The Kernel of Truth Natarajan Shankar, SRI International Computer Science Laboratory Recent advances in automated reasoning have yielded a range of powerful tools for verification, synthesis, and test generation. On the one hand, we would like verification tools to feature powerful automation, but on the other hand, we also want to be able to trust the results with a high degree of confidence. The question of trust in verification tools has been debated for a long time. One popular way of achieving trust in verification tools is through proof generation. However, proof generation could hamstring both the functionality and the efficiency of the automation that can be built into these tools. Another approach to trust is through formally verifying the verification tools, but this is impractical in most cases. We argue that trust need not be achieved at the expense of automation, and outline a lightweight approach called the Kernel of Truth (KoT) where the results of untrusted verifiers are checked by means of verified offline checkers. The KoT approach covers the spectrum of possibilities from verified verifiers to proof generation. We describe the Kernel of Truth architecture along with a few preliminary results, and enumerate some of the practical challenges.