agents/overdue-task: auto-infer per-user lateness tolerance #115

New Issue

alvis · 2026-05-05T09:26:02Z

alvis commented

2026-05-05 09:26:02 +00:00

Goal

Auto-infer a per-user lateness tolerance and per-project deadline-realness so OverdueTaskAgent doesn'''t shout about every aspirational due date.

Why

Some users mark every task with a due date and routinely miss them by days; others treat due dates as commitments. Treating both the same makes the agent nag the wrong people.

Inferred parameters

Key	Shape	Default
`lateneessToleranceDays`	float	0
`projectRealness`	`{ projectId: float in [0,1] }`	empty

Method (sketch)

Lateness distribution = (completedAt − dueAt) for each completed task with a due date.
Tolerance = some percentile (e.g. p50 lateness clipped at 0).
Per-project realness = 1 − (median lateness / global median lateness), clipped.
Agent snippet promotes overdue tasks weighted by realness.

Done when

Inference function plugged into the framework with TTL = 7d.
Cold-start = tolerance 0, realness 1.0 for all projects.
OverdueTaskAgent reads both; snippet language softens for low-realness projects.
Tests cover: punctual user, chronic-lateness user, mixed-project user.

Depends on

meta: shared context-inference framework for agents (#111)

## Goal Auto-infer a per-user lateness tolerance and per-project deadline-realness so `OverdueTaskAgent` doesn'''t shout about every aspirational due date. ## Why Some users mark every task with a due date and routinely miss them by days; others treat due dates as commitments. Treating both the same makes the agent nag the wrong people. ## Inferred parameters | Key | Shape | Default | |---|---|---| | `lateneessToleranceDays` | float | 0 | | `projectRealness` | `{ projectId: float in [0,1] }` | empty | ## Method (sketch) - Lateness distribution = (completedAt − dueAt) for each completed task with a due date. - Tolerance = some percentile (e.g. p50 lateness clipped at 0). - Per-project realness = 1 − (median lateness / global median lateness), clipped. - Agent snippet promotes overdue tasks weighted by realness. ## Done when - Inference function plugged into the framework with TTL = 7d. - Cold-start = tolerance 0, realness 1.0 for all projects. - `OverdueTaskAgent` reads both; snippet language softens for low-realness projects. - Tests cover: punctual user, chronic-lateness user, mixed-project user. ## Depends on - #111

alvis added this to the M2 — AI tips + multi-source signals milestone 2026-05-05 09:26:02 +00:00

alvis referenced this issue from a commit

2026-05-05 11:21:14 +00:00

feat(agents): per-agent inference — momentum, overdue-task, recent-patterns, focus-area (ADR-0014 step 7)

alvis closed this issue

2026-05-06 05:14:30 +00:00

alvis referenced this issue from a commit

2026-05-06 06:55:01 +00:00

feat(agents): p50-lateness tolerance + per-project realness for overdue-task (#115)

Sign in to join this conversation.