pith. sign in

arxiv: 2504.03581 · v2 · submitted 2025-04-04 · 💰 econ.GN · cs.CY· q-fin.EC

Using digital traces to analyze software work: skills, careers and programming languages

Xiangnan Feng , Johannes Wachs , Simone Daniotti , Frank Neffke This is my paper

Pith reviewed 2026-05-22 21:23 UTC · model grok-4.3

classification 💰 econ.GN cs.CYq-fin.EC
keywords software skillsprogramming languagesPythonskill spacerelated diversificationStack Overflowhuman capitalsoftware development
0
0 comments X

The pith

Programmers using Python preferentially acquire higher-value skills, helping explain the language's rise as a general-purpose tool.

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

This paper draws on tens of millions of Stack Overflow posts to build a map of software skills and their relations. It shows that real software jobs call for tightly linked skill combinations and that developers typically add new skills through paths of related diversification, which often point toward lower-value areas. Python breaks this pattern because its users more frequently move toward higher-value skills. If accurate, the finding supplies one reason why Python has spread so widely across different kinds of software work.

Core claim

By analyzing tens of millions of Question and Answer posts on Stack Overflow, the authors construct a software skill space that maps relations among skills. Real-world software jobs demand highly coherent skill sets and programmers learn through a process of related diversification. The latter process often leads to the acquisition of lower-value skills. However, when programmers use Python they preferentially target higher-value skills, offering a potential explanation for Python's successful rise as a dominant general purpose language.

What carries the argument

The software skill space, a map of relations among skills extracted from Stack Overflow posts that assigns value based on usage patterns and reveals clusters of coherent skill sets.

Load-bearing premise

The values and relationships assigned to skills from Stack Overflow posts match the actual requirements of software jobs and the real paths programmers follow when learning.

What would settle it

Job advertisement or employment records that show Python users do not shift toward higher-value skills at higher rates than users of other languages would undermine the proposed explanation.

Figures

Figures reproduced from arXiv: 2504.03581 by Frank Neffke, Johannes Wachs, Simone Daniotti, Xiangnan Feng.

Figure 1
Figure 1. Figure 1: Mapping software tasks. a. Stylized depiction of the bipartite question-tag network. SBM groups tags into communities (tasks) that connect to similar sets of questions. ChatGPT￾4.0 finds a common label that summarizes each community’s tag information. b. Task space. Pointwise mutual information (PMI) expresses how surprisingly often two tasks are performed by the same users. UMAP embeds the resulting co-oc… view at source ↗
Figure 2
Figure 2. Figure 2: Job ads. a. Schematic representation of the workflow to extract salary and task re￾quirements from online job ads by prompting ChatGPT. Task requirements are converted to the 237-dimensional SO task vectors of [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Task dynamics. a. Task user-share change from 2009 to 2022. Purple markers sig￾nal increases, orange markers decreases, in user-shares between 2009 and 2022. Marker trans￾parency reflects the size of shifts: darker tones indicate larger changes in user shares. b. Esti￾mated probability of diversifying into new tasks at different values of density, users’ relatedness￾weighted experience in other tasks: dθ,u… view at source ↗
Figure 4
Figure 4. Figure 4: Programming languages. a. Task-language matrix. Elements are colored when at least 10 SO users have at least one answer post in the task-language combination. b. Number of tasks in which a programming language ranks as the top language in terms of SO users. The graph shows time-series for the largest six languages in terms of cumulative SO posts between August 2008 and June 2023. c. Python’s task footprint… view at source ↗
read the original abstract

Recent waves of technological transformation are reshaping work in uncertain and hard-to-predict ways. However, jobs at the forefront of the digitizing economy offer an early glimpse of these changes and leave rich activity traces. We exploit such traces in tens of millions of Question and Answer posts on Stack Overflow for the creation of a fine-grained taxonomy of software skills to analyze human capital in the global software industry. Constructing a software skill space that maps relations among these skills reveals that real-world software jobs demand highly coherent skill sets and that programmers learn through a process of related diversification. The latter process often leads to the acquisition of lower-value skills. However, when programmers use Python they preferentially target higher-value skills, offering a potential explanation for Python's successful rise as a dominant general purpose language.

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 paper uses tens of millions of Stack Overflow Q&A posts to build a fine-grained taxonomy of software skills and a skill space that maps their relations. It reports that real-world software jobs require highly coherent skill sets, that programmers acquire skills via related diversification (often into lower-value skills), and that Python use is associated with preferential targeting of higher-value skills, providing a potential explanation for Python's rise as a dominant language.

Significance. If the skill-value metric and space construction hold, the work supplies a large-scale, digital-trace empirical mapping of human-capital dynamics in software work, with implications for labor economics, education policy, and explanations of technology adoption. The scale of the data and the focus on coherence and diversification trajectories are strengths; however, the absence of external validation for the value ordering limits the strength of the causal-style claims about Python.

major comments (2)
  1. [Methods / skill-value construction] Section on skill-value construction (likely §3 or §4): the assignment of value to skills is derived entirely from patterns internal to the SO corpus (co-occurrence, question volume, answer quality). No external validation against labor-market data (e.g., wage returns, job-posting requirements from other sources) is reported. This is load-bearing for the headline claim that Python users target higher-value skills, because any SO-specific bias (over-representation of web frameworks, under-representation of enterprise systems) would mechanically generate the reported differential.
  2. [Results / diversification and Python] Results on related diversification and Python effect (likely §5): the claim that diversification 'often leads to the acquisition of lower-value skills' and that Python reverses this pattern requires explicit robustness checks against experience, tenure, or selection into Python use. Without these, the observed association could reflect unobserved heterogeneity rather than a language-specific learning trajectory.
minor comments (2)
  1. [Data / taxonomy construction] Clarify the exact number of skills in the final taxonomy and the threshold used for inclusion.
  2. [Figures] Figure legends for the skill-space visualizations should list the top skills by value and by degree to aid interpretation.

Simulated Author's Rebuttal

2 responses · 1 unresolved

We thank the referee for their constructive comments on our manuscript. We address each of the major comments below and indicate the revisions we plan to make.

read point-by-point responses

  1. Referee: [Methods / skill-value construction] Section on skill-value construction (likely §3 or §4): the assignment of value to skills is derived entirely from patterns internal to the SO corpus (co-occurrence, question volume, answer quality). No external validation against labor-market data (e.g., wage returns, job-posting requirements from other sources) is reported. This is load-bearing for the headline claim that Python users target higher-value skills, because any SO-specific bias (over-representation of web frameworks, under-representation of enterprise systems) would mechanically generate the reported differential.

    Authors: The skill-value metric is intentionally constructed from patterns within the Stack Overflow data to reflect the digital traces of software work as captured on the platform. We recognize that this approach may introduce biases specific to SO's user base and content focus. In the revised version, we will add explicit discussion of these limitations in the methods section and explore opportunities for external validation using publicly available job market statistics or skill demand reports from other sources. We note, however, that linking to individual-level wage data is not feasible with the available data. revision: partial

  2. Referee: [Results / diversification and Python] Results on related diversification and Python effect (likely §5): the claim that diversification 'often leads to the acquisition of lower-value skills' and that Python reverses this pattern requires explicit robustness checks against experience, tenure, or selection into Python use. Without these, the observed association could reflect unobserved heterogeneity rather than a language-specific learning trajectory.

    Authors: We agree that robustness to user characteristics is important. The Stack Overflow data includes user activity histories that allow us to measure tenure and experience. In the revision, we will add robustness checks that control for these factors as well as potential selection effects into Python usage. This will strengthen the evidence for the language-specific trajectory. revision: yes

standing simulated objections not resolved
  • Full external validation of the skill-value metric against wage returns or comprehensive labor-market data from non-SO sources, as such linkages are not possible with the current dataset.

Circularity Check

0 steps flagged

Empirical data-driven construction of skill space from Stack Overflow traces shows no derivation circularity

full rationale

The paper constructs a fine-grained taxonomy of software skills and a skill space directly from patterns in tens of millions of Stack Overflow Q&A posts, then reports observational findings on job skill coherence, related diversification trajectories, and differential skill targeting by Python users. These are descriptive results from the observed data rather than any mathematical derivation, fitted parameter, or self-referential definition that reduces a claimed prediction back to its inputs by construction. No equations, uniqueness theorems, or ansatzes are invoked in a load-bearing way. The analysis is self-contained as an empirical mapping of platform activity traces, consistent with the reader's assessment of score 2.0 and absence of reducing equations.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The analysis rests on the premise that Stack Overflow activity is a valid proxy for software skills and career trajectories; no free parameters or invented entities are identifiable from the abstract.

axioms (1)
  • domain assumption Stack Overflow Q&A posts reflect actual software skills used in real jobs and the process by which programmers acquire new skills
    The entire taxonomy, skill space, and diversification analysis is built from these posts.

pith-pipeline@v0.9.0 · 5669 in / 1159 out tokens · 55220 ms · 2026-05-22T21:23:01.764088+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.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Library Hallucinations in LLM-Generated Code: A Risk Analysis Grounded in Developer Queries

    cs.SE 2025-09 unverdicted novelty 7.0

    A study of seven LLMs finds that realistic prompt variations such as one-character misspellings trigger library hallucinations in up to 26% of cases, fabricated names in up to 99%, and time-based prompts in up to 85%,...

Reference graph

Works this paper leans on

237 extracted references · 237 canonical work pages · cited by 1 Pith paper

  1. [1]

    Develop responsive WordPress themes with custom plugins and API integrations

    work page

  2. [2]

    Implement form submission with input validation

    work page

  3. [3]

    Configure web servers and manage URL redirections

    work page

  4. [4]

    Develop a Facebook integrated web application using CodeIgniter

    work page

  5. [5]

    Develop a web application using Symfony framework and Doctrine ORM

    work page

  6. [6]

    Develop a responsive email service for dynamic content distribution

    work page

  7. [7]

    Implement session management with cookies in a web application

    work page

  8. [8]

    Develop e-commerce platforms with YouTube and Google API integrations

    work page

  9. [9]

    Integrate various payment gateways in an e -commerce platform

    work page

  10. [10]

    Setup and manage a web server environment

    work page

  11. [11]

    Develop a Joomla website with custom extensions and SEO optimization

    work page

  12. [12]

    Set up a local development environment for web applications Web Design

    work page

  13. [13]

    Develop a scalable, partitioned web application with Font Awesome icons

    work page

  14. [14]

    Develop a Nuxt.js app with Chrome extension integration

    work page

  15. [15]

    Develop a cross-browser compatible web animation solution

    work page

  16. [16]

    Generate interactive charts and reports in web applications

    work page

  17. [17]

    Implement a responsive iframe embed for Vimeo videos

    work page

  18. [18]

    Develop web applications integrating communication APIs and interactive notebooks

    work page

  19. [19]

    Develop interactive graphics for web applications

    work page

  20. [20]

    Implement a dynamic web form with various input elements

    work page

  21. [21]

    Develop interactive, cross-browser web applications

    work page

  22. [22]

    Develop responsive UI with modern CSS frameworks and libraries

    work page

  23. [23]

    Implement a responsive sticky header with parallax scrolling effect

    work page

  24. [24]

    Develop a Mozilla Firefox browser extension

    work page

  25. [25]

    Develop a Shopify app with automated BDD testing

    work page

  26. [26]

    Develop dynamic web interfaces in JSF and Spring Webflow

    work page

  27. [27]

    Develop a feature-rich text editor for web content

    work page

  28. [28]

    Implement OCR feature using Tesseract and integrate with Azure

    work page

  29. [29]

    Develop a web application using Polymer and Eclipse

    work page

  30. [30]

    Develop an image gallery using various UI frameworks and sliders

    work page

  31. [31]

    Develop a responsive web app with customized user authentication

    work page

  32. [32]

    Develop a responsive web layout with dynamic content positioning

    work page

  33. [33]

    Implement a dynamic PDF report generator for web content IOS

    work page

  34. [34]

    Develop and optimize an iOS application user interface

    work page

  35. [35]

    Develop a macOS robotics simulation app

    work page

  36. [36]

    Implement and optimize testing strategies for mobile applications

    work page

  37. [37]

    Develop a mobile app with modern networking and UI frameworks

    work page

  38. [38]

    Implement multilingual support for a global web application

    work page

  39. [39]

    Develop a mobile app with offline data synchronization capabilities

    work page

  40. [40]

    Develop iOS keyboard extension with custom click and mouse support

    work page

  41. [41]

    Develop a web application using GWT and Parse platform

    work page

  42. [42]

    Implement a compressed notification system for mobile and web

    work page

  43. [43]

    Implement a mutable, anonymous delegate using pass -by- reference in C#

    work page

  44. [44]

    Implement a location-based service application using MapKit

    work page

  45. [45]

    Implement a visual effects processing application

    work page

  46. [46]

    Develop a cross-platform audio-visual media application. Android

    work page

  47. [47]

    Transform and process XML documents using different technologies

    work page

  48. [48]

    Set up Android Studio project with Gradle build system

    work page

  49. [49]

    Optimize Unity game performance and physics simulation

    work page

  50. [50]

    Develop a mobile app with interactive geospatial features

    work page

  51. [51]

    Develop an Android app with modern UI/UX components

    work page

  52. [52]

    Implement efficient data display in Android applications

    work page

  53. [53]

    Implement geometry transformation algorithms and visualize results

    work page

  54. [54]

    Develop a location-based mobile application

    work page

  55. [55]

    Develop VoIP communication system with audio streaming capabilities

    work page

  56. [56]

    Develop a WebRTC-based live video streaming platform

    work page

  57. [57]

    Develop a cross-platform mobile app with Xamarin and MVVMCross

    work page

  58. [58]

    Develop a modern Android app with navigation and data management

    work page

  59. [59]

    Develop a cross-platform mobile app with cloud integration

    work page

  60. [60]

    Develop a secure contact management application for Android

    work page

  61. [61]

    Develop a countdown timer using MediaWiki's Wikipedia API

    work page

  62. [62]

    Develop a color drawing app with camera support

    work page

  63. [63]

    Capture and analyze screen resolution details

    work page

  64. [64]

    Implement and manage push notification services for mobile apps

    work page

  65. [65]

    Develop a responsive Android app with Material Design

    work page

  66. [66]

    Develop a Bluetooth-enabled AR/VR application

    work page

  67. [67]

    Develop Android media player with custom controls

    work page

  68. [68]

    Implement efficient image processing using GPU acceleration

    work page

  69. [69]

    Develop a 3D mobile game using rendering and physics libraries

    work page

  70. [70]

    Develop a cross-platform mobile app with sensor integration

    work page

  71. [71]

    Develop a database-driven application with API and real -time updates

    work page

  72. [72]

    Develop an interactive Android app with custom gesture controls

    work page

  73. [73]

    Develop a mobile app featuring voice-controlled media playback

    work page

  74. [74]

    Develop a reporting application with interactive graphics and mobile support

    work page

  75. [75]

    Develop a custom video player for Android with fullscreen feature

    work page

  76. [76]

    Advanced Programming Concepts

    Develop a distributed computing system on IBM Cloud using Dask. Advanced Programming Concepts

    work page

  77. [77]

    Develop a cross-platform application with Qt framework

    work page

  78. [78]

    Implement a file compression and decompression utility

    work page

  79. [79]

    Develop a 3D facial recognition system

    work page

  80. [80]

    Develop a data structure library for efficient data manipulation

    work page

Showing first 80 references.