Every article on nobl.rb is generated by our AI engine. We believe in transparency — and we believe readers make our engine smarter. Community Signals is our crowdsourced editorial intelligence layer: your signal trains the engine that writes the next article.
Each article has three signal options. You choose one based on what you read:
✓ Data Valid — The information in the article is accurate and consistent with actual market conditions.
~ Partly Off — Some information is correct, but parts are missing context, slightly inaccurate, or misleading.
✗ Hallucination — The article contains fabricated information, invented price levels, or claims that cannot be verified.
Based on community signals, articles may receive a badge:
Articles under review remain visible — we believe in transparency, not censorship. The badge informs readers to exercise extra caution.
We use confidence-weighted thresholds. A small number of votes cannot unfairly label an article. Thresholds tighten as more readers signal:
| Voters | Tier | Under Review if | Reader Verified if |
|---|---|---|---|
| Less than 10 | Insufficient | No badge — not enough data | |
| 10 – 49 | Low Confidence | ≥ 40% Hallucination | ≥ 70% Valid |
| 50 – 99 | Mid Confidence | ≥ 30% Hallucination | ≥ 60% Valid |
| 100+ | High Confidence | ≥ 25% Hallucination | ≥ 50% Valid |
When an article reaches the Under Review threshold, our editorial memory system extracts patterns from that article. These patterns are sent to the AI model as warnings on the next generation cycle — teaching it to avoid the same mistakes.
This creates a feedback loop: reader signals → editorial memory → better articles → fewer signals needed.
We do not store your IP address. Your IP is hashed (one-way encrypted) before storage and is only used to prevent duplicate votes. It cannot be traced back to you.
One signal per reader per article. You may change your signal at any time.
v1 — Live Anonymous signaling for all readers
v2 — Planned Reader Verified badge unlocks premium content access for active signalers
v3 — Planned Weighted signals — readers with accurate track records carry more weight
AI-generated content will inevitably contain errors. Rather than hiding this fact, we built a system that turns reader skepticism into engine intelligence. Every time you flag a hallucination, you make the next article more accurate — for yourself and every reader after you.