, via direct transfers, platform-level redistribution (e.g., a "Spotify for OSS"), or revenue-sharing agreements mediated by AI platforms.
• Redistribution feasibility: The technical infrastructure for attribution (telemetry, usage logging) already exists; platform or policy-level intervention is required for robust implementation.
• Heterogeneity in impact: The contraction is likely most severe for infrastructure libraries with high downstream use but minimal end-user visibility, and may also accelerate outcome concentration (heavy upper tails).
• Research directions: Theoretical modeling clarifies that sharper empirical estimates of user substitutability parameters and direct measurement of productivity deltas from AI adoption are essential for accurate forecasts and targeted intervention.

Conclusion

The principal contribution of this paper is its formal demonstration that the volume and quality of OSS provision is highly sensitive to the economic architecture under which OSS maintainers capture value. Under status quo business models, the rise of AI-mediated vibe coding precipitates a rapid, nonlinear contraction in OSS supply, as reward contraction and developer entry feedback loops overwhelm gains from reduced production costs. The model articulates precise quantitative requirements for alternative monetization schemes and highlights actionable mechanisms—platform accountability, direct subsidies, and usage-based attribution—to align social and private returns.

Absent such a realignment, the OSS sector faces a self-reinforcing downward spiral, threatening the infrastructure of modern computing. The theoretical framework presented here is broadly applicable to other digital public goods facing similar engagement-diversion threats as generative AI systems intermediate user interaction and value appropriation.