Grounding

Ideas grounded in real markets

Full in-app AI generation is on the roadmap. Today you can copy a structured briefing that points any LLM at the same JSON we use: industries, the kill library on /verdicts, and research notes.

Browse Map for industry scores, then ask your model for niches inside the highest-opportunity verticals.
Study Idea verdicts; killed ideas with reasons are as useful as Strong Potential.
Read Research deep dives for market evidence behind the map.

LLM grounding snippet

Atlas of Ideas, generator grounding (paste into your LLM)

Datasets (repo paths for humans; product loads JSON at build/runtime):
- Industries: src/data/industries.json, 24 industries with current/future market size (billions USD), growth rate (%), problem severity (0–10), competition (0–10), nested problems[] → saasIdeas[], sources[], subreddits[].
- Scored ideas / kill library: src/data/ideas.json, 112 ideas with tier (Strong Potential / Killed), killReason, call (Pursue/Park/Kill), summary, reasons[], nextStep, Solo Code benchmark. Strong Potential: 6. Prefer learning from Killed verdicts (why not to build) as much as survivors.
- Research deep dives: content/research/*.md, industry deep dives and evidence notes (also on /research).

Opportunity Index (0–100, higher is better; same as /map and industry pages):
  marketPts   = min(currentMarketSizeB / 40, 10) * 3     // up to 30
  growthPts   = min(growthRatePct / 5, 10) * 3            // up to 30
  painPts     = problemSeverity * 2.5                     // up to 25
  crowdPenalty = competitionScore * 1.5                   // up to 15, subtracted
  index = clamp(0, 100, round(marketPts + growthPts + painPts - crowdPenalty))

Bands: Strong 45–100, Promising 32–44, Mixed 20–31, Tough 0–19.

Solo Code gate (idea validation): ≥6/7 criteria to pursue; fail Painkiller or ≤4/7 → kill signal.

User focus for this session:
(none, pick an industry on /map or a verdict on /verdicts)