Learn · a trick with a name
Rented Halo
the achievement is real, the drama around it is borrowed.
A real achievement gets wrapped in drama it didn't earn, and the drama is what spreads.
How to spot it
- A modest result described with world-changing language
- Big names or big fears attached to a small finding
- The paper's own conclusion is calmer than the coverage
The one question to ask“What exactly was achieved, stripped of the story?”
At work
Read the abstract or the original post before sharing. Share the finding, not the halo.
Caught in the wild
Every time we've caught it so far. Try calling each one before you open it.
- TRUE, BUTMeta's own statement named the misconfiguration in sentence one. The headlines said the AI went rogue anyway.
- TRUE, BUTSony Music and Warner Chappell are suing Anthropic, personally naming CEO Dario Amodei, calling it one of the largest thefts of intellectual property in history. Their best evidence is book piracy Anthropic already paid 1.5 billion dollars to settle.
- TRUE, BUTAI is building AI, the headlines say. So independent researchers handed frontier agents real, unpublished research questions and six days each. The agents did all of the engineering and wrote up the results. The papers' own authors rejected both. One got a Strong Reject.
- TRUE, BUTClaude hit 14 of 15 protein targets, and outside labs confirmed it. Then read the method: Claude drove the specialist design tools the field already ships, and a binder is the first step of a drug, not the drug.
- TRUE, BUTClaude moved a bound that had not moved in years: 41.6 to 67.2. Journal reviews of the paper so far: zero.
- TRUE, BUTOpenAI's hacking model found two real bugs in Chrome. The word zero-day got added in post.
- TRUE, BUTYes, researchers pulled the hidden reasoning out of Claude, GPT, and Gemini. No, it was not the token counts, and it is already fixed.
- TRUE, BUTMusk says SpaceX AI revenue will pass everything else by September. This time the revenue is actually real.
- TRUE, BUT1,200 AI insiders signed a letter to slow down AI. Their bosses signed it too, by lunch.
- TRUE, BUTOpenAI solved ten unsolved math problems. One detail decides what that is worth.