| name | ems-scan |
| description | Use when scanning communities for early market signals around a behavioral frontier, looking for pain, workarounds, and demand before a market category exists. Use when the user names a behavioral frontier and wants evidence gathered from web sources. |
EMS: Scan Behavioral Frontier
Overview
Scans communities and web sources for behavioral evidence around a named frontier. Updates source_map.csv and weirdness_ledger.md with evidence-backed entries. Separates demand signals from hype.
Inputs
- Behavioral frontier (required): a behavior statement, not a market category
- Weak: "AI agent market"
- Strong: "People delegating multi-step work to unstable agents despite reliability gaps"
When Not to Use
If there is no behavioral frontier yet, reframe before scanning. If the frontier sounds like a clean analyst category, push back and reframe it as awkward behavior.
Process
1. Validate the Frontier
The frontier must describe behavior. Awkward framing = early. Clean category framing = probably late.
2. Select Communities
Open community_tracking_targets.md. Pick 5-10 communities matching the frontier's pain type. Prefer issue trackers, support threads, and complaint-heavy forums over launch boards.
3. Search for Pain
Use query patterns from search_queries.md. Target words: manual, workaround, hack, script, broken, keeps forgetting, why is there no, copy paste, no API, alternative to.
Strong result: specific user, specific pain, current workaround, repeated by others
Weak result (skip): trend essay, launch announcement, market map, speculation without behavior
4. Update source_map.csv
Add one row per useful source with all fields:
community, url, topic_cluster, repeated_pain, workaround, evidence, tool_or_process_replaced, user_sophistication, monetization_hint, category_maturity, notes, next_action
Use exact quotes or tight paraphrases. "People are annoyed" is not enough.
5. Update weirdness_ledger.md
For each behavior that seems too intense for the product quality or category maturity, add an entry using the template already in that file. Every entry must answer: what behavior, why weird, what pain, what workaround, what evidence, what wedge.
6. Report Changes
Updated:
- source_map.csv: added N sources
- weirdness_ledger.md: added S___ entries
Strongest signal:
[one sentence]
Evidence:
[source links, quotes, workarounds]
Next action: ignore / watch / score / interview users
Evidence Standards
Strong:
- "I stopped using X for this."
- "I built a script to make this bearable."
- "This breaks every week but I still use it."
- "I made a guide because everyone kept asking me."
- "I would pay if someone handled this end to end."
Weak (do not log):
- "Looks cool." / "This is the future." / "Huge market." / "Lots of likes."
Common Mistakes
- Scanning 20 communities at once: start with 3-5
- Treating hype as demand: likes and launches are not behavior
- Vague notes: capture exact workarounds and replacement targets
- Mixing inference with evidence: keep them in separate fields
