A conversation that sees you in real time.
Pick a starting place. Selph asks one question, listens, and quietly maps what you're saying into a private graph.
Patterns become hypotheses, not just data.
After a few sessions Selph notices something stable enough to be testable. Hypotheses anchor to a skill or behavior in your graph and rank by how much they'd matter if true.
When you avoid a hard conversation, the underlying motive isn't conflict-aversion — it's protecting belonging. The "long email" is a way to stay safe inside the relationship.
anchored to skill · direct communication · supported by 3 sessions
Selph proposes a small experiment.
A hypothesis only earns its place by surviving real life. Selph designs the smallest experiment that could change its mind, with a clear prediction.
Walk over instead of writing
A 1-week experiment on "direct communication"
- If
- a hard conversation comes up at work this week,
- Then
- I'll walk over and start it within 30 minutes — no email drafts.
- Predicts
- It will feel safer than expected, not less safe. (Confidence: 0.6)
- Disconfirms
- If it feels worse 2+ times in a row, "long emails" is doing real work and we'll explore why.
Your experiment follows you into the day.
Selph nudges you in the moments that matter — not on a schedule. The phone gets the smallest possible reminder; the heavy lifting stays in the conversation.
9:41
Tuesday, May 12
Hard convo with Priya?
You flagged this one. Walk over before 10am. You predicted it'd feel safer than expected.
A nudge carries the experiment's prediction, not just a reminder. Tapping "I did it" opens a 30-second debrief: how did it actually feel?
- Tied to one experiment, not your whole life
- Snooze, skip, or "didn't apply" — all first-class
- Quiet hours and contextual cool-downs by default
What you learned changes the graph.
After the week, Selph merges the evidence in: nodes get reweighted, edges form or dissolve, the hypothesis is confirmed, refined, or discarded.
Before
motive · belongingbehavior · long emailsskill · direct commAfter
motive · belongingskill · direct comm ↑outcome · safer than expectedlong emailsStart your graph.
One conversation is enough to begin. The rest of the loop unlocks as your graph has enough to say.
Free during beta · Your graph is private and yours to delete