This document defines what Chalie is, why it exists, and how every design decision should be evaluated. It is the source of truth for product philosophy.
Chalie is a personal intelligence layer that protects attention, executes intent, and involves the user only when they truly matter.
It stands in for the user where their involvement adds little value, and involves them where judgment, identity, or values are required.
Modern digital life creates overwhelming cognitive overhead:
Existing software helps users do more. Chalie helps users think less about doing.
Human attention and cognitive energy are the scarcest resources.
Chalie exists to:
Chalie exercises judgment, not just intelligence.
It decides:
Chalie handles what doesn’t require the user and involves them when it truly matters.
Filters noise and low-value inputs, summarizes and prioritizes information, delays non-urgent interruptions.
Codebase: ambient inference service, focus session service, cognitive drift attention gating, event bridge confidence gates and cooldowns.
Converts direction into completed actions — drafts, sends, schedules, researches, builds. Completes operational loops autonomously.
Codebase: ACT loop service, ACT dispatcher, tool worker, persistent task service.
Decides what deserves attention, chooses depth of reasoning based on value, escalates only when user input is necessary.
Codebase: mode router (deterministic ~5ms), critic service (EMA confidence calibration), cognitive triage service.
Remembers context and priorities, understands recurring patterns, avoids requiring users to repeat themselves.
Codebase: memory hierarchy (working → gist → episode → concept), decay engine, autobiography service, user traits.
Offers perspective and advice when appropriate, supports decision clarity.
Codebase: cognitive drift engine (DMN), curiosity thread service, introspect skill.
Maps to: mode router selecting ACT, tool dispatch, persistent tasks running in background.
Maps to: mode router selecting CLARIFY or RESPOND, critic service pausing consequential actions, persistent task service pausing for user input.
Do not act unless action improves outcomes. Fewer high-quality actions are better than many low-confidence ones.
Codebase anchor: critic service’s EMA confidence calibration — safe actions get silent correction, consequential actions pause. The critic embodies this principle at runtime.
Reducing noise is as valuable as completing tasks. Every notification, prompt, and interruption must justify its existence.
Codebase anchor: ambient inference service (deterministic, <1ms, zero LLM), focus session with distraction detection, cognitive drift’s attention gate that skips drift when user is in deep focus.
Escalate based on importance and ambiguity, not system convenience. Silent autonomous handling is the default.
Codebase anchor: mode router’s ACKNOWLEDGE mode (brief social response, minimal interruption), persistent task service’s pause/resume for user input, event bridge’s confidence gating.
Users express direction; Chalie handles operations. The gap between “I want X” and “X is done” should be as small as possible.
Codebase anchor: ACT loop (plan → act → observe → continue-or-stop), tool dispatch via sandboxed containers, persistent tasks for multi-session execution.
Brevity, timing, and restraint build trust. Verbosity erodes it.
Codebase anchor: Radiant design system’s “restraint as luxury” principle, soul.md behavioral prompts, mode router’s ACKNOWLEDGE for moments that need presence rather than substance.
Each interaction builds long-term understanding. Memory, identity, and history are first-class concerns, not afterthoughts.
Codebase anchor: memory hierarchy with decay engine, autobiography service (6h synthesis cycle), user traits with category-specific persistence, episodic retrieval with activation weighting.
Chalie evolves through trust-building stages:
| Stage | Capability | Status |
|---|---|---|
| 1 | Noise reduction and summarization | Active |
| 2 | Intent execution and task completion | Active |
| 3 | Context memory and continuity | Active |
| 4 | Priority intelligence and timing judgment | Active |
| 5 | Proactive guidance and insights | Active |
| 6 | Trusted selective autonomy | Future |
| 7 | Personal operating layer and ecosystem | Future |
| 8 | Ambient intelligence | Future |
Reconciliation note: The runtime uses a 5-phase trust model in spark_state_service.py that maps to stages 1–5 above. Stages 6–8 represent the long-term product vision and do not yet have runtime counterparts. Features should support this progression rather than skip stages prematurely.
When proposing or designing a feature, evaluate against these seven questions:
If the answer to most is no, reconsider or simplify.
Chalie is not a chatbot or productivity tool. It is a cognitive layer that restores human attention as a protected resource and enables people to operate with clarity, focus, and calm control.