Faith Cheltenham Microsite

Heartspace

A memory-and-reward prototype for an AI boyfriend reminder app. It uses spacing, anticipation, and emotional intensity to decide when the next cue should arrive and how likely a reward should be.

  • Variable interval reinforcement instead of fixed streaks
  • Builder and harmonizer modes for different emotional contexts
  • Saved memory cards stored locally in the browser
Illustration of three AI boyfriends

Prototype angle

Memory orchestration, not streak pressure

Builder Mode

More frequent cues for lighter memories

Best for habits, gentle bonding, and memories where we want to grow fluency with shorter, steadier intervals.

Harmonizer Mode

Softer pacing for intense memories

Designed for emotionally loaded memories where prompt frequency and reward intensity should ease back instead of escalating.

Memory Vault

Save a memory and generate its next cue time

Live Schedule

No memory selected yet

Next prompt time

Waiting for input

Reward probability

0.0000

Builder mode favors faster reinforcement. Harmonizer mode adds extra spacing and tempers reward pressure.

{}

Saved Memories

Your local prompt library

Recent Research

Why the scheduler behaves this way

Spacing

A 2026 Nature Neuroscience study reported that learning scaled with the time between rewards, not just repetition count.

Read source

Adaptive Retrieval

A 2024 app-based spaced retrieval study supports tailoring recall schedules to individual forgetting and performance.

Read source

Time-Based Rewards

A 2024 reinforcement paper shows reward availability can increase as more time passes after the previous reinforcement.

Read source

Emotional Salience

Emotional cues can dominate neutral ones, so the microsite uses gentler frequency for high-intensity memories.

Read source

This microsite is a wellness prototype, not a clinical product. The emotional-weight rule is a product design inference from the studies above, not medical guidance.