How Fog Room Works: The Clarity Scan

Privacy (the design constraint)

We designed Fog Room around a simple principle: Your inner world is private by default.

We do not store what you write.

To improve the tool, we only log structured signals like Fog Score and interaction taps (e.g., which feedback face you chose). We never log or train on your raw text.

What Fog Room estimates (signals)

Fog Room reads patterns in language to estimate a few interpretable signals:

1) Fog Score (0–100)

A rough estimate of how “foggy” the moment is—based on language patterns that correlate with internal complexity.

Fog tends to increase when we see:

• emotional load
• ambivalence / contradiction (“part of me… but…”)
• uncertainty markers and questions
• longer entries that contain layered context

Fog tends to decrease when we see:

• clear commitments or decisions (“I decided… I’m sure…”)
• language that indicates resolved direction

2) Theme (lightweight context)

A coarse theme classifier (e.g., work / relationship / family / identity / transition / self-blame).

This is not “who you are.” It’s simply a way to select the most relevant reflection templates.

3) Fork detection (Door logic)

Sometimes the scan detects the user is at a near-term decision fork—a moment where two paths are live and emotionally loaded.

When a fork is detected, Fog Room surfaces it explicitly and shows two Doors:

• Door A: Relief-first (what reduces uncertainty fast)
• Door B: Coherence-first (what preserves agency and long-term alignment)

Fog Room does not recommend a door. It clarifies the tradeoff.

4) Operator routing (which kind of move helps)

The scan routes to a micro-step style depending on the dominant mode present in the text:

• Sensing (somatic activation / panic)
• Thinking (rumination / over-analysis)
• Feeling (shame, relational strain, longing)
• Narrative (collapse, hopelessness, “what’s the point?”)

This decides what kind of “micro-step” will be offered (e.g., a body scan vs claim–evidence–alternative).

5) Impulse tendency (momentary)

Fog often comes with predictable impulses:

• seek contact
• numb out
• attack / discharge
• avoid / disappear
• over-control
• freeze

This helps the Doors reflect realistic human trajectories.

What Fog Room outputs (the product)

• Fog Score (and a band label like Low / Moderate / Significant / Heavy)
• Interpretation (grounded, non-dramatic reflection)
• One insight (short, resonant, context-shaped)
• One next clear step (small enough to do today)
• If fork detected: Doors A/B + a micro-step timer (60s–10m)

The output is intentionally limited. The goal is not “more content.”

The goal is more agency.

Known limitations (v0 honesty)

Fog Room v0 is intentionally simple and interpretable. That means it has limits:

• Language varies by culture and personality.
• Some people write “calmly” while highly activated; some write “dramatically” while fine.
• Keyword-based signals can miss nuance or produce false positives.
• This tool does not detect crisis or replace professional care.

So we treat v0 as an instrument, not an authority: helpful, imperfect, improving.

What we’re optimizing for

Our north star metric is not engagement. It’s: Did the user leave with more clarity and less compulsion? Fog Room is built to reduce dependence, not create it. If you use it repeatedly, the goal is that you need it less.