What this system cannot see
A finding that says “three outlets have not covered this” only means something against a stated universe. This is the statement, including the parts that are unflattering.
Every figure below comes from 957 records retrieved from 5 public sources. 15 further sources are deliberately not connected.
No public-conversation source is connected. Everything this system reports about a live story is what was published about it.
Media attention and public opinion are different quantities. A story can dominate coverage while almost nobody discusses it, and a grievance can run through a district for weeks without a single outlet noticing. DRISHTI is built to tell those two situations apart — and in live mode it currently cannot, because only one of the two channels is connected. Every live finding carries this limitation as a named blind spot rather than a footnote.
No category is covered well. Every connected category is partial, and the list beside this one says how.
- Official Telanganatelangana.gov.in press releases (RSS, Telugu and English)
- Local Telugu mediaNTV Telugu — Telangana (full article text)
- English mediaThe Hans India — Telangana desk (full article text) · The Siasat Daily (summary paragraphs) · Deccan Chronicle — Southern States rail (one-line summaries)
- National mediaDeccan Chronicle national rail
- Government documentsGO portal (PDFs behind a search form) · Budget documents · Tabled answers
- Public social conversationX · Facebook · Instagram
- Video and YouTubeChannel feeds · Captions · Public engagement counts
- TelevisionBroadcast transcripts · Running order · Airtime
- Public grievance dataWard-level complaint aggregates
- Political statementsParty press releases with their own timestamps
- Search interestAggregate search indices
Press releases say what was decided without attaching the order that decided it. This is why no live claim in this build reaches the primary-document tier, and why VIGIL can rarely rule a live figure verified rather than merely supported.
The single most valuable gap. Signed orders are the only routinely available primary documents, and they are what allow a circulating figure to be checked rather than merely weighed.
One Telugu outlet is connected, which is enough to read the state in its own language and not enough to characterise how it is covered in it. Eenadu and Sakshi both serve public endpoints, and both carry titles without article text; the refusal below says so precisely, because Phase 2 recorded them as needing a licence and that was wrong.
Three outlets is enough to notice a disagreement and not enough to characterise coverage. Any statement of the form 'the media said' should be read as 'these three outlets said'. Only one of the three syndicates full article text, and a propagation edge can only be evidenced where text exists to compare.
Nothing in the live view measures what citizens are saying. Every live finding describes what was published. Media attention and public opinion are different quantities and this system will not present one as the other.
youtube.com/robots.txt disallows the feed endpoint under the wildcard group. Community video is usually where a local issue surfaces first, so this gap costs early warning specifically.
NTV Telugu's website is connected; the channel is not. A broadcaster's web desk is not evidence that anything was broadcast, and this system will not treat it as such. A story reaching television is a distinct escalation step and no live chain here can show it.
Without the wire, twenty outlets carrying one agency story look like twenty independent origins. Syndication detection on live data is weaker than on the demonstration corpus for exactly this reason.
Administrative counts are the closest thing to a primary record of citizen experience. Their absence is why a live complaint can be described but not corroborated.
No party in Telangana publishes a lawfully retrievable press-release feed — all three major parties were checked directly and the results are recorded below. Opposition and government argument reaches this system only as reported speech in media, which is a later and weaker record than the statement itself. Every response and counter-response hop in a live chain carries that limitation.
Would be a relative index within a query cluster in any case — never comparable across topics, and never a measure of opinion.
Each was evaluated and refused. Where a publisher’s robots.txt names this system, the refusal is not a judgement call. Where it blocks AI crawlers without naming us, a literal reading would have permitted the fetch and it was honoured anyway.
Evidence robots.txt groups ClaudeBot, Claude-Web, Anthropic-ai, GPTBot, CCBot and others under a single `Disallow: /`.
What it would add A Telangana-specific feed from a national paper of record, with strong coverage of irrigation, education and the legislature.
Evidence robots.txt contains `User-agent: Claude` followed by `Disallow: /`.
What it would add Independent south-India reporting with a distinct editorial position from the outlets connected.
Evidence robots.txt contains `User-agent: ClaudeBot` and `User-agent: anthropic-ai`, each followed by `Disallow: /`.
What it would add A second national desk covering Telangana, useful mainly for detecting syndication.
Evidence robots.txt blocks CCBot, GPTBot and ChatGPT-User with `Disallow: /` but does not name this system. A literal reading would permit the fetch under the wildcard group.
What it would add A high-volume Telangana-focused English feed — the single largest coverage gain available. Left unconnected because the publisher's evident intent is to exclude automated AI ingestion, and reading the omission of our name as permission would be lawyering rather than listening.
Evidence youtube.com/robots.txt contains `Disallow: /feeds/videos.xml` under the wildcard group.
What it would add Video metadata and captions from community and news channels — the ecosystem where local issues usually surface first.
Evidence Public post retrieval requires a paid API tier or a licensed provider. No unofficial route is used.
What it would add Public conversation — the entire citizen-voice half of this product. Its absence is the single largest limitation on every live finding.
Evidence Item links are Google redirect URLs and descriptions contain only an anchor tag, not article text.
What it would add Broad aggregation across publishers. Rejected because a finding could not be traced to an original public URL, which is a requirement rather than a preference.
Evidence robots.txt permits the wildcard group and a public feed exists at /rss.xml, but its items carry a title, an author token and a timestamp only — the description element repeats the headline and the byline, and content:encoded is absent. Phase 2 recorded this source as needing a licence, which was wrong: the feed is public. It is refused because it carries no article text, not because of policy.
What it would add A second Telugu daily. Headline-only records could be counted but never quoted, verified or used as evidence of propagation, which is what this build needs Telugu sources for.
Evidence robots.txt explicitly allows /sitemap-news.xml, which carries URLs, titles and lastmod timestamps but no article body. No RSS or Atom feed was located. Obtaining text would mean fetching article pages, which this system does not do for any publisher.
What it would add The largest Telugu daily in the state, and district editions no connected source reaches.
Evidence robots.txt is a single wildcard group with `Allow: /` and names no AI agent, so retrieval would be permitted. The main /feed carries timestamps and canonical URLs but empty descriptions, and no Telangana-section feed exists. Assessed and not added: it would repeat the Deccan Chronicle problem — a national rail with little Telangana content and no text to reason over.
What it would add Little that is not already covered. Recorded so the assessment is visible rather than the absence being read as an oversight.
Evidence No public syndication feed exposing article text; commercial licensing required.
What it would add Further Telugu-language depth, including district pages.
Evidence No robots.txt is served (the path returns an HTML error page), so retrieval would be permitted. The RSS endpoint at /RssMain.aspx returns items carrying a title and a link and nothing else — no pubDate, no description, no body — for every language and region parameter tried. A record with no publication timestamp cannot be placed in a chronology, and a chronology is the whole of what this phase is for.
What it would add Union government releases touching Telangana — the department-response tier at the centre, and the counterpart to a state release on any shared subject.
Evidence brsparty.in is a site-builder page whose robots.txt disallows only /404; it exposes no press-release listing, no feed and no dated items. telanganacongress.in serves a client-rendered application with no public data endpoint. telanganabjp.org does not resolve. All three were checked directly.
What it would add Political statements in the parties' own words, with their own timestamps. Their absence is why the response and counter-response hops in any live narrative can only be observed as reported speech in media, which is a weaker and later record than the statement itself.
Evidence robots.txt permits everything but /search. The site is a client-rendered application and the CKAN-style API paths tried returned the application shell rather than JSON. No documented public API was found.
What it would add Administrative datasets — the primary-record tier for figures that are currently only ever quoted.
Evidence robots.txt is `User-agent: * / Allow: /`, so policy permits retrieval — this refusal is technical, not a matter of permission. Orders are reachable only by submitting an ASP.NET postback search form; there is no feed, no dated listing endpoint and no stable item URL obtainable without driving that form, which is scraping a search interface rather than consuming a published endpoint.
What it would add The primary-document tier. Without signed orders, VIGIL can rarely raise a live claim above 'supported' — every live verification in this build is limited by exactly this gap.
Every source has coverage only from 2026-09-01 19:57 — 3 of 5 feeds expose recent items densely and older ones not at all. Volume before that point reflects what was retrievable, not what was published, and is not compared.
- media.hansindiahistory truncated by feed depth60 records from 2026-09-01
- media.siasat59 records from 2026-08-18
- media.deccanchroniclehistory truncated by feed depth45 records from 2026-08-31
- media.ntvteluguhistory truncated by feed depth21 records from 2026-09-01
- tg.gov.press12 records from 2026-08-21
- What the public thinks. No public-conversation source is connected.
- Whether a change in volume before the comparison horizon is real. Feed retention shapes the record count, not events.
- That any event caused any change. Findings say 'followed by' and mean it.
- Anything about a district with no attributable record. Under-collection and quiet are indistinguishable from here.
- How Telangana reads in Telugu. Only one connected source publishes in it.
21% of retrieved records were admitted to the pipeline; the rest named no Telangana entity or place. Watch what enters the system → How every figure is computed →