What Hamming Offers

Hamming is the execution engine and result store for voice tests: it drives the caller side of a real phone call (dials, plays a persona, listens), and it stores test cases, guardrails, and run history. Its grading layer is guardrails: natural-language pass/fail criteria scored by an LLM over the transcript, with settings for when they run and whether they wait for tool-call or webhook data. It also ships a full webhook-driven orchestration framework, Test Hooks, with four lifecycle phases (before run, before case, after case, after run) that can seed state, validate it, and chain responses between hooks. That is a genuinely strong voice-execution product, and rebuilding it would be a multi-quarter distraction with no differentiated payoff. The limits are equally structural: Hamming judges probabilistically, it judges the transcript, and it has no access to Avoca's data, so anything that is a matter of fact or lives outside the conversation is out of reach.