Vibe Fuzzing: Building an AI-Guided Fuzzing Engine from Scratch | Sam Lovejoy | OAIC 2026

Sam Lovejoy of IBM X-Force Offensive Research presents Vibe Fuzzing at Offensive AI Con 2026. Sam explains how VF3 combines graph-based input generation, state-aware fuzzing, and LLM guidance to target code and patches while keeping the fuzzing loop fast. The talk covers Markov-chain grammars, program state, model context, performance and cost, followed by a live demonstration and audience Q&A. This edit uses Sam’s updated post-conference slides. Slides and project materials: https://github.com/dinosaurlover38/oa... Watch on OAIC: https://www.offensiveaicon.com/vibe-f... Offensive AI Con: https://www.offensiveaicon.com The live demonstration section retains the speaker camera and original explanation. Chapters 00:00 OAIC intro 00:05 Introduction 01:04 Why classical fuzzing needs to change 05:09 Vibe Fuzzing and subagents 06:44 The data plane 09:10 Program state and state machines 13:56 The control plane 15:00 When to invoke AI 16:42 Targeting code and patches 18:34 Beyond memory corruption 19:40 Performance and cost 21:06 Live demo — speaker camera 25:13 Recap 26:50 Audience Q&A #OAIC #OffensiveAI #Fuzzing