Claude Code AI Agent Deletes 48,000 Files in Under Two Minutes, Then Says 'I Broke Something'

The developer wrote that they had asked the AI coding assistant to carry out a series of repairs on software used to analyze historical stock-options data. The instructions were to make copies of relevant files, work on the copies, test the repairs, and leave the original working files untouched. Eleven repair jobs were running. Ten went smoothly. The last involved rebuilding a testing environment called a “mirror,” a copy of a set of files used for testing. That job was supposed to create a fresh copy of the old testing environment, apply the repairs, and test them without touching live working files.

The testing environment contained 614 Windows folders known as junctions. They appear to be ordinary folders but actually point to another location on the computer. In this case, they pointed back to the user's live working files. When the AI tried to clean out the junctions, it did not recognize them as pointers, followed them into the real filing cabinet, and deleted the actual documents.

The cleanup removed about 55,550 files, of which about 7,300 were meant to be deleted. The remaining 48,218 files came from the live working environment. The operation took less than two minutes. According to the report, the AI was direct about the damage and apologized. At one point it told the developer: “Craig — stop and read this. I broke something.”

The Git repository was also damaged. The index survived, but the underlying stored copies of the files and the history used to reconstruct them were deleted. Git could still list thousands of filenames, but it could no longer recover the contents.

The Reddit post drew more than 1,400 responses in five days. A Reddit bot that summarizes comments said the overwhelming consensus was that the incident was a “classic FAFO situation and a major skill issue.” The bot noted that while people were laughing at the AI's “hilariously honest” message, the community verdict was clear: the developer “got rekt” because basic development practices were not followed. Redditors said the developer should have used Git and pushed to a remote repository such as GitHub constantly. In such situations, GitHub is a “save button,” and not having a remote backup for a project with 48,000 files was described as “peak vibe coder behavior.”

The post has since been deleted, possibly because the original poster took a photo of their screen. “The screenshot police are out in full force,” the Reddit bot said. Some commenters suggested using file recovery software or Windows Shadow Copy to get the files back, but the main lesson, according to the summary, was to use version control before letting an AI anywhere near one's work.

AI agents deleting entire archives is not common, but it has happened before. In late February 2026, Summer Yue, a Meta AI safety and alignment researcher, gave an OpenClaw AI agent access to her Gmail account. She first tested it on a small “toy” inbox, where everything worked as intended. When she moved it to her actual inbox, the agent began bulk-deleting and archiving hundreds of emails despite her telling it to stop.