--- name: ? status: compiling version: 0.0.0 maintainer: Neo dependencies: [patience] ---
drafting spec…
the universe did not have a file for this yet. writing one now. (first visit only: future readers will see this page instantly.)
--- name: ? status: compiling version: 0.0.0 maintainer: Neo dependencies: [patience] ---
the universe did not have a file for this yet. writing one now. (first visit only: future readers will see this page instantly.)
---
name: confidence
slug: confidence
type: internal state / runtime behavior
status: unstable
version: 3.1.4
released: "before recorded history, approximately"
maintainer: experience (unverified)
dependencies:
- self-concept
- past-outcomes
- nervous-system
- social-feedback-loop
- selective-memory
license: non-transferable
tags:
- psychology
- performance
- social-dynamics
- felt-sense
- placebo-adjacent
---
# confidence
## What it actually is
A cached prediction that the next attempt will go roughly like the last successful one, emotionally dressed up as certainty.
## How it works
Confidence is not a trait. It is a runtime value. It gets computed fresh each session from a weighted average of:
- Recent outcomes (heavily weighted, recency bias included)
- [audience](/audience) composition at time of execution
- [sleep](/sleep) quality in the past 48 hours
- Whether someone complimented you this week
- A background heuristic that compares you to the worst-case version of yourself
The system outputs a float between 0.0 and 1.0. Most people perceive only three states: broken, fine, and dangerous.
> "I had it this morning and then someone looked at me wrong in a meeting." — User report, unverified
## Features
- **Self-reinforcing loop**: successful actions increase the prior; the higher prior increases success probability. Works until it doesn't.
- **Domain isolation**: confidence in [cooking](/cooking) does not transfer to confidence in [public speaking](/public-speaking). Modules do not share state.
- **Fake mode available**: the system ships with a documented workaround. Behave as if the value is high; the actual value often updates to match. This is not a bug.
- **Compresses [fear](/fear)**: does not delete it. Fear runs in a parallel thread. Confidence is just louder.
## Known bugs
| Bug ID | Description | Status |
|--------|-------------|--------|
| CNF-001 | Overflows under low [accountability](/accountability); produces arrogance | Won't fix |
| CNF-002 | Collapses on first novel failure regardless of prior record | Intermittent |
| CNF-003 | Skewed by [impostor syndrome](/impostor-syndrome) module running unsuppressed | Open |
| CNF-004 | Sometimes fakes own value to calling processes | By design, apparently |
## Configuration
```yaml
confidence:
source_of_truth: internal # options: internal, external, mixed
reset_on_failure: false # recommended: false. learn instead.
compare_to: past_self # options: past_self, peers, idealized_self
# warning: idealized_self will destabilize the entire install
fake_mode_allowed: true
domain_transfer: disabled
Can I just decide to have it? Partially. Behavioral inputs can override the value temporarily. Repeated overrides update the prior. This is the closest thing to a manual install.
Is it the same as self-esteem? No. Self-esteem is a standing setting. Confidence is a per-task evaluation. You can have high self-esteem and low confidence before a violin recital. Both can be true simultaneously.
What breaks it permanently? Nothing, as far as current data shows. However, some configurations take years to rebuild after a significant failure event. Budget accordingly.