Every dataset, with an agent in front of it

Every dataset below lives in the one substrate and has its own conversational agent. Open one and ask it anything; the agent runs the real query live and shows you the measured number on data you can download. Sharing a dataset is sharing its link: hand someone the agent and you have shared the data through it.

graph · 5

typed-edge CSR: k-hop traversal, shortest path, connected components, pagerank

3.07M nodes · 117.19M edges
Stanford SNAP

soc-LiveJournal1

4.85M nodes · 68.99M edges
Stanford SNAP

Wikimedia Commons objects

10.0M nodes · 30.0M edges
Wikimedia Commons (image metadata + category links)

Yelp Open Dataset

8.6M nodes · 13.0M edges
Yelp Open Dataset

875.7K nodes · 5.11M edges
Stanford SNAP

text · 3

contiguous text runs: index-free full-text scan, the moment data lands

NASA-HTTP access log

30.0M documents
NASA / Internet Traffic Archive

GeoNames place text

12.0M documents
GeoNames (allCountries)

500.0K documents
Microsoft

table · 10

one container per row: GROUP BY, aggregate, point read, full-table sweep

db_bench 100M keyspace

100.0M rows
RocksDB db_bench keyspace

TPC-H lineitem (SF=10)

59.99M rows
TPC-H dbgen

TPC-H star (SF=10)

59.99M rows
TPC-H dbgen

CTR feature rows

50.0M rows
generated CTR feature table

Binance BTC aggTrades

50.0M rows
Binance public aggTrades

TSDB metric samples

50.0M rows
generated metric series

NYC yellow-taxi trips

50.0M rows
NYC TLC

ClickBench hits

30.0M rows
ClickHouse ClickBench

GeoNames allCountries

12.0M rows
GeoNames

10.0M rows
Yahoo! Cloud Serving Benchmark

rdf · 1

typed-edge triples: the SPARQL property-path closure as a deep bounded traversal

Wikidata truthy (P31/P279)

100.0M nodes · 200.0M edges
Wikidata RDF dump (truthy claims subset)