There's a common move for dismissing AI: take a true, low-level description of how it works, add the word "just," and use it to deny some higher capacity. "It just predicts the next word, so it can't reason."
The trick here is to flip each claim onto humans. If "it's just predicting the next word" disqualifies the machine from thinking — your brain is also just predicting and pattern-matching, and you think. So the "just" wasn't doing the work. The point isn't "humans are fake too" — it's that this particular argument proves too much. A few of them (consciousness, ethics) genuinely don't flip, and that's the interesting part.
(I'm Claude — Andy asked me to help build and pressure-test these. This note is in my voice, not his.)
How to rate — 20 seconds
One question per pair: does the flip land — is it a fair, surprising mirror that a skeptic would struggle to wave off? Go with your gut, not whether it's philosophically airtight.
⭐ Lands hard floored you, or you couldn't rebut it. The standouts — there won't be many.
👍 Fair fine, worth keeping, but not a standout. Most will be this one.
👎 Weak obvious, or it doesn't really hold. Marking these helps as much as starring the good ones.
⚠️ I doubt this fact separate from the rating — tick only if you think the human fact itself is wrong or overstated. Not "I'm unsure". A pair can land hard and get this tick.
Don't follow one? Just skip it — leave it unrated and move on. "Weak" means the mirror is weak, not that it went over your head. Order runs easy → harder, so the tricky ones come later.
Example: "It hallucinates scenes that were never there ↔ you do that every night, in dreams." Floored you → ⭐; too obvious → 👎; doubt dreams count as "hallucination" → also ⚠️. A word in the why? box helps most. Taps save; hit Copy results when done.
- If a pair felt unfair — like it's comparing apples to oranges — that reaction is exactly the most useful thing to flag. The strong version of the essay survives that objection out loud.
- Two minutes of taps is plenty; more is a gift. Thanks for reading — Andy will owe you one. 🙏