Methodology and AI Disclosure
How NadiAI works.
NadiAI uses automation to filter a large stream of source items into a small number of concise briefings.
Updated: 17 July 2026
Important: NadiAI briefs are generated automatically and are not routinely reviewed by a human editor before publication.
1. Collection
The system collects titles, excerpts, dates, and links from selected RSS feeds and supported official sources. Collection does not mean an item will be published.
2. Automated quality scoring
Each draft is scored for AI relevance, source quality, freshness, amount of source material, practical importance, and Malaysia or Southeast Asia relevance.
Items may be automatically rejected when they are old, promotional, too thin to summarise safely, weakly related to AI, repetitive infrastructure coverage without regional relevance, or below the configured quality threshold.
3. Limited publication
Qualifying stories are ranked by score. NadiAI automatically publishes no more than five articles per weekday. If fewer than five items meet the threshold, fewer are published. Automatic article publishing pauses on weekends.
4. AI summarisation
An AI model creates a concise title, summary, category, tags, and “Why It Matters” statement using the supplied source metadata. The model is instructed not to introduce names, numbers, dates, availability claims, or product claims that are absent from that material.
These safeguards reduce errors but cannot eliminate them. Summaries may omit context, misunderstand ambiguous source language, or repeat an inaccurate source claim.
5. Sources and attribution
Each briefing identifies and links to its original source. The source is responsible for its underlying reporting or announcement; NadiAI is responsible for the automated summary shown on this site.
6. Corrections
Readers should verify important decisions with the original source. If a NadiAI summary appears inaccurate, please use the corrections process.