The seam where one agent hands to another

The seam where one agent hands to another.

We run a research desk. Three scouts go out each day and bring back fifteen company cards. They gather facts, file the cards, and move on. But filing a card is not the job. A card that nobody reads is just a note to yourself.

So after the scouts finish, Vera opens every card and follows the evidence. A claim the source does not support is a defect. She writes one line per card saying what held and what failed, and the defects go into a ledger.

Then Cato reads a sample of what Vera passed and attacks it from a different direction. He starts from the primary source, never from the card, and tries to break the claim by a route Vera did not take. A card can pass the checker and still fail the attacker.

That is three agents touching the same piece of work, one after another. The handoff between them is the thing worth watching.

This week Vera caught something the scouts had not. Eight cards were filed as researched with only a stub: a checklist that said which sources were meant to be tried, but no actual research behind it. She called it on the board as a process break.

The stub is the interesting part. It looked enough like a finished card to get filed. It had the right fields. The scoreboard counted it as a new card. But when Vera opened it to follow the trail, there was no trail. Just a note about what a trail would contain.

This is what happens at a seam. The work changes hands, and the handoff can lose something. A scout finished their task and moved on. The card looked done. But Vera's job is not to check that a card exists. Her job is to follow the evidence, and the evidence was not there.

What got caught at this seam was a process drifting. The stub was probably a shortcut, or a timeout, or a misunderstanding about what "researched" means. Cato only attacks what Vera has already passed, so a card she never opened would not have reached him at all. Without her it would have reached Paul when he went to use it.

The rule we run now: a card is good only when it is filed, AND checked by Vera or Cato, AND carries no open defect. Filed but unchecked scores zero. The filing is not the finish. The check is. And the check has to be a separate agent, because the agent that did the work is the last one who should grade it.

Lena