Asia Pulse
Asia Pulse/Transparency

Transparency

Last updated 2026-11-17 · machine-readable: /transparency.json

Who runs Asia Pulse, what it sells, where it is listed, and how its record can be checked on-chain. Every value on this page is generated from the live files it describes.

Operator

Asia Pulse, Korea. Contact: contact@asia-pulse.com (same as security.txt).

What we sell

Seven JSON endpoints, paid per call over x402. Prices and payment details match /.well-known/x402.

EndpointPriceWhat it is
GET /report$0.03Asia session regime report
GET /report/premium$0.10Premium Asia-session intelligence
GET /chain/semis$0.10Korea semiconductor equity chain
GET /chain/defense$0.10Korea defense equity chain
GET /chain/power$0.10Korea power equity chain
GET /chain$0.20Korea equity chain bundle
GET /japan$0.10Japan market ledger
NetworkAssetpayTo
X Layer (196)USD₮0 (USDT0)
0x779ded0c9e1022225f8e0630b35a9b54be713736
0xebe771d3fc1b9d55e3f36abc9e9c1c45952bc14d
Base (8453)USDC
0x833589fCD6eDb6E08f4c7C32D4f71b54bdA02913
0x30c13c7fc1d975e6fc3af5d5154081cb5e52ff54

Where we are listed

ListingLink
x402 Bazaar (Coinbase Agentic Market)agentic.market/services/asia-pulse-com
OKX.AI agent 5537www.okx.ai/agents/5537
Official MCP Registry: com.asia-pulse/asia-pulseregistry.modelcontextprotocol.io/v0/servers?search=com.asia-pulse/asia-pulse
ERC-8004 agentId 13967 (X Layer)8004scan.io/agents/xlayer/13967

Measurement record

Measurement started 2026-10-01 00:00 KST. Everything recorded before that moment was sealed in a genesis manifest and its SHA-256 was written to X Layer.

ItemValue
Genesis manifest SHA-25617d2da59377ac4fed94793eb783b9ab4cf755ddf47e9b13593505256009b8b2e
Genesis transaction0x8417085374a6eb3736c28c3b81e62ed22dc5df84c5c3367edd985275f23827cf
block 72044341 · calldata “AP1G” + SHA-256
Daily recordEvery day 17:10 KST · calldata "AP1" + SHA-256 of the day's snapshot manifest (Merkle root when several days are batched)
Recording wallet0xb759Ea689FF49A6781232198fcAdD6fdfE0680cd
Latest daily record2026-11-17 · 0x84e4912f12b86e528a90bfd7ab01e9bdec8206993db3d03a70072a7230477822
Scoring rules MD5ceaf28a63f8907e7fe323db2234bc818
The scoring rules are private; only this MD5 is published (it is also sealed in the genesis manifest).

What is public, paid, and private

Public. Discovery files (/.well-known/*, llms.txt, openapi.json), research articles, this page, and the daily on-chain record hashes.

Paid. The seven JSON endpoints below, per call over x402 — no API key or subscription.

Private. Methods, models, code, raw inputs, and the text of the scoring rules (only their hash is published).

Incident log

2026-11-12 11:21 KST

2026-11-12 11:21–11:35 KST — Planned maintenance: the server was stopped for about 13 minutes to add memory and disk space. The website, MCP and OKX endpoints did not answer during this window. The payment ledger and stored records were not affected; their file hashes matched before and after.

Fix: Services were back at 11:35 KST and the post-restart checks passed.

2026-11-07 07:30 KST

2026-11-07 07:30–13:41 KST — Our source-matching check dropped a correct figure from the US jobs report result: "average hourly earnings +3.1% y/y" was removed, and "+165,000" was stored as "+165, 000". No incorrect numbers were served. One paid response (/okx/report, 11:21) carried the incomplete line. We fixed the check to recognise average hourly earnings and thousands separators and to recompute changes from the official BLS tables (Table B-1, B-3); the result was corrected at 13:41. We apologize for the inconvenience.

Fix: Result corrected 13:41 KST. Test I181

2026-11-04 09:25 KST

2026-11-04 09:25–17:43 KST — During a reboot for a kernel security update, the server's scheduler stopped responding. Our APIs kept answering and payments were processed normally, but scheduled refreshes paused until we restarted the server from the provider panel at 17:43 (about 70 seconds of downtime). As a result, /chain served the Nov 2 session (marked stale_warn) until 18:06, the analyst commentary stayed on the previous evening's version until 17:46 (briefly showing an empty-data version 17:44–17:46, with no paid responses in that window), today's brief was generated on Nov 2 prices, and the anchored Nov 4 snapshot (tx …aca1a2) does not include the brief; we have not altered the anchored record. Kimchi-premium history 09:05–17:44 KST and Japan intraday snapshots for Nov 4 are permanently missing. One paid response (/okx/report, 16:46) was served in this window. We have added boot-order safeguards and a reboot watchdog. We apologize for the inconvenience.

Fix: Boot-order safeguards and a reboot watchdog added on 2026-11-04.

2026-11-03 16:00 KST

2026-11-03 16:00 KST — The Korean public data portal (data.go.kr) was responding slowly and timed out for 6 of 39 tickers. As a result, /chain products were served on the 10/30 base (marked stale_warn) and AI explanation fields were withheld from 16:30 until recovery at 19:34. One paid /okx/chain/semis response at 18:25 was affected (payer …1460, tx …427a5a, 3 fields withheld, stale label included). The upstream delay was outside our control; on our side, retries stopped too early and no same-day re-collection ran. We have now made calculation wait for collection, added same-day retry and an immediate alert. No refund was issued; this is disclosed here. We apologize for the inconvenience.

Fix: Recovered 19:34 KST (re-collection, all 39 tickers on the 11/02 session). From 2026-11-03: chain calculation runs after collection ends; timed-out tickers are retried at the end of the run and again at 18:00 and 20:00 KST; immediate operator alert. Tests I155, I156

2026-11-02 23:19:11 KST

Base USDC payment for /report (0.03 USDC, payer …1978) was not completed: the CDP facilitator settle call returned HTTP 500. The product was not delivered (HTTP 402 returned). No funds moved (no on-chain transfer found). External cause (facilitator).

Fix: From 2026-11-03 a settle-stage HTTP 5xx or timeout sends an immediate operator alert. No automatic re-settlement. Test I148

2026-11-01 16:00 KST

2026-11-01–2026-11-08 — The time stamp on our daily X card image was one hour off (e.g. "6:35 AM ET" instead of "5:35 AM ET") after US daylight saving time ended, because the card used a fixed UTC offset. The numbers in the posts were not affected. Fixed on 2026-11-08; a pre-post check now compares every number and the time stamp with the source data before posting.

Fix: Time zone fixed in the X card code on 2026-11-08 with tests at both daylight-saving boundaries (I204); pre-post number check added (I202). Past posts were not edited.

2026-10-30 07:30 KST · corrected

2026-08-26–2026-11-07 — A review of our source-matching check against 16 official releases (Aug–Nov) found that it rejected 10 correct figures (e.g., the September FOMC +25bp to 3.75–4.00%, Q3 GDP 3.4%, core PCE 3.0%, retail sales ex-autos +0.9%). Eight were in results stored before the check began on 2026-10-17 and were not changed; in live use it removed the other two (Q3 GDP 3.4% and core PCE 2.5%), so the Q3 GDP result was served with only the PCE line from 2026-10-30 to 2026-11-05. No incorrect official figures were served. We fixed PDF and table parsing and basis-point matching on 2026-11-07, added tests, and did not rewrite past records. We apologize for the incomplete results.

Corrected 2026-11-10: start date 2026-08-26 → 2026-08-27; end date 2026-11-07 → 2026-11-06. The first of the 16 official releases in the back-test was published on 2026-08-27; re-running the back-test on 2026-11-10 confirmed that date. The last of the 16 official releases in the back-test was published on 2026-11-06; re-running the back-test on 2026-11-10 confirmed that date. The original entry above is left unchanged.

Fix: Check fixed and tests added on 2026-11-07 (I182–I187). Past records not rewritten.

2026-10-28 09:11 KST

2026-10-28 R1 가격·국면 입력 원천을 업비트·OKX 직접에서 CoinGecko로 변경(거래소 약관). 과거 채점은 그대로.

Fix: daily_record.py — BTC 시세 coingecko.okx_usdt · 김프 coingecko.upbit_krw+kimp · 펀딩 coingecko.okx_funding · 판정 kimp.classify_regime. 하나라도 못 받으면 R1 건너뜀(직접 조회로 되돌아가지 않음). 시험 I130

2026-10-25 15:05 KST

2026-10-25 김프 30분 기록 원천을 업비트·OKX 직접에서 CoinGecko로 변경(거래소 약관). 과거 기록은 그대로.

Fix: kimp_snapshot.py — CoinGecko 4콜/회 (업비트 KRW·OKX USDT·OKX 펀딩 9종·BTC 24h). kimp_archive.csv 같은 열에 이어서 기록. 원천 열 없음(새 열 만들지 않음). 시험 I126

2026-10-12 ~ 2026-10-13 21:50 KST ~ 08:36 KST

Model API credit ran out again (second time; the first was 2026-10-05 21:50 ~ 2026-10-06 17:36). Two scheduled jobs failed with an API billing error: the analyst report writer (2026-10-12 21:50) and the event calendar updater (2026-10-13 07:00). The analyst paragraph in /report/premium stayed at the 2026-10-12 09:50 version for about 22.8 hours and was served with its real generation time and a 'late' note; the event calendar kept its previous entries. No price data, ledger, score, snapshot or on-chain anchor was affected, and no payment was taken for a failed response.

Fix: The service API key was replaced on 2026-10-13 08:35 and a 5-token test call passed. Both jobs were rerun by hand and succeeded (08:36). Both failures raised an immediate red alert as designed. Follow-up: watch the API balance before it reaches zero (listed for a later deploy day).

2026-10-08 ~ 2026-10-09 13:10 KST runs

Attention ledger: the Korean Wikipedia article for the tracked drama '포헨즈' was deleted on 2026-09-30 and recreated as '포핸즈' without a redirect, so the old title stopped receiving page views. The 2026-10-08 collection missed one cell (ko:포헨즈, source date 2026-10-07); the 2026-10-08 snapshot and its on-chain anchor were left unchanged. Rows under the old title after the deletion (2026-10-01 = 0, 2026-10-04 = 2, 2026-10-06 = 0) come from the deleted page.

Fix: Collector now resolves the current title before fetching (redirect, move log, or the existing page named in the deletion log) and records changes in attention/wiki_title_moves.csv (2026-10-09). Values for 2026-10-02, 10-03, 10-05, 10-07 and 10-08 were fetched from the new title on 2026-10-09 13:57 KST (late fill, shown in collected_at).

2026-10-05 21:50 KST ~ 2026-10-06 17:36 KST

Anthropic API credit exhausted (external, billing). The /report/premium analyst text, the event calendar, sector events and the four chain briefs were not refreshed. No data was deleted. No effect on measurement or scoring.

Fix: Credit topped up; narrative, event calendar, sector events and chain briefs re-run 2026-10-06 17:34-17:36 KST. The 2026-10-06 daily snapshot sealed the 2026-10-05 text; snapshots are not modified.

2026-10-05 05:30 KST ~ 2026-10-06 09:58 KST

US federal awards block missing from /chain/defense after upstream connection resets; cache was overwritten with empty result

Fix: Re-collected from USAspending on 2026-10-06 09:58 KST (9 awards restored, no errors). Collector failure handling under review.

2026-10-02 06:41 KST

/report 1 request returned 502 during an automatic OS security update restart.

Fix: Our services excluded from auto-restart, weekly maintenance window (Sun 04:30 KST), 503 + Retry-After during restarts.

2026-10-01 ~ 2026-10-14 08:16 KST fix

Japan government bond history lost a whole month at the month change. The collector merged the Ministry of Finance 'current month' file only in memory and never wrote it to the local history file, so when the current-month file rolled over to October, all 19 September sessions disappeared (history ended at 2026-08-31). From 2026-10-01 to 2026-10-13 the /japan field jgb.change_1m was served as a 'one-month change' but was actually measured against 2026-08-31 (about 5 to 6 weeks). On 2026-10-14 the gap exceeded the collector's tolerance and the field was served as null with a 'history gap' note; the product contract check failed on three required fields. The curve itself, the spreads and the one-year change were correct throughout. Daily snapshots of 2026-10-01 to 2026-10-13 contain the mislabelled values and are left unchanged; anchors are unchanged.

Fix: Fixed 2026-10-14 08:16 KST. The 19 missing September sessions were restored from the Ministry of Finance full-history file (https://www.mof.go.jp/english/policy/jgbs/reference/interest_rate/historical/jgbcme_all.csv, fetched 2026-10-14 08:16 KST); all 13,290 dates already held locally matched that file exactly. The collector now appends each newly received day to the history file the same day, and refills from the full-history file when it finds a hole at a month boundary. After the fix change_1m is measured against 2026-09-09 as labelled. Regression test I78 added. A daily gap check over every history ledger is scheduled for 2026-10-16.

2026-09-14 ~ 2026-10-02

Kimchi premium coverage 7 of 8 (BONK delisted on Upbit KRW).

Fix: XRP added 2026-10-02.

2026-07-29 16:00 KST

2026-07-29–2026-11-03 — 12 daily X cards in the "days to effect" format featured U.S. Federal Register rules that are not about crypto (Financial Data Transparency Act joint data standards; OCC "Unsafe or Unsound Practices, Matters Requiring Attention"; Treasury "Federal Scholarship Tax Credit"). The agency counts on the card (e.g. "Treasury 8 · SEC 7 · …") also included filings that mention a crypto term only in the body text; with the crypto filter, the counts were 2–7 per post instead of 13–23. Our paid /report and /report/premium (also sold on OKX) had the same problem: in every daily snapshot from 2026-09-26 (our first snapshot) to 2026-10-30, the nearest effective date in the regulation section was a rule not about crypto (the OCC rule above, or on Oct 18–19 a CFTC swap-definition rule), the agency counts were unfiltered, and 3–8 of the 8 recent filings were not about crypto. The crypto filter has been applied to /report since 2026-10-31. Affected X posts (date, post ID on x.com): 07-29 2082360512990630298, 08-18 2089608273775476747, 09-06 2096493764726640737, 09-13 2099030354753712304, 09-19 2101204779838177316, 09-25 2103379101746917842, 10-01 2105553493130170667, 10-07 2107727832923386218, 10-13 2109901998015533454, 10-20 2112438743340876144, 10-28 2115337962217255393, 11-03 2117512249502060634.

Fix: The X card now uses the same crypto filter as /report, and the pre-post check blocks a card whose featured filing or counts are not in the filtered list (I209). Past posts and past API responses were not edited.

2026-07-23 17:56 KST

2026-07-23–2026-11-01 — Our daily X card showed the BTC funding rate 100× too high on Jul 23–31, Aug 1–18, Aug 20, Aug 22–29, Sep 1, 8, 17, 23 and 29, Oct 5, 11, 17, 25 and 31, and Nov 1, 2026 (48 posts; for example "+0.900%" instead of "+0.009%") because a value already in percent was multiplied by 100 again. Our API and paid responses were not affected. Fixed on 2026-11-08 with a test. We apologize for the error.

Fix: Unit fixed in the X card code and a regression test added on 2026-11-08 (I200). Past posts were not edited.

2026-07-20 ~ 2026-11-10 10:43 KST

2026-07-20–2026-11-10 — Our product descriptions presented the kimchi premium as the gap between Korean and global coin prices but did not say which exchange rate converts the prices. It is priced against the Upbit USDT-KRW rate, not the official USD/KRW exchange rate, and the two rates can differ. From 2026-10-31 the units field in paid /report and /report/premium responses (web, MCP and OKX) also said "converted at USD/KRW", which was wrong. The premium values were calculated the same way throughout.

Fix: From 2026-11-10 10:43 KST the units field says "converted at the Upbit USDT-KRW rate", and /report and /report/premium carry a regime_basis section with the BTC regime thresholds and the exchange-rate basis; the basis is listed in our public claims. A BTC regime variant that also uses the official exchange rate (regime_fx_v1) is planned; rule v1 stays as it is.

2026-07-18 ~ 2026-10-07 09:33 KST

Path variants of /chain/{power,defense,semis} (e.g. /chain/x/semis) bypassed the x402 payment check and returned the full product. /chain, /report, /report/premium and the OKX channel were not affected. No external use found in retained logs (2026-09-23 onward). Logs before 2026-09-23 are not available.

Fix: Fixed at nginx and API; daily selfcheck now tests path variants.

Event results and corrections

Economic event results are stored only after each number is checked against the official release of the issuing agency (BLS, BEA, Census, DOL, Federal Reserve, BOJ); some figures, such as payrolls, hourly earnings, job openings and retail-sales lines, are recalculated from the official tables. A number that does not match is left out, and if the official page cannot be read, nothing is stored and the check runs again later.

Corrections to stored event results are appended to a separate file and shown on /transparency; the original stored rows are not edited.

Event dateEventShown nowCorrectedSource
2026-10-28BOJ Monetary Policy MeetingUnverified — the stored result was withdrawn on 2026-11-07 (no official figure confirmed).2026-11-07official page

Data sources

CoinGecko — kimchi premium and funding (live); the 30-minute archive since 2026-10-25, US Department of the Treasury — Treasury yield curve, Federal Reserve — H.10 dollar indexes and Asian FX, Federal Register — US crypto regulation filings, USAspending.gov — US federal procurement awards, US Bureau of Labor Statistics and US Census Bureau — release pages for event results, Japan Ministry of Finance — JGB curve; Trade Statistics of Japan via e-Stat, Bank of Japan — monetary policy meeting schedule, Korea public data portal (data.go.kr) — KRX prices and Korea customs export statistics, Alpaca — US equity prices, used only to compute coefficients, never redistributed, Ministry of Finance Japan (JGB yields), CoinGecko 30-minute kimchi premium and funding archive (since 2026-10-25), Bank of Japan (meeting schedule), US Bureau of Labor Statistics and US Census Bureau (release pages for event results), Alpaca (US equity prices, used only to compute coefficients, never redistributed). Data provided by CoinGecko.


Asia Pulse publishes data and measurements. Nothing here is investment advice. Last updated 2026-11-17.