Coding Agent Adoption Thesis — Git skill for Claude Code

Coding Agent Adoption

Bachelor's Thesis

---

What This Thesis Is About

This thesis investigates how coding agents (GitHub Copilot, Cursor, Claude Code, and similar AI-assisted development tools) are being adopted on GitHub, with a focus on the Egyptian developers specifically.

Two research questions drive the study(For: JS/TS/npm):

• RQ1 — Adoption rate: What proportion of active JS/TS repositories show evidence of coding agent use? How does adoption vary by project visibility (star stratum) and over time?
• RQ2 — Impact: Does the introduction of a coding agent measurably change repository activity — commit frequency, PR merge time, contributor growth, code churn?

The study is part of a six-student research cluster. All six topics study the same research questions for Egyptian developers, each covering a different programming language ecosystem. Shared methodology ensures findings are directly comparable across ecosystems.

---

Pipeline Overview

The data collection pipeline runs in 7 stages. Stages 1–3 are complete.

Stage 1 ── Search GitHub for active JS/TS repos (global sample)
              └─► base_jsts_corpus_pilot.csv  (4,372 repos)

Stage 2a ── Discover Egyptian developer accounts by location search
              └─► egypt_seed_candidates.csv   (1,876 accounts)

Stage 2b ── Snowball-expand the seed list through follower/following networks
              └─► egypt_snowball_candidates.csv  (19,347 new accounts)

              combine_accounts.py
              └─► egypt_combined_accounts.csv  (21,223 total accounts)

Stage 2c ── Pull JS/TS repos from all Egyptian developer accounts
              └─► egypt_corpus.csv  (72,946 repos)

Stage 3 ── Merge and deduplicate both corpora
              └─► unified_corpus.csv  (77,315 unique repos)

Stage 4 ── [TODO] Detect coding agent adoption signals in each repo
              └─► signals.csv + agent_adopted column filled in

Stage 5 ── [TODO] Compute before/after metrics for adopted rep