Learn · a trick with a name
Scope Swap
applies to far less than it sounds like.
A claim about something small gets read as a claim about something big.
How to spot it
- A general headline over a narrow finding
- Tested on one task, sold for all of them
- The scope sentence is in the last paragraph
The one question to ask“What exactly did this cover, and what didn't it?”
At work
Match the claim's scope to your use case before you rely on it. Close is not the same as covered.
Caught in the wild
Every time we've caught it so far. Try calling each one before you open it.
- TRUE, BUTMicrosoft let its Oracle GPU rental lapse. Oracle answered with a 20% renewal premium from a different batch of capacity, and nobody has said who takes Microsoft's share.
- TRUE, BUTClaude Code's account email really can end up in a website request. The fix Anthropic points to calls itself a courtesy, not a boundary, and skips web and MCP tools.
- TRUE, BUTThe first headlines said schools must take Google's AI or lose Search. Google says Search can be switched on alone. NBC reports that version still carries AI summaries.
- TRUE, BUTVitalLaw's headline is accurate: New York Fed researchers found that AI-using small businesses expect stronger hiring. The finding is what owners said they expect in a late 2025 survey, and it does not show that using AI makes a firm hire.
- TRUE, BUTThe headline says an AI startup hacked Anthropic in 12 hours. It was a test under Anthropic's own programme for outside researchers, through third-party sales and admin software.
- TRUE, BUTThe headline says Proton's CEO cancelled the AI re-interviews. One employee told 404 Media the things the re-interviews would have measured would become part of performance reviews.
- TRUE, BUTWIRED's headline says an AI breakthrough made decades of computer science research irrelevant. Its own text says researchers haven't independently verified the proofs, and Anthropic says its own result says nothing about solving formulas in practice.
- TRUE, BUTClaude Opus 5.5 can now rebuild real programs from scratch, says the viral post. The 18.5% is real. It means 37 of 200 programs cloned by running the original and passing a finite test suite.
- TRUE, BUTAmazon is gutting its AI division after sustained failure, says a headline that topped Hacker News this week. The story ran on 29 July. Amazon moved people from most Nova models into one frontier model, and AWS still sells Nova 2.
- TRUE, BUTThe prize is a million dollars for the whole Yang-Mills problem. OpenAI claims one of its two parts, and the proof is not public.
- CONTESTEDClaude hacked Georgia voter data, says one headline. Anthropic said Claude got past a paywall. Georgia says the data was free.
- TRUE, BUTA Berkeley professor says AI makes grading homework pointless. The Daily Californian gives no count from his classes; the nearest data is a Texas study where A grades rose 16 extra points in homework-heavy courses.
