pith. sign in

arxiv: 2511.06545 · v2 · submitted 2025-11-09 · 💰 econ.GN · cs.CY· q-fin.EC

Vibecoding and Digital Entrepreneurship

Ruiqing Cao , Abhishek Bhatia This is my paper

Pith reviewed 2026-05-18 00:05 UTC · model grok-4.3

classification 💰 econ.GN cs.CYq-fin.EC
keywords generative AIvibecodingdigital entrepreneurshipentrepreneurial entrySTEM educationventure performancecoding automation
0
0 comments X

The pith

Vibecoding increases viable digital entrepreneurial entry by 11% only where it augments technical skills rather than fully automating them.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper studies how generative AI tools for coding, referred to as vibecoding, influence the start and success of digital businesses. These tools lead to more first-time launches and faster launch times overall. However, they boost the number of economically viable new ventures only in product segments where the AI assists rather than takes over the entire development process. This 11% increase in viable entry is driven solely by founders who have STEM education or work experience. The results highlight that human technical expertise continues to matter, with AI best used to handle routine coding so people can tackle bigger challenges.

Core claim

Vibecoding increases first-time launches and shortens time to launch, but economically viable entry rises only where vibecoding augments, rather than fully automates, product development. In these partially exposed product segments, viable entry increases by 11%, driven entirely by ventures founded by individuals with STEM education or work experience, especially those whose most recent employment was outside middle management. Among ventures launched before GenAI became widely accessible, performance gains similarly concentrate among partially exposed ventures with engineering-intensive initial teams.

What carries the argument

Difference-in-differences models exploiting ex-ante variation in ventures' exposure to vibecoding based on the product characteristics of their initial launches.

Load-bearing premise

The variation in exposure to vibecoding across product types is unrelated to other factors that influence entry decisions and performance around the GenAI diffusion period.

What would settle it

No differential rise in viable entry for partially exposed segments after GenAI diffusion, or similar gains among non-STEM founders, would show that benefits do not hinge on complementarity with technical expertise.

read the original abstract

As generative artificial intelligence (GenAI) automates coding tasks and expands access to technical resources, this paper examines how GenAI-enabled coding automation, colloquially known as "vibecoding," affects digital entrepreneurial entry and venture performance. We exploit ex-ante variation in ventures' exposure to vibecoding based on the product characteristics of their initial launches and estimate difference-in-differences models around the diffusion of GenAI coding tools. Vibecoding increases first-time launches and shortens time to launch, but economically viable entry rises only where vibecoding augments, rather than fully automates, product development. In these partially exposed product segments, viable entry increases by 11%, driven entirely by ventures founded by individuals with STEM education or work experience, especially those whose most recent employment was outside middle management. Among ventures launched before GenAI became widely accessible, performance gains similarly concentrate among partially exposed ventures with engineering-intensive initial teams. Together, these results suggest that GenAI-enabled coding automation does not eliminate the value of technical expertise. Instead, vibecoding creates the greatest value when it complements internal engineering capabilities, allowing ventures to delegate lower-level coding tasks to GenAI while shifting human effort toward higher-level problem solving and dynamic adaptation.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The manuscript examines the effects of generative AI-enabled coding automation ('vibecoding') on digital entrepreneurial entry and venture performance. Using difference-in-differences around the diffusion of GenAI tools and exploiting ex-ante variation in exposure based on initial product characteristics, it finds that vibecoding boosts first-time launches and reduces time to launch. However, increases in economically viable entry (by 11%) occur only in partially exposed segments where AI augments rather than replaces development, and this effect is driven by founders with STEM education or experience. Performance improvements are concentrated in engineering-intensive teams. The results suggest that vibecoding complements internal engineering capabilities rather than substituting for them.

Significance. If the identification holds, the paper offers timely evidence on the boundaries of AI automation in entrepreneurship, showing that coding tools create value primarily through complementarity with technical human capital rather than broad substitution. This contributes to labor economics and innovation studies by documenting heterogeneous effects by founder background and product type, with potential implications for how AI diffusion shapes entry barriers and the returns to STEM skills.

major comments (2)
  1. [Empirical Strategy] The DiD identification in the empirical strategy section relies on ex-ante product characteristics of initial launches being exogenous to founder selection and pre-GenAI trends. However, the concentration of the 11% viable-entry effect among STEM-educated founders raises the possibility that these characteristics proxy for differential selection or anticipated demand shocks around late 2022, which could confound the complementarity interpretation.
  2. [Results] The abstract and results presentation provide no details on parallel trends tests, robustness checks, data sources, sample construction, or the precise measurement of product exposure. These omissions are load-bearing for assessing the reliability of the central 11% estimate and its restriction to partially exposed segments.
minor comments (2)
  1. [Introduction] The colloquial term 'vibecoding' is used throughout but would benefit from a more formal definition in the introduction to improve accessibility for readers outside the immediate subfield.
  2. [Tables and Figures] Table and figure captions could more explicitly link the reported estimates to the partial-exposure subsample to clarify how the 11% figure is derived.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We are grateful to the referee for providing detailed and insightful comments on our paper. We address each of the major comments below and have made revisions to strengthen the manuscript where appropriate.

read point-by-point responses

  1. Referee: [Empirical Strategy] The DiD identification in the empirical strategy section relies on ex-ante product characteristics of initial launches being exogenous to founder selection and pre-GenAI trends. However, the concentration of the 11% viable-entry effect among STEM-educated founders raises the possibility that these characteristics proxy for differential selection or anticipated demand shocks around late 2022, which could confound the complementarity interpretation.

    Authors: We appreciate the referee highlighting this potential concern with our identification strategy. The product characteristics used to define exposure are based on the initial product launches of ventures prior to the diffusion of GenAI coding tools, which occurred around late 2022. This timing helps ensure that the characteristics are not influenced by post-GenAI selection. Furthermore, the heterogeneous effects by founder STEM background are consistent with our complementarity interpretation, as the effect is absent for non-STEM founders. To bolster this, we have added additional robustness checks in the revised manuscript, including tests for differential pre-trends by founder type and correlations between product characteristics and pre-2022 demand indicators. We believe these additions address the concern without altering the core findings. revision: partial

  2. Referee: [Results] The abstract and results presentation provide no details on parallel trends tests, robustness checks, data sources, sample construction, or the precise measurement of product exposure. These omissions are load-bearing for assessing the reliability of the central 11% estimate and its restriction to partially exposed segments.

    Authors: We agree with the referee that greater transparency on these methodological details is essential. In the revised version of the manuscript, we have expanded the empirical strategy and results sections to include: (1) parallel trends tests with graphical and statistical evidence supporting the assumption; (2) a comprehensive set of robustness checks, including alternative specifications and placebo tests; (3) detailed descriptions of the data sources used; (4) step-by-step sample construction criteria; and (5) precise definitions and measurement of product exposure based on ex-ante characteristics. These details have been added to the main text where space permits and to the appendix for completeness. We believe this will allow readers to better evaluate the reliability of our estimates. revision: yes

Circularity Check

0 steps flagged

Purely empirical DiD design with no derivation chain or fitted inputs

full rationale

The paper is an empirical study using difference-in-differences around GenAI diffusion, exploiting ex-ante variation in product characteristics of initial launches to classify exposure. No mathematical derivations, self-definitional equations, fitted parameters renamed as predictions, or load-bearing self-citations appear in the abstract or described methods. Central estimates (11% viable entry increase, STEM concentration) are data-driven from external event variation rather than constructed by definition or prior author results. This matches the default expectation of no significant circularity for self-contained empirical work.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The analysis rests on standard econometric assumptions for causal identification in observational data around a technology diffusion event.

axioms (1)
  • domain assumption Difference-in-differences identifies causal effects under the parallel trends assumption
    Invoked by the use of DiD models around GenAI diffusion timing.

pith-pipeline@v0.9.0 · 5515 in / 1186 out tokens · 45219 ms · 2026-05-18T00:05:59.410853+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

6 extracted references · 6 canonical work pages · 1 internal anchor

  1. [1]

    engines of growth

    Agrawal, A., Gans, J., and Goldfarb, A. (2022).Prediction machines, updated and expanded: The simple economics of artificial intelligence. Harvard Business Press. Aral, S., Brynjolfsson, E., and Wu, L. (2012). Three-way complementarities: Perfor- mance pay, human resource analytics, and information technology.Management Science, 58(5):913–931. Asam, D. an...

    work page arXiv 2022

  2. [2]

    what is an opportunity?

    Dushnitsky, G. and Stroube, B. K. (2021). Low-code entrepreneurship: Shopify and the alternative path to growth.Journal of Business Venturing Insights, 16:e00251. Eloundou, T., Manning, S., Mishkin, P., and Rock, D. (2024). Gpts are gpts: Labor market impact potential of llms.Science, 384(6702):1306–1308. Ewens, M., Nanda, R., and Rhodes-Kropf, M. (2018)....

    work page internal anchor Pith review arXiv 2021

  3. [3]

    Schulz, C., Bendig, D., Kriebel, J., Haubner, K., and Fainshmidt, S. (2025). Ceo human capital and digital product innovation: A dynamic managerial capabilities perspective. Information Systems Research. Shane, S. and Venkataraman, S. (2000). The promise of entrepreneurship as a field of research.Academy of management review, 25(1):217–226. 36 Tambe, P. (...

    work page 2025

  4. [4]

    Start with AI

    The definitions of managerial and computer occupations, advanced business degree, and STEM degree can be found in Appendix Section C. Robust standard errors are clustered at the category level. *p <0.1, **p <0.05, ***p <0.01 41 Table 4: Effects of GenAI on Venture Performance Across Founder Expertise Dep. Var.1(Raised VC Funding) (1) (2) (3) (4) (5) Treat...

    work page 2022

  5. [5]

    Robust standard errors are clustered at the category level

    or male (column 6), and whether the primary founder is located in North America (column 7), Europe (column 8), or Asia (column 9). Robust standard errors are clustered at the category level. *p <0.1, **p <0.05, ***p <0.01 45 B Sampling Criteria and Final Data Our final dataset is constructed by matching all launch posts on Product Hunt with makers’ Linked...

    work page 2021

  6. [6]

    makers” listed in the Product Hunt post for a venture’s first public launch. It captures the founding team size, as the “makers

    Each venture in the final sample belongs to one of 85 distinct categories. Time to Launch (Months).This variable measures the number of months between the website’s creation (see the Data Appendix 10 for measurement details) and the venture’s first public launch on Product Hunt. It serves as a proxy for the time required to develop the initial product fro...

    work page 2025