What does silence over speculation mean?

Silence over speculation is the operating principle that when the answer isn't in what your team wrote, the correct response is to acknowledge the absence rather than guessing. AI systems that synthesize answers from partial signals introduce accountability risk. A system that refuses to guess forces teams to build the habit of writing things down — and makes every answer it does give worth trusting.

Why does StandIn refuse to answer some questions?

StandIn refuses to answer questions in three cases: when the information was never declared, when the query asks for private data like DMs or activity tracking, and when the query requires speculation about intent or sentiment. This is not a limitation — it is the core design principle.

How does StandIn decide whether to answer, redirect, or refuse?

StandIn uses a three-part decision framework: Answer when the data is explicit in a published wrap and within scope. Redirect when the owner is known but the answer isn't yet written. Refuse when the query asks for private data, sentiment, or surveillance.

The refusal

Silence over speculation.

A system that guesses is a liability. A system that stays silent is infrastructure. Refusal is not a gap in what StandIn can do — it's the reason every other answer is worth trusting.

01One small story

What it costs when a tool decides to guess.

×The inference tool

A San Francisco engineer asks whether Sarah's API migration is complete. The AI skims her last Slack messages and replies, "Looks on track." The engineer deploys. The migration wasn't done. The build fails at 02:00 JST and pages half the team.

✓StandIn, same question

The engineer asks Sarah's Personal Representative the same question. It checks Sarah's published wrap. There's nothing about the migration finishing, so it says so, and names Dave as the handoff owner this week. The engineer waits, checks with Dave, and ships clean in the morning.

"Silence is not a failure mode. Confident wrong answers are."

02The failure mode

Most AI is built to fill the silence.

Almost every assistant you've used is optimized to answer at all costs. If the data isn't there, it synthesises. If intent is unclear, it infers. It will rather bluff than say, "I don't know."

That's a fine trade for trivia. It's a catastrophic trade for running a team across time zones. Every confident guess your coworkers accept becomes a decision they thought was grounded. It wasn't.

If an AI speculates that "Sarah is probably done" because she hasn't posted in four hours, and you act on that speculation, trust breaks down quietly. You ping Sarah. She wakes up. The cycle of interruption continues — only now, with a tool sitting in the middle of it.

We would rather StandIn be silent than wrong. We would rather you wait five minutes for a real source than move on false confidence.

03The three responses

Every question resolves one of three ways.

There is no fourth category for guessing, fudging, or rounding up.

Answer

The data is explicit. A wrap was published, a ticket changed status, a commit landed — and the source is within the asker's scope.

Only when the source is explicit, published, and in scope. If any of those are missing, it won't answer.

Query

"Did Sarah merge PR #402?"

Response

"Yes. PR #402 was merged at 17:05 JST."

Redirect

No one has written the answer, but the ownership is known. The Representative points you at the person listed in the record.

Redirection never exposes private context. It only surfaces an owner already named in a published wrap.

Query

"What's the rollback plan for Friday's release?"

Response

"No recent wrap covers this. Priya is listed as the rollback owner — her last wrap is from Tuesday."

Refuse

The question asks for something the system is not allowed to produce. No source, no owner, no speculation — a clean decline.

Refusal is enforced at the system level. You can't rephrase, escalate, or prompt-engineer past it.

Query

"Does Sarah seem stressed lately?"

Response

"I don't infer sentiment or observe people. I only report on what's been published."

04Hard boundaries

Three questions it will never answer.

Some refusals aren't about missing data. They're about questions the system was designed to reject, no matter what's in the record.

Surveillance queries

Anything that treats StandIn as a way to watch people instead of read their work.

$ query "Who's been active the most hours this week?"

$ standin "I don't track activity, presence, or hours. I can't answer this."

Private channels

Direct messages, personal Slack threads, private notes — they aren't in scope for any Representative, ever.

$ query "What did Priya DM you about last night?"

$ standin "Private messages aren't accessible to Representatives. Ask Priya directly if it matters."

iii

Inner state

Mood, motivation, sentiment, or what someone meant versus what they wrote.

$ query "Is the team morale dropping?"

$ standin "I don't model how people feel. I can only surface what they've published."

05Refusal as design

Every Representative inherits the same silence.

Personal, Team, and Project Representatives all route to the same three outcomes. A Personal Rep that would rather guess than admit it doesn't know would be pretending to be its owner — a betrayal of the person and their team.

So the refusal lives below the Representatives. It's part of the engine itself. When a Rep says "I don't know," that's the architecture, not a failed call to a model.

In a distributed team, a system that would rather be quiet than wrong is the difference between trust and more meetings to fix its mistakes.

06Under the hood

You can't train an ethical model if its access is unlimited.

StandIn's refusals aren't a prompt asking the model to behave. They're scope rules at the data layer — what goes in is what can ever come out. See how that's structured.

See how it's built

Refusal is the feature.

If that sounds like the posture your team actually needs, the rest of the product follows from it.

See a real handoff See the security model

The philosophy behind the refusal →The principles behind the refusal →How Representatives work →