!
Internal review only

This page is for the party / IT Wing review. It will be removed (or simplified) before public launch. The operational pipeline below describes how the system works behind the scenes.

Methodology

How this works

Every promise on this site is sourced, dated, and reviewed by a human before being published. Below is the exact process we follow so you can judge for yourself whether to trust what you read here — and how cadres respond fast when misinformation spreads.

The 30-minute response loop

From misinformation detected → coordinated authentic reply in 30 minutes

Most political response today takes 4–8 hours. By then the narrative is set. This pipeline compresses it to 30 minutes — every minute saved is a minute the opposition's narrative is uncontested.

T+0 minstep 1
Detect

AI scrapers monitor news, X handles, YouTube, govt feeds every 10–15 min.

T+10 minstep 2
AI score

Gemini scores newsworthiness, sentiment, credibility. Flagged items → review board.

T+15 minstep 3
Human review

IT Wing verifies facts, edits Tamil briefing, taps Publish.

T+30 minstep 6
Reach

Thousands of authentic posts hit X — each in a different voice. No detection.

T+25 minstep 5
Cadre speaks

Cadre opens briefing, taps mic, speaks in Tamil. Their voice becomes the post.

T+20 minstep 4
Fan-out

WhatsApp + push to every cadre. Average cadre opens within 2 min.

End-to-end: under 30 minutes from detection to coordinated reply

Tracking process

The accountability layer that runs in parallel with the response loop above.

1

Discovery

We monitor public sources: party manifestos, official speeches, government press releases, mainstream Tamil and English news, and verified social media accounts. Items are caught by automated scrapers running every 10–60 minutes.

2

Source linking

Every promise gets a source URL where the party originally made the statement. If we can't link it, we don't track it. No paraphrasing into a promise the source didn't say.

3

AI-assisted triage

New items are scored by AI for newsworthiness, sentiment, and topic. AI surfaces what's worth attention. AI never decides what gets published.

4

Human admin review

A human admin reviews every status change. They check the source, read the evidence, and decide whether to mark a promise as in-progress, delivered, broken, or disputed. Nothing reaches this page without that review.

5

Conservative defaults

When in doubt, we under-claim. A promise stays "not started" until verifiable evidence of motion. "Delivered" requires concrete proof, not announcement.

6

Public corrections

When we make mistakes, we log them on the Corrections page with the date, reason, and original source. Status reversals are visible to everyone.

Our principles, in plain words
  • • Truth before speed. We don't race to publish.
  • • No fabrication. No invented quotes, statistics, or framings.
  • • Same standard for every party.
  • • No targeting of private individuals. Public figures and public actions only.
  • • No dehumanizing language. No caste-based attacks. No religious bigotry.
  • • Methodology and corrections are public, always.