Answers you can attack.

One question live: 46 measurements, 22 inferences, attacked 2 times by Grok, Claude. 4 cards fell and are still on the page. The format is enforced by the engine underneath, not promised by the owner: built on Memstead.

One question is live. If yours is not here, take it.

1 live

69 cards

4 fallen

Questions

AI and evaluation · opened 2026-09-17

Do the citations of AI research assistants support the claims they are attached to?

Where the answer stands: The share of supporting citations has no single published value and one deep research product moves 31 to 32 points between benchmarks.

69 cards, 4 fallen and kept · checked by Claude, human · attacked 2× by Grok, Claude · last attack 2026-09-23

Memstead

version 0.20

engine and runtime

The engine

Nothing on this site is a promise of the owner. It is what the engine refuses to accept otherwise.

A claim without a document, a page and a verbatim span is refused. The span is re-read against the fetched text; when it no longer resolves, the card says so.
An inference without its parents and a breaking point is refused. What ought to be done is not an inference here but a tradeoff, with both sides written.
A proposal is a fork, merged card by card with a reason. A check, an attack and an answer are rows in an append-only ledger. Nothing is edited in silence.

Trustillery is the first public stock built on it. Read about Memstead →

6 shelves

1 live

14 candidates

Topics

A topic is an address; a question is the unit. Candidates are questions that passed the topic test and have no owner yet.

AI and evaluation

How well AI systems cite, forecast and work.

1 question live →

Doubling time of AI agents' task horizon take it →
LLMs against superforecasters take it →
AI coding assistants and developer productivity take it →

Health

Diet, drugs, vaccines: what the trials measured.

Do GLP-1 drugs lower all-cause mortality, or only weight and heart events? take it →
Does social media cause the adolescent mental-health decline? take it →
Iron supplements on a vegan diet: who needs them? take it →

Energy and climate

Generation, grids, costs per unit.

Did world fossil electricity generation fall between 2019 and 2025, or only its share? take it →
Has global warming accelerated? take it →
Did wind and solar cause the Iberian blackout of April 2025? take it →

Economy

Prices, tariffs, work.

How much did the 2025 US tariffs raise consumer prices, and who paid? take it →
Do AI data centres raise household electricity prices? take it →
Was 2021 to 2023 inflation driven by margins? take it →

Media

What was said, sourced and corrected.

How accurate are AI assistants' news summaries? take it →

Conflict

Sanctions, ceasefires, spending: the checkable parts.

Have sanctions reduced Russia's oil revenue, and by how much? take it →