Self-Evolution Center
What ORANIS wants to change about itself. It tunes its own parameters within the bounds you set, watches whether the metrics it optimizes still predict real money, turns failures into general rules, and earns autonomy by being right. You approve anything risky.
Pending parameter changes (0)
None — nothing needs your approval right now.
Live parameters
| Parameter | Now | Default | Range | Mode |
|---|---|---|---|---|
| differentiation_min appstore | 55 | 55 | 40–80 | auto |
| four_three_min appstore | 60 | 60 | 45–85 | gated |
| remediation_max_iter appstore | 3 | 3 | 1–5 | auto |
| submit_approval_min appstore | 85 | 85 | 70–95 | gated |
| anti_template_threshold generation | 72 | 72 | 60–85 | auto |
| winner_score_threshold portfolio | 70 | 70 | 50–90 | gated |
Am I optimizing the right thing?
Correlation between each internal score and the real outcome it should predict. Low + flagged = the metric has drifted from reality.
| Metric | Corr. | Samples | Status | Suggested reframe |
|---|---|---|---|---|
| differentiation_score | 0.00 | 0 | ok | — |
| quality_score | 0.00 | 0 | ok | — |
| winner_score | 0.00 | 0 | ok | — |
Failures turned into rules
None yet — generated when apps fail the pre-submission gate.
Earned autonomy
The better-calibrated ORANIS proves on a decision, the lower the confidence it needs to act alone — but never below your ceiling. Irreversible spend stays gated.
| Decision | Calibration | Threshold | Proposed | Your ceiling | |
|---|---|---|---|---|---|
| kill_app | — | 0.90 | — | 0.85 | |
| paid_ua_spendirreversible | — | 0.95 | — | 1.00 | |
| price_changeirreversible | — | 0.95 | — | 1.00 | |
| submit_app | — | 0.85 | — | 0.80 |
Self-evolution runs on the /api/cron/evolution schedule. It is conservative by design: it requires real outcome data before changing anything, steps in small bounded increments, records every change for rollback, and routes anything risky to you here.