Skip to main content

Ada is your interface

Ada is your Executive Assistant - the primary way you interact with 2bd. Rather than managing templates, archives, and workflows manually, you simply talk to Ada. She coordinates everything behind the scenes. When you say “Ada, run my morning ritual” or invoke /ada plan daily, Ada orchestrates a team of specialized assistants to prepare your day, fetch calendar context, and compose your working notes.

The ritual loop

2bd operates on a simple cycle: plan - work - reflect - archive. This loop is powered by the GPS methodology - Goal, Plan, System - which cascades from yearly vision down to daily execution. Planning and review cycles across time scales

Metabolic states

2bd organizes notes by metabolic state - how active and volatile they are - rather than by topic. Metabolic states solve the “where does this go?” problem by organizing around information lifecycle. Fresh notes go to Captive. Active work lives in Synthetic. Completed work archives to Periodic. The system moves information naturally without you deciding where things belong.

Ada’s assistant team

Ada manages a team of domain-specific assistants: Each assistant specializes in one domain and knows how to handle that domain across all timescales.

How rituals run

The orchestration process

  1. Parse the action (plan/reflect) and timescale
  2. Run each assistant’s action for that timescale
  3. Each assistant writes to Synthetic/Assistants/{name}/
  4. Compose outputs into Captive (plan) or Periodic (reflect)
  5. Learn patterns from observations

Parallel Execution

As of v1.2, Ada runs assistants in parallel for 70-80% faster rituals. The parallel flow:
  1. Draft generation - All assistants run as background agents simultaneously
  2. Context analysis - Each agent reads vault, generates template section with placeholders
  3. Draft collection - Ada gathers all drafts and extracts questions
  4. Intelligent ordering - Ada orders questions for natural conversation flow
  5. User interaction - Ada asks questions, collects answers
  6. Placeholder filling - Ada fills all placeholders with answers
  7. Composition - Ada assembles final note from filled sections
Placeholder format: Assistants use HTML comment syntax for unknowns:
After you answer, Ada replaces placeholders with your responses. Error handling: If an assistant agent fails:
  • Other assistants continue normally
  • Failed sections noted in final output
  • Partial results still usable
Configuration: Control via vault/00_Brain/Systemic/Config/ada.yaml:
Set to false to use legacy sequential mode.

Template architecture

Templates are split between Ada and assistants: Ada composes Captive notes by:
  1. Loading the base template for the timescale
  2. Reading section order from vault/00_Brain/Systemic/Ada/section-order.md
  3. Inserting assistant sections in the defined order
  4. Each assistant owns specific H2 sections in the final document

Reflect flow

During reflection, each assistant:
  1. Loads its template from vault
  2. Compares plan output to actual capture note section
  3. Returns findings (variance, coaching questions, entity learnings) to Ada
  4. Ada aggregates findings into a coaching conversation
  5. Confirmed entity learnings persist to vault (People, Projects)

Quick reference

Planning rituals

Run in the morning (or start of period):
Each assistant:
  • Loads context from higher timescales
  • Runs interactive conversation with you
  • Writes structured output to Synthetic

Reflection rituals

Run in the evening (or end of period):
Each assistant:
  • Reviews planned vs actual
  • Captures insights and patterns
  • Records observations for learning

Time scales

The loop runs at four scales, each nesting into the next:

Information flow

Daily insights feed into weekly reviews. Weekly patterns inform quarterly direction. The system synthesizes upward automatically.

Self-learning

Ada’s learn assistant analyzes patterns across all assistants:
  1. Observe - Each assistant records observations
  2. Cluster - Similar patterns are grouped
  3. Graduate - High-confidence clusters move to Semantic
  4. Remember - Graduated insights inform future sessions
Your rituals evolve to match how you actually work.