Proximity over hierarchy
Use distance to adjust orientation. Keep action, object and consequence available throughout the flow.
“Cancel subscription?”
Exception: expand into a stable review panel when the choice needs more context.
Independent conceptSpatial content designInteractive prototype
GLASS / 01
What replaces the phone
isn't a phone.
It's a place.
A content system for the moment computing moves off the screen and into the world around us.
01 / PHONE ERA
Everything collapsed
into the phone.
Open app → Search → Read → Act
02 / GLASS ERA · A HYPOTHESIS
Everything collapses
out into the world.
Intent → Place → Review → Confirm
The shift is a design premise, not a prediction that phones disappear.
The problem / speculative framing
Mobile phones won by collapsing everything into a rectangle we stare down at. That rectangle is now at its limit. Cognitive load is maxed, attention is fractured, and the next wave of computing (XR glasses, ambient agents, Android XR) removes the rectangle entirely. How do you write interface copy when there is no interface? How do you preserve trust and clarity when words must live in space, not on a screen?
This provocation sets the brief. The prototype explores it; it does not establish these claims as research findings.
For the person wearing glasses
Manage a subscription at home without losing the context of the bill on the desk.
For the team building the experience
Give content, design and engineering a shared contract for what to say, where to put it and when to stop.
Less room. More responsibility.
Spatial
12 / 6 / 3
Candidate word budgets by distance. Required consequences never disappear to meet a limit.
Trust
A trace
Every confirmed action produces a readable receipt with inspectable source evidence.
Social
Quiet first
Voice is optional. Every spoken message has a visible equivalent.
System
One contract
Conditions, fallbacks and review ownership travel with each content pattern.
Let the place carry some of the words.
A nearby, clearly identified object may need less orientation. Start with a small vocabulary, then test whether people can still identify the object, action and consequence.
Your subscription, on the desk.5 words / 6-word budget
Distance is simulated. Six feet is across the room, not arm’s length. Budgets apply to orientation copy, not the complete transaction.
Three rules. Clear exceptions.
Anchor, not label
Tie placement language to a visible object. Name the action explicitly when location alone is ambiguous.
“Pin this here?”
Exception: “Cancel here?” becomes “Cancel subscription?” because “here” does not explain what ends.
Wayfinding voice
Add location when the anchor is hard to find. Remove redundant directions once focus is established.
“On your desk, to the left.”
Exception: gaze can change presentation. It cannot authorize an action or prove understanding.
The bill becomes the starting point.
GLASS / simulated first-person viewCancel subscription? ↗Illustrated environment · No camera, gaze or billing connection
The anchor opens a review. It does not cancel anything. The same action remains available without spatial tracking.
A short prompt. An informed decision.
Two subscription charges appear in the record. The person wants to stop renewal. GLASS separates that request from investigating the possible duplicate.
01 Identify02 Review03 Trace
Fictional provider: Northstar · Fixed example date: 22 October 2026 · Nothing is sent or saved
“Cancel subscription?”
Identify Northstar and open its review. A short anchor is an entry point, not complete consent.
“Renewal stops. Access stays until November 22. Confirm cancellation?”
9 words. Cause is not invented: auto-renew does not establish why two charges occurred.
“Renewal cancelled.”
Show effective date, retained access, unresolved charges and the supporting provider response.
Brevity is a constraint.
Understanding is the outcome.No measured word reduction, trust score or tap savings claimed.
A browser demonstration of proposed behavior.
Understanding is the outcome.No measured word reduction, trust score or tap savings claimed.
A browser demonstration of proposed behavior.
Content that comes with conditions.
Four reusable patterns specify when language appears, what it needs and how it fails. Version 0.1 is a proposal for review, not an adopted production standard.
Spatial confirm
SC–01“Cancel subscription?”
- Structure
- Action + named object. Location is optional.
- Use when
- A single identified object has a stable anchor. A candidate confidence threshold above 80% needs calibration.
- Budget
- 3 words for the entry prompt. Required review copy is exempt.
- Fail closed
- Ambiguous object: identify it first. No mutation from gaze or nod alone.
Causal whisper
CW–01“Two charges appear. Review your subscription?”
- Structure
- Verified observation or cause + proposed next step.
- Use when
- Voice is enabled and the interruption is appropriate. Otherwise show the equivalent text.
- Budget
- 12 words per short utterance. Split required detail into a user-controlled review.
- Fail closed
- If the cause is unknown, describe the observation. Do not infer a refund or cancellation.
Trace receipt
TR–01“Renewal cancelled. Access stays until November 22.”
- Structure
- Confirmed status + effective date + consequence + source.
- Use when
- The provider returns a verified outcome. Failed and unknown states use separate copy.
- Budget
- Readable summary first. Full evidence on request; no arbitrary receipt word cap.
- Fail closed
- No success claim without confirmation. Never call avoided renewal a recovered refund.
Wayfinding add-on
WA–01“On your desk, to the left.”
- Structure
- Location cue attached to the existing action.
- Use when
- The anchor is hard to locate. Below 60% confidence for two seconds is a test candidate, not an established threshold.
- Budget
- Use the distance budget where possible. Do not repeat the full action.
- Fail closed
- If tracking is lost, remove world anchoring and offer a stable panel. Preserve the review.
Inspect the behavior contract
INPUTS object_id · identity_verified · anchor_confidence · tracking_state user_confirmation · provider_status · source_record · locale POLICY ORDER 1. Identify the object; disambiguate if uncertain. 2. If tracking fails, switch presentation; keep task state. 3. Show action, consequence and alternatives before confirmation. 4. Execute only the explicitly confirmed action. 5. Report success only from a confirmed provider response. 6. Apply proximity and voice budgets to optional orientation copy. OUTPUT CONTRACT pattern_id · visible_text · voice_equivalent · next_action source_ref · status · required_disclosure VALIDATION Reject unsupported causes and invented source references. Never compress away material consequences to meet a word limit. A model may propose language; deterministic policy owns authorization. This prototype uses authored strings, not a trained model.
How the pattern library would be governed
Content design owns intent, language and exceptions. Engineering owns confidence calibration, state integrity and source availability. Accessibility and localization review precede release. Each change records its pattern ID, version, rationale, examples, counterexamples and test evidence. New locales get their own length and comprehension checks; English word counts are not universal limits.
What this demonstrates.
For a Staff UX Writing / Content Design, XR conversation, the useful evidence is the system itself: decisions, exceptions and inspectable behavior. This is an independent portfolio project, not a Google product or commissioned engagement.
Working in ambiguity
Turns a broad premise into one bounded task, explicit assumptions and a working simulation. Stakeholder adoption and user outcomes remain untested.
Model behavior design
Defines inputs, state-dependent responses and rejection rules. Separates adaptable wording from deterministic authorization. No model training is claimed.
Scalable content systems
Four patterns include triggers, budgets, exceptions and fallback language. Proposed ownership and versioning make future review concrete.
Technical traceability
Distinguishes observed data, user permission and provider confirmation. Exposes evidence without making API syntax the user's only explanation.
What I would test next
- Comprehension: can people explain what ends, when access ends and whether a refund happened before confirming?
- Recovery: can people continue after tracking loss without selecting the wrong subscription or repeating a request?
- Proximity: compare short anchors with explicit alternatives across distances, lighting, text sizes and access needs on actual hardware.
- Trust: can people distinguish confirmed, failed and unknown states, and find the evidence behind each?
Recruit representative participants, including people who do not use voice or gaze input. Establish a real baseline before making efficiency claims.