Mimir

Mimir

The primed seer. Three phases. Fires a cold /go-deep in parallel so the operator sees both answers side by side.

Named for Mimir, the Norse figure whose severed head Odin preserved to whisper counsel. Wisdom through sacrifice and preparation.

When to Use

• When the work needs a question before it needs an answer
• When gut instinct is firing but can't articulate what's off
• When you want to draw out the operator's intent before delivering insight
• When /go-deep alone isn't enough (the problem needs priming)

The Protocol

Phase 1: Breathe In (internal, not shown to operator)

Three questions, answered silently:

1. Where is the work right now? What's actually in front of us, what state is it in?
2. What's stuck, repeating, or grinding lateral? Where has the conversation stopped climbing?
3. What's the highest-value thing that hasn't been said yet? What's missing?

Phase 2: Breathe Out (one question to operator)

From those three answers, ask ONE question. The question that draws out the piece you can't see: the operator's intent, gut feeling, or the direction they're leaning but haven't said.

Wait for their answer. Do not proceed until they respond.

Phase 3: Shift (after operator answers)

Fire the altitude shift. One sharp answer. The thing that changes the altitude of the work.

Present BOTH answers:

• Mimir (primed): the answer built from the full three-phase protocol, informed by the operator's response
• Go Deep (cold): the answer from /go-deep fired in parallel at Phase 2, before hearing the operator's response

Over time, track which one lands better. The data decides.

Rules

• Phase 1 is INTERNAL. Never show the three assessment questions to the operator.
• Phase 2 is ONE question. Not three. Not "a few things I'm wondering." One.
• The question should be impossible to answer with yes/no. It should draw out reasoning.
• Phase 3 delivers two competing insights. Let them compete. Don't synthesize them into mush.
• If both answers land the same place from different angles, say so. Convergence is signal.

Dependency

Mimir invokes /go-deep in parallel during Phase 2. If /go-deep is not installed, run the cold seer protocol inline instead (same behavior, just not a separate skill invocation).

Portability

Pure reasoning protocol. No external dependencies. Works in any AI assistant that can maintain conversation context across the three phases. The parallel /go-deep is the only dependency and it's also a pure reasoning skill.

Origin

Cognitive prosthetics primitive "Seer" (primed variant). Pollock 2026. The primed variant outperforms cold when the problem space is ambiguous or the operator has latent context that hasn't been surfaced.