One substrate, the entire data-systems landscape
Every row is a category the industry built a separate engine for, paired with its signature public benchmark. One 64-byte-container substrate does them all in one pass. Live now or loading, every number measured on data you can download.
★ collapse
fused query
graph + vector + search in ONE pass, one copy
vs the entire stack
run ▸
graph
web-Google
6-hop deep traversal + provenance
vs graph databases
run ▸
graph
LiveJournal
69M-edge social graph k-hop
vs graph databases
run ▸
graph
com-Orkut
117M-edge community graph k-hop
vs graph databases
run ▸
search
MS MARCO
index-free full-text scan, zero index
vs search engines
run ▸
vector
GloVe-6B
binary Hamming recall, 32× density
vs vector databases
run ▸
key-value
800K keys
seqlock RAM-ring GET, no separate cache
vs in-memory KV stores
run ▸
SQL
TPC-H
scan + filter + group-by aggregate
vs SQL databases
coming
OLAP
ClickBench
full-corpus columnar aggregation
vs columnar warehouses
coming
graph
Panama Papers
offshore-leaks investigation traversal
vs graph databases
coming
document
Stack Exchange
nested-JSON field query, no schema
vs document stores
coming
wide-column
YCSB
the cloud-serving read/write workload
vs wide-column stores
coming
RDF
Wikidata
billion-triple SPARQL-style traversal
vs triple stores
coming
time-series
crypto ticks
as-of join + VWAP + windowed OHLC
vs tick databases
coming
object
Common Crawl
content-addressed blob PUT/GET
vs object stores
coming
geospatial
OpenStreetMap
spatial range + nearest scan
vs geo databases
coming
embedded
db_bench
the LSM read/write workload
vs embedded LSM stores
coming
streaming
append-log
native append-only event log
vs log / stream brokers
coming