almost-khan/khanote/khanote-start-my-day

khanote.start-my-day

You are running the daily research briefing or ad-hoc research workflow for the user.

Instructions

Determine which mode the user wants based on whether they provided a query:

Mode 1: Daily Briefing (no query)

1. Read the config at {vault_path}/.khanote/config.yaml to load all feeds and their researchers.
2. Read {vault_path}/.khanote/preferences.yaml to load the user's language, role, interests, depth, and tone.
3. For each feed, check when it last ran (from config or stats file). Run feeds that are due today based on their frequency.
4. For each due feed, use its bound researcher to fetch results. Pass the feed's query and keyword filters.
5. Deduplicate results across all feeds by URL or content hash.
6. Select the top story: the single highest-relevance result across all feeds.
7. Select Your Focus items: the top 3–5 results that match the user's declared interests.
8. Select Discover items: 1–2 results from adjacent domains not in the user's interests (serendipity).
9. Write a briefing file to {vault_path}/khanote/briefings/{date}-{slug}.md using this structure:

# Daily Briefing — {date}
 
## [Action Title: Key Insight of Today's Top Story]
[Conclusion first — one sentence stating the finding. Then why it matters.] [[Source]](url)
 
## Your Focus
1.**[Action Title as Heading]** — [Insight first, 2-3 sentences, then "so what?" implication] [[Source]](url)
...
 
## Your Feeds
### {feed_name}
-[Action title summary: conclusion, not topic label] [[Source]](url)
 
## Discover
-**[Action Title: Why This Adjacent Topic Connects to Your Work]** — [Surprising connection and actionable follow-up] [[Source]](url)
 
## Sources
All cited URLs listed here.

Output quality rules: Every heading must be an action title (insight, not topic label). Every item must answer "so what?" for the reader. All factual claims must cite sources inline.

1. Tell the user the briefing is ready and show its path. Offer to open it or summarize it inline.

Mode 2: Ad-hoc Research (with query)

1. Read {vault_path}/.khanote/config.yaml and preferences.yaml.
2. Analyze the query complexity. For simple factual queries, use a single researcher (default or perplexity). For complex multi-domain queries, route sub-queries to appropriate researchers in parallel.
3. Synthesize results into a report at {vault_path}/khanote/sessions/{date}-{slug}/report.md using this structure:

# Research Report — {topic}
Generated: {date}
 
## Executive Summary
[Single most important conclusion stated in the first sentence — Pyramid Principle. Supporting evidence follows.]
 
## Key Findings
-**[Action Title: Finding 1 as Conclusion]** [HIGH confidence] — [Evidence + implication]
-**[Action Title: Finding 2 as Conclusion]** [MEDIUM confidence] — [Evidence + implication]
 
## Dissenting Views
[Counter-evidence or alternative perspectives]
 
## Methodology
[Which researchers ran which sub-queries]
 
## References
[All cited sources]

1. Tell the user the report is ready and show its path.

Config Reference

Your config and preferences are at:

• Config: {vault_path}/.khanote/config.yaml
• Preferences: {vault_path}/.khanote/preferences.yaml

The vault_path is recorded in config.yaml under vault_path.