Decisions

  • Pending: Hook-based vs memory-based vs hybrid approach?
  • Pending: How prominent should the learning nudge be? (inline note vs separate section)

User Tasks


Summary

Automatically surface relevant learning topics from vault/20_knowledge/personal/learning-topics.md when the user works on something related — so they learn by doing instead of studying separately.

Problem / Motivation

The user has a list of topics to learn (agentic programming concepts). These topics are only useful when they come up naturally during work — not as a study list. Without a surfacing mechanism, they sit in a file and get forgotten. The inbox should stay clean (not hold learning items), and topics should be completable (check off and remove).

Proposed Solution

A hybrid approach:

  1. Memory (immediate) — Claude remembers the topics and explains relevant concepts when they come up in conversation. Already active.
  2. Hook/rule (future) — a trigger that detects when work touches a learning topic and adds a brief explanation to the output. E.g., when creating a sub-agent, explain how agent specialization works.

Open Questions

1. Surfacing mechanism

Question: How should topics be surfaced?

OptionDescription
A) Memory onlyClaude remembers and explains when relevant — no automation
B) Memory + hookMemory for now, add a hook later that checks the topic list
C) Briefing integrationDaily briefing includes “today’s learning opportunity”

Recommendation: Option B — memory works today, hook adds reliability later.

2. Nudge format

Question: How prominent should learning nudges be?

OptionDescription
A) InlineBrief note within the normal response (“btw, this is related to your learning topic X — here’s what’s happening…“)
B) Separate sectionA boxed “Learning Note” section at the end of responses
C) Only on requestUser asks /learn to get the explanation

Recommendation: Option A — least disruptive, most natural.


Phase Overview

PhaseDescriptionStatus
Phase 1Memory-based surfacing (manual)in-progress
Phase 2Hook that cross-references topic list with current work
Phase 3Auto-check-off topics after sufficient exposure

Phase 1: Memory-Based Surfacing in-progress

Goal: Claude knows the learning topics and explains concepts when they come up.

File / FeatureDetailsOwnerStatus
vault/20_knowledge/personal/learning-topics.mdTopic checklistopusdone
Claude memory entryMemory linking topics to triggersopusdone
Remove vault/90_inbox/learning-goals.mdClean inboxopusdone

Phase 2: Hook-Based Trigger —

Goal: A hook or rule that detects relevant work and reminds Claude to teach.

File / FeatureDetailsOwnerStatus
Trigger rules mappingWhich actions/files map to which topicsopus
Hook script or ruleAutomatic detectionopus

Test

Manual tests

TestExpectedActualLast
pending-

AI-verified tests

ScenarioExpected behaviorVerification method

E2E tests

ScenarioAssertion

Integration tests

ComponentCoverage

Unit tests

ComponentTestsCoverage

History

DateEventDetails
2026-03-12CreatedSplit from learning-goals.md, which moved to knowledge

References

  • FR-064 (Personal Learning Feed) — delivers learning content via messaging; this FR surfaces topics during work sessions
  • FR-062 (Learning Mode) — related concept