I installed a git hook that quizzes me on my AI's code before every push. It's… · @ReignSwords · TwKit

@ReignSwords

I installed a git hook that quizzes me on my AI's code before every push. It's called Sphinx. 252 points on Hacker News over the weekend. The pitch: you can outsource your thinking, but not your understanding. Here's what actually happened: 1. Install took 28 seconds. One flag scopes it to a single repo. Good thing, because the default hooks every push on your machine. 2. Pushed 91 lines to main. Instantly intercepted: "understanding check needed." Push frozen. A questionnaire opened on localhost. 3. For the reviewer brain, I tried the cheapest setup possible: a free anonymous API. No key, no signup. It passed the compatibility checks. Tool calls worked. Structured output worked. Then it died on the first real question. The free model thinks out loud so much it burned its entire 1500-token output cap before finishing. I doubled the budget like the docs suggest. Same wall. The cap is on their side, not mine. My takeaway: the hook itself is genuinely well built. The $0 brain is a fantasy. The honest price is about half a cent per push with a cheap model. That's what it costs to make sure you actually read the code. I break tools so you don't have to. Stick around for more honest tests. When's the last time you read every line your agent wrote before hitting merge? 🔵 Fellow Premium members, let's connect! I post about AI & tech — follow for more 👥

曝光 38 · 评论 7 · 点赞 1 · 书签 0