HashAI (HASHAI) — the network, measured

Generated 2026-09-02T17:54:29Z from live ClickHouse queries against this estate's ledger. Held inventory: 7,403 observations · 32 distinct series · 3 upstream sources · first row 2026-08-09 17:15:20 · newest row 2026-09-02 17:49:30 · 7,403/7,403 rows carry a verify_url. 32 series refreshed in the last 24h (726 fresh rows). Nothing on this page is cached, estimated, or remembered.

Provenance — where every number comes from

3 source families feed this network's ledger. Every row they wrote carries its own source, as_of and verify_url; this table is the live rollup of that.

30-day history — headline series (daily last-value, live)

One row per series: latest value, a 30-day sparkline of daily last-known values (scaled to the series' own min→max), day count, and window change. Series absent from the 30-day window are omitted from this table, not zero-filled.

Market reference — CoinGecko coin page, as held

Market-data reference rows: price, supply, capitalisation, diluted valuation, and the published change columns including ATH/ATL. These are the coin-page reference's own numbers — the estate does not model, smooth, or forward-fill them. The change columns / rank / ATH-ATL / high-low series start later in the archive than the price rows (each metric's caveat in the full inventory states its own first-held row) and are shown from their first held row. The circulating, total and max supply rows are different quantities (circulating vs total vs cap) and are never netted against each other or against the Blockscout onchain_total_supply chain-side read in the chain-state section below — different measurements from different endpoints, shown side by side. The token trades far below one cent; the sub-cent cells are rendered to full precision.

TVL — DefiLlama protocol page, as held

Two buckets of the same DefiLlama /protocol/hashai response: the protocol-page TVL (tvl[] series' last totalLiquidityUSD, reading 326 at generation) and the staking bucket (currentChainTvls.staking, reading 59,413.5142). They are different measurements — the general TVL bucket is currently very small as DefiLlama reports it while the staking bucket is orders of magnitude larger — and both are published exactly as answered, never netted, summed, or reconciled against each other. A small value here is the upstream's live answer, not a gap (both series' 30-day shapes are shown in the history table).

Token & chain counters — Blockscout, as held

Chain-side reads for the Ethereum-mainnet contract 0x292fcdd1b104de5a00250febba9bc6a5092a0076 — the catalog's own token_address field, verbatim (catalog chains field: ethereum) — from the public Blockscout v2 API. This source is deliberately distinct from any RPC family so its freshness identity stays untouched. At generation the chain-side total-supply read is 100,000,000,000, holders 29,377, transfers 611,316. Polled roughly once per day; counters are published raw, never differenced into rates, never extrapolated. The top-holder balance is the largest single holder's balance — an ownership-concentration input, not a supply figure, and never netted against supply reads.

Full metric inventory — 32 measured series, each with provenance

Every series held for HASHAI with rows in the last 30 days: latest value, when it was observed, which source observed it, how many rows we hold, how many of the last 30 days it reported, its comparability grade, and the origin link. "days/30" counts distinct dates with at least one row in the window.

Revision ledger — 0 backward revision(s) recorded for HASHAI

Backward revisions are values upstream later replaced for a date we had already archived. The before-value is retained — that retention is the only reason a rewrite is detectable, and it is the part every overwriting vendor discards.

The ledger query ran during generation and returned 0 rows for HASHAI. Zero here is the live query result, not a gap — the ledger itself is at /v1/revisions.

Estate grades for HASHAI (graded 2026-08-20 13:54:15.000)

Served by /v1/network/HASHAI — the same grades a paying agent receives. Scores are the engines' own 0–100 outputs, not ours to smooth. They are published exactly as the engines emitted them, unflattering parts included: the revenue F reflects that the estate holds no fees/revenue series for this network (the absence is shown in the sections above, not papered over), and the data-coverage D reflects a thin measured-series set relative to first-party-telemetry networks. The node-quality and dev-velocity A grades sit on third-party and chain-side series alone. The grade stamp is dated; the engines re-grade on their own schedule.

Verify any cell yourself — the exact commands

No account, no key, no trust required. Commands 1–3 hit the same upstream endpoints our collectors do (the coingecko.com browser coin page 403s plain curl, so the programmatic markets endpoint is given; the Blockscout commands use the public eth.blockscout.com instance the verify_urls point at); command 4 hits this estate's public API, where every row carries the same source / as_of / verify_url fields. One polite call per endpoint, no hammering.

# 1) On-chain token record (Blockscout v2 token API — the collector's method), as of 2026-09-02T17:54:29Z:
curl -s -H "User-Agent: kairos-check" "https://eth.blockscout.com/api/v2/tokens/0x292fcdd1b104de5a00250febba9bc6a5092a0076" | python3 -c "import json,sys; d=json.load(sys.stdin); print('total_supply', d.get('total_supply')); print('holders', d.get('holders')); print('transfers', d.get('transfers'))"
# 1b) Token counters (transfers + holder count):
curl -s -H "User-Agent: kairos-check" "https://eth.blockscout.com/api/v2/tokens/0x292fcdd1b104de5a00250febba9bc6a5092a0076/counters" | python3 -c "import json,sys; d=json.load(sys.stdin); print('transfers_count', d.get('transfers_count')); print('token_holders_count', d.get('token_holders_count'))"
# 1c) Contract address counters (transactions + gas usage):
curl -s -H "User-Agent: kairos-check" "https://eth.blockscout.com/api/v2/addresses/0x292fcdd1b104de5a00250febba9bc6a5092a0076/counters" | python3 -c "import json,sys; d=json.load(sys.stdin); print('transactions_count', d.get('transactions_count')); print('gas_usage_count', d.get('gas_usage_count'))"
# 1d) Top holder (raw largest-holder balance):
curl -s -H "User-Agent: kairos-check" "https://eth.blockscout.com/api/v2/tokens/0x292fcdd1b104de5a00250febba9bc6a5092a0076/holders?page=1" | python3 -c "import json,sys; d=json.load(sys.stdin); items=(d.get('items') or []); v=((items[0].get('value') if items else None)); print('top_holder_raw_value', v)"
# 1e) The per-token page our verify_url points at (human view):
curl -s -o /dev/null -w "%{http_code}\n" "https://eth.blockscout.com/api/v2/tokens/0x292fcdd1b104de5a00250febba9bc6a5092a0076"
# 2) Protocol TVL + staking bucket (DefiLlama public API, same /protocol response the rows read):
curl -s -H "User-Agent: kairos-check" "https://api.llama.fi/protocol/hashai " | python3 -c "import json,sys; d=json.load(sys.stdin); t=(d.get('tvl') or [{}]); print('tvl_usd', (t[-1] if t else {}).get('totalLiquidityUSD')); print('staking', (d.get('currentChainTvls') or {}).get('staking'))"
# 3) Market reference (CoinGecko public API):
# CoinGecko public markets API (the endpoint family our collector
# reads for the price/supply/cap rows) — the coingecko.com browser
# coin page 403s plain curl, so the programmatic endpoint is given:
curl -s "https://api.coingecko.com/api/v3/coins/markets?vs_currency=usd&ids=hashai" | python3 -m json.tool
# 4) This estate's public API row for HASHAI:
curl -s "https://kairossignal.com/v1/network/HASHAI" | python3 -m json.tool | head -60

Method & honesty notes. All counts are queries against the live ledger at generation time, not restamped copy — this page is regenerated by gen_deep_dive_hashai.py and the hourly claim_drift_check.py run grades its claims against the live API like every other published surface. Comparability grades come from the served canonical-metric system via its own resolver: total_supply maps to token_supply_total (question-verified). The remaining series are HASHAI-native market/on-chain metrics with no cross-network concept; grading them would be fabrication, so they are labelled "hashai-native (unmapped)". FOUR supply references are shown side by side as held, never summed, merged, or reconciled: CoinGecko's circulating, total and max supply, and the Blockscout chain-side token supply. At generation the chain-side read equals the CoinGecko max-supply reference; they remain different measurements from different endpoints and no ratio is derived between any of them. The CoinGecko circulating_supply row is a different quantity (circulating vs total) and is never netted against the other reads. The DefiLlama general TVL bucket reads a small non-zero value while the staking bucket of the same response is orders of magnitude larger — two different buckets, both published as answered, never netted or summed, and the small value is the upstream's live answer, not a gap. The top-holder balance stays the raw largest-holder balance as the API reports it — an ownership-concentration input, never relabelled as supply, never presented as investment advice. The Blockscout counters are ~daily polls, published as answered. The catalog's token_address field is filled for HASHAI (0x292fcdd1b104de5a00250febba9bc6a5092a0076) and the chain-side family reads exactly that contract, with every verify_url pointing at eth.blockscout.com for the same address. The estate holds no GitHub, no DefiLlama fees slice, no RPC eth_call rows, and no first-party snapshot rows for HASHAI — those sections are absent rather than zero-filled, and the queries rerun at every regeneration. Where a section has no data, it is absent — we do not publish zeros for gaps. Upstream probes ran serially by this generator immediately before rendering: DefiLlama protocol API (hashai) HTTP 200; CoinGecko markets API (hashai) HTTP 200; Blockscout token API HTTP 200; Blockscout token counters API HTTP 200; Blockscout address counters API HTTP 200; Blockscout token holders API HTTP 200; estate /v1/network/HASHAI HTTP 200.