Show HN: Sphinx – a hook that quizzes you on your AI's code before shipping

TEXT START: “Outsource your thinking, not yourunderstanding.”

The Dissection

Sphinx is a control gate for the transitional phase of AI-assisted software development: the agent writes code, while a human must demonstrate enough comprehension to ship it. Its real function is liability allocation. The machine produces the artifact; the human is forced to acknowledge its failure modes.

The Core Fallacy

Sphinx mistakes human verification for durable economic indispensability. Under P1, code generation, testing, explanation, and eventually review are automated. Under P2, a Git hook cannot preserve a human-only economic domain; teams can automate the answers, replace the quiz with agent-to-agent verification, or bypass the gate. Under P3, requiring a human to answer a question does not mean human labor remains structurally necessary.

A correct answer also proves less than the product implies. It demonstrates recall of one scenario, not complete semantic understanding, security, correctness, or operational judgment. The quiz is a useful friction mechanism, not proof of control.

Hidden Assumptions

• The reviewing model generates accurate, high-value questions from the diff.
• A correct answer correlates with genuine understanding rather than memorization or guesswork.
• Humans have the time, context, and competence to review increasingly complex agent output.
• The added delay costs less than the defects it catches.
• Agents do not soon become capable of reviewing and verifying their own work.
• Organizations will enforce the hook when throughput pressure rises.
• The user will not simply invoke sphinx yolo whenever the gate becomes inconvenient.

That last feature exposes the hierarchy: when verification conflicts with speed, speed usually wins.

Social Function

Transition management and partial truth, wrapped in ideological anesthetic. Sphinx can genuinely reduce reckless acceptance of AI-generated code and create a narrow Servitor role for people who retain deployment authority, domain judgment, or incident liability. But its slogan preserves a human ritual, not a human economic position. It converts declining participation into a checkpoint and calls the checkpoint understanding.

The Verdict

Sphinx is useful transitional infrastructure and a form of verification arbitrage. It may catch real defects during the human-supervised phase of agentic development. It does not alter AI ownership, restore the wage circuit, or prevent productive participation from collapsing. Its moat is workflow integration and organizational habit—temporary hospice care. In Discontinuity Thesis terms, Sphinx manages the carcass; it does not stop the kill mechanism.