tool · [1 source] · 2026-05-23 08:17 · Français(FR) Un papier arXiv récent a étudié 302 600 commits vérifiés comme écrits par IA dans 6 299 repos GitHub. Il trouve 484 366 problèmes distincts par analyse statique

AI-generated code on GitHub shows persistent issues, study finds

By PulseAugur Editorial · Summary by gemini-2.5-flash-lite from 1 source

A recent arXiv paper analyzed over 300,000 commits identified as AI-generated across nearly 6,300 GitHub repositories. The study found more than 480,000 distinct issues through static analysis, with a significant portion persisting in the latest code versions. This suggests that while AI lowers the barrier to producing plausible code, the bottleneck is shifting towards human code qualification and review. AI

Summary written by gemini-2.5-flash-lite from 1 source. How we write summaries →

IMPACT Highlights potential challenges in qualifying AI-generated code, shifting focus to human review.

RANK_REASON The cluster describes findings from an academic paper published on arXiv. [lever_c_demoted from research: ic=1 ai=1.0]

Read on Mastodon — fosstodon.org →

arXiv
GitHub

paper
other

COVERAGE [1]

Mastodon — fosstodon.org TIER_1 Français(FR) · [email protected] · 2026-05-23 08:17

A recent arXiv paper studied 302,600 verified commits as AI-written in 6,299 GitHub repos. It finds 484,366 distinct issues by static analysis

Un papier arXiv récent a étudié 302 600 commits vérifiés comme écrits par IA dans 6 299 repos GitHub. Il trouve 484 366 problèmes distincts par analyse statique, et 22,7% des problèmes suivis survivent jusqu'à la version la plus récente. Ça ne veut pas dire "l'IA écrit du mauvais…

COVERAGE [1]

A recent arXiv paper studied 302,600 verified commits as AI-written in 6,299 GitHub repos. It finds 484,366 distinct issues by static analysis

RELATED ENTITIES

RELATED TOPICS