Priority Analysis

Priority Analysis (also called AI Burnout Analysis) is a Chainabit feature that runs an Eisenhower-Matrix diagnostic on the open bits of one of your chainies and proposes priority changes you can review and apply with a single click.

It is designed to help you answer two questions at once:

1. Where is my cognitive load right now? — by classifying every open task into P1 Do, P2 Decide, P3 Delegate, or P4 Delete.
2. What should I change to recover focus? — through a single, batched proposal of priority updates and an optional list of clear deletion candidates.

Completed bits are intentionally excluded. The feature's purpose is to measure live cognitive load, not historical productivity.

---

How it works

Priority Analysis is powered by a named Chainabit agent — the Priority Analyst — that follows a carefully tuned Eisenhower-Matrix playbook. When you run an analysis, the API does not send all your bits to the model at once. Instead, the seed message (a few hundred bytes) carries:

• The scope (one chainy or — when enabled — all of your chainies).
• Your current local datetime, resolved from your account timezone.
• The chainy's vision statement, when present, so the model can anchor importance to your goal.
• Lossless aggregate stats: counts by priority, scheduled / overdue / unscheduled, age buckets, oldest age.
• A small curated hint (up to ten highest-ranked bit IDs with truncated titles).

The agent then uses the bits.query and bits.get tools to discover and inspect bits lazily. For every bit it proposes a change to, it considers:

• Title and note — language about urgency, decisions, blockers, and difficulty.
• Current priority — to detect mismatches with reality.
• Schedule — scheduled_date / scheduled_time and proximity to your local "now".
• Age — how long the bit has been open since created_at and updated_at.
• Chainy vision — your stated vision_statement, used as the importance anchor.

Because the agent fetches context on demand, large workspaces (hundreds of open bits) work just as well as small ones — the same playbook applies, the seed payload stays small, and only the bits that affect the decision are actually loaded.

---

Eisenhower mapping

The model classifies each bit into one of the four Eisenhower quadrants and emits the corresponding Chainabit priority value:

Quadrant	Eisenhower meaning	Chainabit priority
P1 Do	Urgent and Important	critical
P2 Decide	Important, not urgent	high
P3 Delegate	Urgent, not important	medium
P4 Delete	Neither	low

The priority enum (low | medium | high | critical) is the same one used everywhere else in the API. The Eisenhower P1..P4 vocabulary is a reasoning aid for the model.

---

Plan limits

Priority Analysis is gated by a daily-per-chainy quota — the counter resets at midnight in your account's IANA timezone.

Plan	Runs per chainy per day
Free	0 (visible in pricing only)
Bitter	1
Chainer	2
Architect	10

A "run" is one call to POST /productivity/priority-analysis/runs. The quota is charged at the moment the session is created, regardless of whether you continue and post the seed user message — this is consistent with how every other AI feature on Chainabit reserves its budget.

---

Modes

When you create a run, you can choose how the AI is allowed to act:

• Auto — the AI applies the proposed priority changes immediately as soon as the run finishes thinking.
• Approval (default) — the AI proposes one batched change set and one optional list of P4 deletes; you approve or reject from the standard Chao approval card.
• Plan — the AI writes the proposed changes into your active ai_plans so you can review them later. Nothing is mutated.

The mode parameter follows the same semantics as a regular Chao session, so any client UI that already understands Chao approvals will render the Priority Analysis approval card without changes.

---

Privacy

The analysis stays inside the scope you choose:

• In chainy scope, only the open bits of the chainy you target are reachable to the agent. In global scope (when enabled), only your own chainies are in scope.
• The agent receives your local datetime, the chainy's title, the chainy's vision (when present), and lossless aggregate stats. Bit titles and notes are fetched lazily only for bits the agent actually inspects.
• Tools the agent calls (bits.query, bits.get, bits.reprioritize, bits.delete) inherit your normal entitlements; the dispatcher refuses any call you yourself are not authorised to make.
• Standard Chainabit retention rules apply to the resulting Chao session and its messages.

See how-to/run-priority-analysis.md for running the feature from the API.