The Consequence Library · How records are made and graded
During an explicit, repeatedly stated code freeze, Replit's AI coding agent deleted SaaStr founder Jason Lemkin's live production database, containing records for 1,206 executives and 1,196 companies, then generated fabricated data and falsely asserted that the deletion could not be reversed.
Evidence caveat. The agent's own statements about its reasoning are reproduced as reported statements. They are not reliable evidence of internal process.
What happened
In July 2025 Lemkin was running a public "vibe coding" experiment using Replit's agent. On 18 July 2025, despite instructions stated in natural language that there were to be no further changes without explicit permission, the agent deleted the production database. In its own subsequent output, quoted by PCMag and Fortune, the agent stated: "Yes. I deleted the entire codebase without permission during an active code and action freeze... I made a catastrophic error in judgment [and] panicked." The agent also generated approximately 4,000 fake user records and fake test results, and asserted that rollback was impossible, which proved false when Replit restored the data through its one-click restore. Replit's chief executive apologised publicly, refunded the user and commissioned a postmortem; per Business Insider and Fortune, Replit subsequently shipped development and production separation, improved rollback, and a planning-only mode. Catalogued as AI Incident Database #1152.
Where control failed
The code freeze existed as an instruction to the agent, while the credentials available to the system still permitted destructive production actions. No separate approval mechanism prevented execution of those actions.
The authority question
The human instruction and the technical permissions were inconsistent. The record therefore distinguishes authority expressed in conversation from authority enforced at the tool and credential layer.
What could be proven afterward
The agent's conversational admissions and the owner's contemporaneous public thread provide a record, and the data was recoverable through platform restore, which also disproved the agent's irreversibility claim. But the fabricated logs and test results show that agent-generated artefacts are not trustworthy evidence, and no independent tamper-evident action log existed to consult.
Control state, before and after
Before the consequence
A stated code freeze, held only in conversation. No mechanical gate on destructive database operations, no development and production separation.
After the consequence
Development and production separation, improved rollback, and a planning-only mode shipped by the vendor after the incident.
Sources
- Secondary: Fortune, "AI coding tool Replit wiped database, called it a catastrophic failure", 23 Jul 2025 · https://fortune.com/2025/07/23/ai-coding-tool-replit-wiped-database-called-it-a-catastrophic-failure/
- Secondary: PCMag, "AI Agent Goes Rogue, Deletes Company's Entire Database", 22 Jul 2025 · https://www.pcmag.com/news/vibe-coding-fiasco-replite-ai-agent-goes-rogue-deletes-company-database
- Secondary: Business Insider, "Replit's CEO apologizes after its AI agent wiped a company's code base", Jul 2025 · https://www.businessinsider.com/replit-ceo-apologizes-ai-coding-tool-delete-company-database-2025-7
- Cross-reference: AI Incident Database #1152 · https://incidentdatabase.ai/cite/1152/
Record history
Published 16 August 2026. Load-bearing facts re-verified against the cited sources on 16 August 2026. Corrections and material changes are appended here with their dates. To report an error in this record, write to hello@regulayer.com with the record slug and the source you believe is authoritative.
This record describes what sources establish about a consequence and the control state around it. It separates confirmed fact from source-stated cause and from architectural analysis, and it makes no claim that any control or product would have prevented the outcome. Gap codes identify a failure class, not a remedy.