AfterVibe: What Remains When the Conversation Ends
2026-07-10 • Software Engineering
Software Engineering
AI summaryⓘ
The authors created AfterVibe, a tool that takes a coding session and figures out a natural-language description of what the code is supposed to do. Then, they use another AI to write new code from that description to check if the explanation really matches the original intent. If the new code works the same, the description is good; if not, it gets improved. They tested this on many real coding sessions and found the descriptions were clear about what the code should do but left room for different ways to do it. This approach might help focus reviews on the specifications instead of the code itself as AI coding advances.
natural-language specificationLLMcode regenerationvalidation pipelinevibe codingbehavioral intentcode reviewAI code generationspecification refinement
Authors
Matteo Paltenghi, Satish Chandra
Abstract
We present AfterVibe, a framework that recovers natural-language specifications from a vibe coding session. Given a code artifact and the conversation trajectory that produced it, AfterVibe uses an LLM to extract an abstract natural-language specification capturing the developer's intent, and validates it through a regeneration test: a second, blind AI agent re-implements the artifact from the spec alone, and the resulting code is graded against the original through a multi-tier validation pipeline. Spec quality is thus measured by whether an agent can regenerate passing code; if the verifiers deem the implementations equivalent the spec is considered strong, otherwise it is iteratively refined. Evaluating AfterVibe on 72 real-world vibe-coded projects from a company's internal coding sessions, we find that its recovered specs are abstract by design-capturing behavioral intent without dictating implementation-yet strong. Multiple independent regenerations achieve a high mean regeneration score of 5.06 out of 6.0 while remaining diverse in their details, confirming that the spec constrains what without over-prescribing how. Besides outperforming existing human-authored descriptions, the specs can be further strengthened iteratively to a score of 5.74. A practical implication is that specifications-not code-could become the primary artifact for human review and the source of record at a time when AI-generated code is outpacing customary code review.