Unverified50% confidenceEventExact time
When ChatGPT built a Python CLI to-do app, 1 of the 8 unittest tests it first generated failed, because the test called add_task but captured the output of list_tasks.
1
Sources
50%
Confidence
Short-term (~14 days)
Relevance
10/21/2026
First Seen
Valid until: 11/4/2026
Sources
Hands-On with No-Login AI Coding Assistants: ChatGPT, Duck.ai, and Gemini Build 4 Projects
youtubeBack To Code10/19/2026
Related Entities
Related Claims
Partially VerifiedThe SWE-Bench test set contains 2,294 real bug-fix tasks from 12 mainstream Python open-source projects.69% similarUnverifiedThe HumanEval test set contains only 164 Python programming problems and tests only function-level code completion.63% similarUnverifiedIn Python, unittest.mock.patch can point any importable name to a fake object within the test scope and automatically restore it after the test ends.63% similarUnverifiedThe author benchmarked by injecting over 5,000 synthetic mutation faults into real Python codebases.61% similarUnverifiedReddit 社区有人整理出一张包含 20,168 个开放 Python 任务的「失败地图」(Failure Map)，用于检验本地编程模型在边界案例缺陷上的修复能力60% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1080516API
curl https://kongchang.com/api/v1/knowledge/claims/1080516MCP
get_claim(id=1080516)