NammaPedia reads the live web, primary documents and regional reporting, then writes one answer — every claim tied to its source, every disagreement left visible.
Trending now, from today's NammaFeed
Live NammaFeed
Compressed to what changed, what it means, and what is still open.
Indian AI labs are shipping smaller models that hold their own on retrieval benchmarks
Three sub-10B models released this month are tuned for document retrieval rather than general chat.
Two labs publish per-task benchmarks; the third has released no evaluation set.
Synthesized from 6 primary sources
11 min ago
Karnataka · PolicyBengaluru
A state incentive package cleared cabinet — what it actually changes for local manufacturing
Eligibility thresholds, subsidy ceiling and application window, read from the cabinet note itself.
Press coverage of the package omits the phased disbursement schedule in the annexure.
Synthesized from 4 primary sources
38 min ago
Global · InfrastructureGlobal
Compute export rules shifted again. Here is what changed in the text
Clause-level diff of the new rule against the previous one, limited to non-US cloud capacity.
How enforcement will apply to joint ventures remains unresolved in the filing.
Synthesized from 5 primary sources
1 hr ago
How the answer is built
Retrieval, verification, synthesis. The order is the product.
01
Real-time retrieval
Live web results plus a primary-source index: government gazettes and notices, court filings, official blogs, preprints and earnings documents.
02
Cross-source verification
Each claim is matched against independent sources. Where they disagree, the split stays visible instead of being averaged into a smooth non-answer.
03
Structured synthesis
One answer, inline citations, gaps labelled. Syndicated copies and SEO aggregation are demoted rather than summarised.
The line we hold: if a claim can't be traced to a retrievable source, it doesn't go in the answer. Paywalled sources are marked as inaccessible rather than inferred, and thin coverage is labelled thin.