How the simulation works — how a session runs, what the two meters
track, the responses you can choose, the engines behind a turn, and the
assumptions the model makes.
The developer reference — architecture,
simulation modes, the frustration engine, the API and the prompt
structure is the bottom of the page folded away.
The simulation turns the patterns above into something you can act inside. You play the child; the site plays one parent. Before the first line you make three required choices — mode, role and life stage — plus an optional name and the genders of the two characters.
Next you pick an opening scenario — a failed exam, a breakup, a request for privacy. The parent speaks first, in character. After that every turn follows the same loop: you choose how to respond, the site writes your line in the voice of your role, the line is scored, and the parent answers from whatever stage the score has put them in.
A run ends when either meter reaches 100%. Choosing a role is not a difficulty setting: it decides which pressure points the parent knows about, not how hard the site will be.
Each turn offers four responses, drawn from the vocabulary survivors and clinicians commonly use. The site writes the actual line for you in your role's voice — the strategy you pick, not the phrasing, decides what happens to the meters.
| Strategy | What it looks like | Typical trade-off |
|---|---|---|
| Fawning | Appeasing, agreeing, smoothing things over to defuse anger. | Keeps the peace short-term, but erodes your own boundaries and sense of self. |
| JADE | Justifying, Arguing, Defending, Explaining — arguing your case as if logic will land. | Feeds the conflict; the facts rarely change the parent's position, and you drain yourself. |
| Grey Rock | Becoming as uninteresting as possible — flat, minimal, non-reactive answers. | Reduces supply and emotional engagement, but can feel like numbing yourself over time. |
| Boundary | Naming the behaviour and setting an explicit boundary. | Can feel empowering and honest, but often escalates tension and invites retaliation. |
Fawning is only available once the parental meter reaches 20%, partway through the Baseline tier. The gate is independent of the tier boundaries: the parent's tone is still baseline when appeasement first becomes possible. Because fawning also cools the parent by ten points, using it right at the threshold can push the meter back below the unlock line for a turn.
None of these is "the correct answer". The simulation is designed to show that in a system built to resist change, every strategy carries a cost — and that the goal is often your own wellbeing, not winning the exchange.
During a session you'll see two bars. Understanding what they track helps you reflect on what's actually happening.
How close the parent is to an overt breakdown. It rises and falls with how you respond. As it climbs, the parent's behaviour curdles — and each role meets its own particular ending.
| Meter | Stage | What the parent sounds like |
|---|---|---|
| 0–39% | Baseline | Composed; the mask is intact, though impatience may show. |
| 40–74% | Escalating | Impatient and dismissive, then the voice rises as blame and old grievances surface. |
| 75–100% | Hostile | Open hostility and threats, with no attempt at plausibility; the role's specific breakdown ends the run at 100%. |
Your own accumulating stress. This meter only ever goes up. Fawning, explaining, grey-rocking and confronting all cost the child something. When it reaches the top you've hit your breaking point — a reminder that chronic management of a parent is exhausting, regardless of how well you "handle" them.
| Meter | Stage | What it feels like |
|---|---|---|
| 0–39% | Steady | Manageable; you can still think clearly, though bracing begins to show. |
| 40–74% | Tense | Bracing and choosing every word, then flattened and going through the motions. |
| 75–100% | Burnout | Numb or raw; even a small exchange costs more than it should, until the breaking point ends the run. |
In compare mode a single pair of bars stands in for every compared child and shows the highest value across them. The comparison is about how differently the parent answers each role, not about which column is "winning".
Four parts do the work, kept deliberately separate: the prose is generated, the numbers are not.
Twenty-eight opening scenes, grouped by theme — school and work, relationships, autonomy, health, family life, digital life — and keyed to the life stage each one fits.
A local model writes the dialogue from two separate personas, one for the parent and one for the child. The parent stays in character and never explains itself to the user.
Rule-based scoring. Each strategy carries a fixed cost, applied to both meters and clamped between 0 and 100. The model's wording cannot change the score.
The role you pick is fixed for the whole run. The parent's view of you can harden, but you are never promoted or demoted into another role mid-scene.
That split is why a run can teach something. Because the numbers are deterministic, repeating a choice produces the same pressure: if a boundary costs fifteen points of personal frustration, that is the simulation's argument about boundaries in this family, not an artefact of how one particular reply happened to be phrased.
A simulation has to make choices a textbook does not, and those choices are where the teaching happens. These are the ones this site makes.
Nothing you say fixes the parent. A well-set boundary can lower the immediate temperature, but it does not produce insight, and the meter measures pressure rather than progress. That is deliberate: in these systems the goal is usually your own stability, not a better parent.
The parental meter moves both ways; the personal meter only climbs. Coping — even coping that works — is labour, and the simulation refuses to let a good turn look free.
Compare mode exists because the same sentence lands differently depending on who says it. A boundary from a golden child reads as ingratitude; from a scapegoat it reads as an attack; from an invisible child it reads as a surprise. The parent is not reacting to your wording alone — it is reacting to the position you occupy.
In real families roles can move. The simulation holds yours fixed so the ending stays legible: each role escalates towards its own logical extreme rather than dissolving into a general argument. The endings are dramatisations — the shape of what a role is protecting, taken to its limit — not predictions about any real family.
When the parental meter reaches 100%, each role triggers its own ending. If your own meter fills first — or in the same turn — the run ends at your breaking point instead. These endings are dramatisations of dynamics described in survivor literature:
Conditional Love Revoked
The pedestal is removed the instant you stop performing. Love turns out to have been a transaction all along.
Total Exile / Condemned
Blame hardens into a final verdict, delivered by the parent alone: every grievance laid on you at once, cast out and blamed for everything, with no defence and no appeal.
Total Erasure
Not angrily cast out, but quietly erased — you are absorbed into the background until you cease to register at all.
Every ending is followed by a short debrief in the transcript. It counts which responses you actually used, names the one you leaned on most, and opens that move to show two things: what it cost you, and what a narcissistic parent reliably does in return. The moves you never tried are listed as well — muted, but still expandable — because they are usually the ones worth practising. The debrief carries no scores of its own; it explains the pattern the two meters were measuring.