WORKSPACE

Bench

The bench's work: what was asked, what was found, and what is going stale. Evidence has a shelf life — a bench that admits it is better than one that does not.

Press /

Agents at the edges, arithmetic at the core. Citations only, never prose. 'Unestablished' is a valid and often correct answer. Propose, never author. What the bench finds informs a verdict; it never becomes content to adapt.

The bench has not been asked anything yet.

A workspace that has asked nothing shows what the bench is for. The Market Researcher answers questions about competitive timing and angle exhaustion — what was covered this week, whether a subject is current. The cases:

Has this angle been covered this week elsewhere?

competitive timing — whether the subject is exhausted before you ship

Is this subject still current, or has the conversation moved on?

angle exhaustion — a stale angle is a wasted slot

What is the market saying about this topic right now?

editorial advantage — what the bench found, with citations and dates

Cost per specialist

No usage records yet. A specialist that has never been called is different from one that was called for free — cost appears here once the bench runs. Where a model has no rate entry, usage is recorded as null, disclosed — never estimated.

Staleness bands
fresh — within 7 days
aging — within 30 days
stale — beyond 30 days

A rights position from March is not current in September. A bench that accumulates stale answers is worse than one that admits it.