Topic Brainstormer
Generate blog topic ideas that align with a content identity built around solving frustrating technical problems. Three phases: ASSESS, DECIDE, GENERATE. Every topic must pass the three-question content quality filter before presentation.
Deep References
| Signal | Load | Content |
|---|---|---|
| Filtering topics: three-question test, category examples | references/content-filter.md | Full quality filter with green/red/yellow topic categories, salvage patterns |
| Scoring: impact, vex, resolution matrix | references/priority-scoring.md | Rubrics per dimension, score ranges, calibration, worked examples |
| Mining: problem sources, gap analysis, expansion | references/topic-sources.md | Mining prompts, signal strength, gap types, expansion strategies |
Phase 1: ASSESS
Gather context about existing content and available sources.
- Read all posts in the content directory. Document post count, content clusters, technologies covered, last post date.
- Identify available sources: problem mining (recent debugging, errors, config struggles), gap analysis (cross-references leading nowhere), tech expansion (adjacent technologies).
- Extract all "see also" and cross-reference mentions. Flag any pointing to content that does not exist.
Gate: Content landscape documented, at least 2 sources identified with material.
Phase 2: DECIDE
Generate candidates and filter through the content quality test.
- Mine 5-10 raw candidates from at least 2 sources. Capture: source type, raw topic, initial vex signal.
- Apply the content quality filter -- every candidate must answer YES to all three:
- Was there genuine frustration? Real time lost, failed attempts, unclear docs, unexpected behavior.
- Is there a satisfying resolution? Clear fix, understanding gained, prevention strategy, "a-ha moment."
- Would this help others? Reproducible problem, actionable solution, relatable frustration.
- Reject failing candidates. Document each rejection: topic, failed question, reason.
Failure modes to catch:
- Tutorial-only topics: "How to Set Up X" with vex listed as "learning a new tool" is not genuine frustration. Find the specific friction. "Hugo Local Build Works But Cloudflare Deploy Fails" has real vex.
- Opinion without experience: "Why Go Is Better Than Python" is debate, not experience. Ground in measurement.
Gate: At least 3 candidates pass. Fewer than 3 -- return to Step 1 with different sources.
Phase 3: GENERATE
Score, prioritize, and present the filtered list.
Scoring: Priority = Impact(1-5) x Vex(1-5) x Resolution(1-5). Thresholds: 60-125 HIGH, 30-59 MEDIUM, 15-29 LOW, 1-14 SKIP.
Titles: Replace vague categories with failure-mode titles. Bad: "Kubernetes Networking Issues". Good: "Pod-to-Pod Traffic Works But Service Discovery Fails".
Output format: Grouped by priority tier. Each topic: title, vex, joy, content cluster fit, word estimate, score breakdown. End with recommendations: top pick, quick win, deep dive.
Tie-breaking: prefer topics that fill an existing gap, complement recent posts, use already-covered technologies, have clearer narrative structure.
Gate: All topics scored, prioritized, presented with recommendations.
Angle Lenses Mode
When the task asks for angles on a single topic (not fresh candidates), apply these six lenses instead of the content quality filter:
| Lens | Question |
|---|---|
| Perspective shift | Whose view changes the story? |
| Ladder of abstraction | One rung up (the trend) or down (the single case)? |
| News values | Which value carries it: conflict, proximity, novelty, impact, human interest? |
| Data angle | What does the dataset say that the narrative misses? |
| Contrarian | What if the consensus framing is wrong? |
| Timeliness peg | Why now? What event or deadline makes this urgent? |
Rules: one lens per kept angle (distinctness is structural). Each kept angle must pass the "so what?" gate -- one sentence naming a concrete reader payoff. Log refused angles with reason.
Error Handling
| Error | Cause | Solution |
|---|---|---|
| No existing posts | Content directory empty | Focus on problem mining; ask about recent debugging sessions |
| All candidates fail filter | Sources lack frustration signals | Ask probing questions: "What broke recently?" Shift sources. |
| Topic too broad | Category, not a specific problem | Break into failure modes: "[A] works but [B] fails" |
| Resolution unclear | Ongoing issue, no fix yet | Defer until resolved; or assess if "understanding the workaround" suffices |
