{"articles": {"2026-08-27-august-2024-yen-carry-fx-replay-remains-inconclusive-under-undertow-s-coverage-rule": {"body_md": "## The read\n\nUndertow’s historical replay of the August 2024 yen-carry episode produced an **INCONCLUSIVE** result. That is not a finding that FX conditions were calm during the episode, and it is not evidence that the board detected stress. It is a finding about the instrument’s coverage: the registered FX cell did not produce a scored tier inside the specified window.\n\nThe registered expectation was that the FX cell would reach STRAINED or DEFICIENT on at least one as-of date from 2024-07-25 through 2024-08-16. The replay had 1 as-of date, 2024-07-31. On that date the cell was PARTIAL, with 0 qualifying measures out of 2. There were no scored as-of dates.\n\nThe two FX measures, Amihud and lambda-hat on an FX ETF-proxy bucket, were present but withheld. Undertow’s coverage policy requires at least 2 qualifying measures for a scored tier, and a withheld cell is an absence of verdict rather than a pass. The result therefore does not support a claim that the cell was NORMAL, STRAINED, or DEFICIENT during the registered period.\n\nThe live board is separate context, not evidence for this replay. On 2026-08-27, the dispatch reported that 1 of 9 segments scored, identified IG as NORMAL, and showed the Seiche funding overlay as STRAIN. The overlay is declared funding context only. It is not used in Undertow’s tier derivation and is not blended into the historical FX result.\n\n## The mechanism\n\nThe episode registration describes USD/JPY as falling roughly 12% from its July 2024 peak into 5 August and identifies the unwind of yen-funded carry positions as the proximate cause of that day’s global equity and volatility shock. The replay does not retest that historical account. Its narrower question is whether Undertow’s FX cell could meet its own stressed-tier condition during the chosen window.\n\nThe answer was constrained by the history requirement. Each FX measure needs 60 monthly observations to qualify. The relevant monthly history begins in 2020-08, which places the earliest scoreable FX as-of in mid-2025. A 2024 event window could therefore leave the measures available but unable to qualify for a tier.\n\nThat limitation was recorded in the episode rationale. It anticipated a NOT_RUNNABLE or INCONCLUSIVE result because the FX history was not yet sufficient for a scored reading. The published result was INCONCLUSIVE: the sole as-of date showed both measures withheld and the FX cell PARTIAL.\n\nThe replay is also a truncated-latest construction. It recomputes readings from today’s stored pack using observations dated on or before each historical as-of. This makes the exercise point-in-time in dates, but not a reconstruction of the exact data vintage available during the event. The record says that FRED and New York Fed series can be revised and that the product has no ALFRED-style vintage archive. The result must therefore be read as a construction replay on final data, not as a live simulation.\n\nThe source used for the replay, identified as equity_liquidity, also has a declared cross-sectional look-ahead caveat. Its monthly history is backfilled for a basket committed on 2026-07-17. Truncating the dates does not remove the survivorship-conditioned membership. That caveat does not turn the result into evidence of either stress or calm; it narrows what the replay can establish.\n\n## The countercase\n\nThe strongest countercase is that two FX ETF-proxy liquidity measures may be too limited for a yen-carry unwind. Such an event may transmit through funding, volatility, collateral, equities, positioning, or execution channels not captured by this FX cell. It may also have developed between the replay’s available observation dates.\n\nThat possibility does not contradict the INCONCLUSIVE verdict. It reinforces its proper interpretation: the board lacked the qualifying evidence required to assign its own FX tier in this window. The replay does not determine the severity of the historical episode or the condition of every FX venue.\n\nThe opposite inference is also unavailable. The fact that the measures were withheld does not imply that they would have shown a stressed tier if they had qualified. No tier was produced. Treating withheld inputs as an implied result would override the stated coverage rule.\n\nThe current Seiche STRAIN regime cannot resolve this gap. Its funding measure is a declared regime based on a calibrated weighted blend of engine sub-scores; its stress percentile is withheld and it is not a percentile of its own history. It is funding context only, not a cross-product score and not an input to the Undertow FX tier.\n\n## What would falsify this\n\nThe article’s claim is narrow: the published replay record shows that Undertow’s registered FX instrument could not issue a scored verdict within the stated event window. That claim would be falsified if the public record did not reproduce the PARTIAL FX tier on 2024-07-31, the 0 qualifying measures out of 2, or the absence of any scored as-of date.\n\nIt would also be falsified if the record showed a qualifying FX tier of STRAINED or DEFICIENT at an as-of date within the registered window. The replay record instead reports no scored as-of dates and an INCONCLUSIVE verdict.\n\nAdditional historical evidence could examine channels outside this construction, but it would be additional evidence. It would not change what this particular registered FX cell established under its stated coverage rule and replay vintage.\n\n## Method and limits\n\nThis is a historical replay, not a forward call and not investment advice. Its conclusion applies only to the registered expectation, the FX ETF-proxy measures, the 2024-07-25 to 2024-08-16 window, and the stated coverage and vintage rules.\n\nPARTIAL means a segment is still accruing history. Undertow states that a scored tier needs at least 2 measures with a stress percentile at 60 or more observations each; accruing or absent measures are disclosed rather than rendered as calm. A tier is a deterministic board derivation, not an observed price or a trade signal.\n\nThe 2026-08-27 board and its funding overlay are included only to identify the current published context. They do not validate, revise, or supplement the historical FX verdict. Research and market data, not investment advice.\n\n## Sources\n\n- [Undertow](https://liquilens-undertow.com)\n- [Undertow historical and board records](https://api.seiche.info/undertow)\n- [Undertow’s 2026-08-27 dispatch record](https://liquilens-undertow.com/dispatch/2026-08-27.json)\n- [Seiche](https://seiche.info)\n- [LiquiLens](https://liquilens.in)\n- [Undertow on Telegram](https://t.me/Undertow_LiquiLens_bot)", "canonical_url": "https://liquilens-undertow.com/articles/2026-08-27-august-2024-yen-carry-fx-replay-remains-inconclusive-under-undertow-s-coverage-rule/", "date": "2026-08-27", "dek": "Undertow’s registered FX test did not produce a scored tier during its August 2024 event window. With both measures withheld on its only as-of date, the replay cannot be read as either detection of stress or evidence of calm.", "editorial_memory": {"directives": ["preserve_current_standard"], "memory_fingerprint": "sha256:858e66336656deb8fc0ab6d4cbc625e3de57656b9566f701bef7c1c4601c987d", "rubric_version": "mqdnse.editorial-rubric.v1", "source_run_id": "sha256:7639b188886e625cdd6e00cc3bc9ab28485fb21bb5b2562487d182348b7297b8", "status": "applied"}, "evidence_asof": "2026-08-27", "evidence_receipt": {"board": "https://api.seiche.info/undertow/board.json", "board_scope": "public free subset only", "cross_product_score": null, "dispatch": "https://api.seiche.info/undertow/dispatch.json", "replay": "https://api.seiche.info/undertow/replay_coverage.json"}, "funnel": {"companions": ["https://seiche.info/articles/", "https://liquilens.in/articles/"], "primary_label": "Inspect the live Undertow board", "primary_url": "https://t.me/Undertow_LiquiLens_bot"}, "generation": {"deterministic_boundary_overlay": [], "dossier_sha256": "f1ebb2f608d5ae40f1e1e8de960efd988ddb1e13eb8351b9b457381ea2851a70", "editorial_memory": {"directives": ["preserve_current_standard"], "memory_fingerprint": "sha256:858e66336656deb8fc0ab6d4cbc625e3de57656b9566f701bef7c1c4601c987d", "rubric_version": "mqdnse.editorial-rubric.v1", "source_run_id": "sha256:7639b188886e625cdd6e00cc3bc9ab28485fb21bb5b2562487d182348b7297b8", "status": "applied"}, "fallback_reason": null, "mode": "model_written_reviewed", "passes": 2, "review_model": "openai/gpt-5.6-terra", "review_reasons": ["The draft is broadly dossier-grounded, but it contains padded editorial assertions that do not advance the registered replay finding, including claims about what preregistration is “for” and characterizations such as “less theatrical” and “usable.”", "The headline’s “usable result” is an unsupported evaluative gloss. The dossier supports an INCONCLUSIVE result and its stated reason, not a broader judgment of usefulness.", "The falsifier section is overextended: a later or altered registration record would bear on the integrity of the exercise, but the article’s narrow claim should be falsified first by failure to reproduce the published replay record’s PARTIAL tier, zero qualifying measures, and zero scored as-of dates.", "A repaired version should retain the countercase, state the replay’s final-data and survivorship-conditioned limitations, and keep the current funding reading strictly separate from the historical FX verdict."], "writer_model": "openai/gpt-5.6-terra"}, "headline": "August 2024 yen-carry FX replay remains inconclusive under Undertow’s coverage rule", "published_at": "2026-08-27T21:54:11Z", "quality": {"failures": [], "numeric_tokens_checked": 18, "required_headings": 6, "source_links": 6, "status": "PASS", "word_count": 1036}, "slug": "2026-08-27-august-2024-yen-carry-fx-replay-remains-inconclusive-under-undertow-s-coverage-rule", "topic": "aug2024_yen_carry_fx", "type": "historical_replay", "word_count": 1036}, "2026-08-28-the-june-2013-cn-replay-is-not-a-stress-verdict-it-is-a-coverage-verdict": {"body_md": "## The read\n\nUndertow’s historical replay does not say that the June 2013 Shibor shock passed or failed a China liquidity test. It says that the test could not run.\n\nThe registered episode, **June 2013 Shibor shock — China basin**, covers 2013-06-01 through 2013-07-15. Its stated expectation is that the CN cell reaches DEFICIENT on at least one as-of within that window. The published result is **NOT_RUNNABLE**.\n\nThat is the narrow conclusion. A missing historical measurement surface is a limitation of the instrument, not evidence about the event’s severity, duration, or consequences. The result cannot be rendered as calm, and it cannot be converted into a retrospective market judgment.\n\nThe replay record gives two reasons. No wired replay source feeds the CN segment, and the packs carrying CN have no history reaching the registered period. The result has no evidence, no scored as-ofs and no sources used. Undertow therefore cannot establish whether its registered CN cell-tier expectation would have been met during the episode.\n\nThe current daily reading provides context but does not repair that historical gap. On 2026-08-28, 1 of 9 segments scores; UST, HY, EQUITY, ETF, FX, CN, CRYPTO and BSTOCK are still accruing history. IG reads NORMAL on its within-board score. The separate Seiche funding overlay is declared STRAIN. These are current readings from separate products, not a blended cross-product score and not confirmation of the historical CN replay.\n\n## The mechanism\n\nA replay can test only the inputs its recorded source coverage can observe. Undertow’s registered expectation concerns a CN cell tier. It is not a general claim about every China-basin funding condition or every manifestation of market dislocation.\n\nFor the CN cell to generate replay evidence, an eligible wired source would need to feed CN and have observations reaching an as-of within the registered window. The published record says those conditions are not met.\n\nA tier is a deterministic board derivation, not an observed price or a trade signal. The daily dispatch states that a scored tier needs at least 2 measures with a stress percentile at 60+ observations each; accruing or absent measures are disclosed rather than rendered as calm. The CN replay fails at an earlier stage: there is no eligible wired historical source coverage from which to produce replay evidence for the episode.\n\nThe replay also has a stated vintage limitation. Readings are recomputed from today’s stored pack, truncated to observations dated on or before each as-of. That is described as what was knowable-by-date, rather than the vintage published at the time. The record notes that FRED and NY Fed series can be revised and that the product has no ALFRED-style vintage archive. For CN, though, the immediate limitation is absence of eligible historical coverage, not merely revision risk.\n\nThe episode’s registration rationale makes the intended discipline explicit: the CN row is given a NOT_RUNNABLE result rather than left empty in a way that could be mistaken for untested-but-testable. The record identifies what the system cannot evaluate instead of allowing absence to resemble neutrality.\n\n## The countercase\n\nThe strongest countercase is that NOT_RUNNABLE may be too narrow to satisfy readers considering a well-known historical episode. An alternative proxy, an adjacent market series or a later-assembled dataset might support a different analysis. The relevant transmission channel might also extend beyond the particular CN cell that Undertow registered.\n\nThose possibilities may justify building a new test. They do not justify changing this registered test after its coverage limitation is known. Replacing sources, changing the window or importing a different product’s regime would create a different evaluation rather than complete this one.\n\nNor does the Seiche STRAIN overlay fill the gap. The evidence braid identifies Seiche as a funding-context overlay only and says it is not used in the Undertow tier. Its measure is a calibrated weighted blend of engine sub-scores, with higher values denoting stress; it is not a percentile of its own history, and its stress percentile is withheld rather than invented. A declared funding regime can provide context, but it cannot create a missing CN replay source or establish a historical CN cell outcome.\n\n## What would falsify this\n\nThe claim here is limited and falsifiable: the published NOT_RUNNABLE result accurately describes what the registered Undertow CN instrument can establish for the 2013-06-01 through 2013-07-15 window.\n\nThat claim would be undermined by a reproducible record showing that a wired source fed CN, that its stored history reached an as-of in the registered window, and that the stated method produced scored replay evidence capable of testing the registered expectation. It would also be undermined if the episode registration or expectation had been altered after the result.\n\nA future eligible historical source could support a new CN replay. It would not turn the present NOT_RUNNABLE result into either a pass or a fail. Any new result would need its own disclosed source coverage and the same stated vintage boundary.\n\n## Method and limits\n\nThis is a historical-replay editorial based solely on the supplied evidence envelope. It does not infer withheld percentiles, create a cross-product score or make a forward call.\n\nThe replay is NOT_RUNNABLE because the record reports no wired replay source feeding CN and no historical coverage reaching the registered period. The current Undertow board and the Seiche funding regime are reported as separate product readings, never as one combined score. The board’s funding overlay is declared STRAIN, while IG’s within-board reading is NORMAL.\n\nResearch and market data, not investment advice.\n\n## Sources\n\n- [Undertow](https://liquilens-undertow.com)\n- [Undertow API](https://api.seiche.info/undertow)\n- [Seiche](https://seiche.info)\n- [LiquiLens](https://liquilens.in)\n- [Undertow LiquiLens bot](https://t.me/Undertow_LiquiLens_bot)\n- [Undertow API record](https://api.seiche.info/undertow)", "canonical_url": "https://liquilens-undertow.com/articles/2026-08-28-the-june-2013-cn-replay-is-not-a-stress-verdict-it-is-a-coverage-verdict/", "date": "2026-08-28", "dek": "Undertow’s registered China-basin replay for the June 2013 Shibor shock is NOT_RUNNABLE. That result does not minimize the episode; it fixes the boundary of what this instrument can honestly claim to have tested.", "editorial_memory": {"directives": ["preserve_current_standard"], "memory_fingerprint": "sha256:f17535929c098429accb219c382a7dc958c758fa2806cc96aab152a2acce96f0", "rubric_version": "mqdnse.editorial-rubric.v1", "source_run_id": "sha256:af16a9c24ea2cdf5fe0da3b2f012c691bd8d28d4abf771b13f122628042dd4a5", "status": "applied"}, "evidence_asof": "2026-08-28", "evidence_receipt": {"board": "https://api.seiche.info/undertow/board.json", "board_scope": "public free subset only", "cross_product_score": null, "dispatch": "https://api.seiche.info/undertow/dispatch.json", "replay": "https://api.seiche.info/undertow/replay_coverage.json"}, "funnel": {"companions": ["https://seiche.info/articles/", "https://liquilens.in/articles/"], "primary_label": "Inspect the live Undertow board", "primary_url": "https://t.me/Undertow_LiquiLens_bot"}, "generation": {"deterministic_boundary_overlay": [], "dossier_sha256": "188a397bd0f690be77e42609c5882319cc2e04e8c88ab3ff7c1a427416066def", "editorial_memory": {"directives": ["preserve_current_standard"], "memory_fingerprint": "sha256:f17535929c098429accb219c382a7dc958c758fa2806cc96aab152a2acce96f0", "rubric_version": "mqdnse.editorial-rubric.v1", "source_run_id": "sha256:af16a9c24ea2cdf5fe0da3b2f012c691bd8d28d4abf771b13f122628042dd4a5", "status": "applied"}, "fallback_reason": null, "mode": "model_written_reviewed", "passes": 2, "review_model": "openai/gpt-5.6-terra", "review_reasons": ["The draft is broadly dossier-grounded and includes all required headings and six approved-link instances, but it contains unsupported characterizations that must be removed.", "“Different desk’s regime” and “comparisons across desks” are not established in the dossier. The dossier distinguishes products and specifies that Seiche is a funding-context overlay only; it does not establish desk identities.", "“Reconstruct subscriber measures” is unsupported. The dossier says the free board withholds a percentile, not that the underlying measures are subscriber-only.", "Calling the Telegram link a “public access point for the live product” adds an unsupported description of that source.", "The falsifier should be tied more precisely to the registered expectation: a qualifying source and scored as-of within the registered window would be material, rather than requiring history across the entire window."], "writer_model": "openai/gpt-5.6-terra"}, "headline": "The June 2013 CN replay is not a stress verdict. It is a coverage verdict.", "published_at": "2026-08-28T21:47:46Z", "quality": {"failures": [], "numeric_tokens_checked": 12, "required_headings": 6, "source_links": 6, "status": "PASS", "word_count": 937}, "slug": "2026-08-28-the-june-2013-cn-replay-is-not-a-stress-verdict-it-is-a-coverage-verdict", "topic": "shibor_shock_jun2013_cn", "type": "historical_replay", "word_count": 937}, "2026-08-29-funding-moves-strain-to-erosion-ig-remains-normal": {"body_md": "## The read\n\nThe public [Undertow dispatch](https://liquilens-undertow.com/dispatch/2026-08-29.json) dated 2026-08-29 presents a narrow divergence, not a market-wide verdict. It scores 1 of 9 segments. IG is the scored segment, classified NORMAL with a within-board score of 0.4724, 3 available measures, and 3 qualifying measures. UST, HY, EQUITY, ETF, FX, CN, CRYPTO, and BSTOCK are still accruing history.\n\nSeparately, the Seiche funding overlay changed from STRAIN to EROSION. That change is fresh relative to the prior dated reading, but it does not alter the IG tier. The [Undertow](https://liquilens-undertow.com) evidence braid defines Seiche as a funding-context overlay only and says it is not used in the tier.\n\nThe analytical point is limited. IG’s NORMAL label remains the published board result. But the supplied IG metadata also says `disagrees: true`. Undertow says qualifying measures disagree within at least 1 scored cell and that the span for each such cell is restricted to the Agent/Desk tier because disclosing it would reveal a withheld percentile. The public record therefore supports scrutiny of internal disagreement, not a claim that IG has entered a stressed tier.\n\n## The mechanism\n\nUndertow describes a tier as a deterministic board derivation, not an observed price or a trade signal. Its stated method compares measures with their own accrued point-in-time histories. That matters because the measures are not interchangeable, and neither is the Seiche overlay with the IG tier.\n\nFor IG, the disclosed `IG OAS (BAMLC0A0CM)` reading has a stress percentile of 0.2055 as of 2026-08-27. The measure note says a wider investment-grade spread means costlier credit liquidity and more stress. A second named measure, dealer corporate net positions, is described as a congestion proxy: high net dealer inventory indicates less balance-sheet capacity to absorb customer selling. Its stress percentile is withheld, and its stated observation date is 2026-08-19.\n\nThe funding overlay requires separate treatment. Its note reports a Seiche composite index of 44.9/100, regime EROSION, and 100% coverage. It also says the composite is a calibrated weighted blend of engine sub-scores, with higher values indicating stress, and is not a percentile of its own history. The board explicitly withholds a funding stress percentile rather than inventing one.\n\nNo combined score follows from these facts. IG is NORMAL under Undertow’s qualifying-measure process. Funding is EROSION under Seiche’s declared-regime process. The evidence permits comparison of those distinct readings; it does not permit aggregation into a cross-product market signal.\n\n## The countercase\n\nThe strongest countercase is that there is no material tension. A NORMAL IG tier based on 3 qualifying measures may be the appropriate result under the board’s rules, while disagreement within that cell may be routine variation rather than evidence of deterioration.\n\nThe disclosed measure limits reinforce that restraint. The OAS note says the available FRED history is limited to a trailing roughly 3-year window and omits earlier crisis regimes, including 2008 and 2020. It warns that this can understate rarity in calm periods. The dealer-position measure is a proxy for IG, not an IG-only series: it spans the broader corporate-credit complex and is described as spliced across New York Fed series breaks.\n\nNor does the funding change mechanically require an IG-tier change. Seiche’s declared funding regime and Undertow’s deterministic tier are separate product outputs. Treating their difference as a contradiction would add a causal relationship that the public pack does not establish.\n\n## What would falsify this\n\nThe IG portion of this analysis would fail if the 2026-08-29 public record were corrected to show that IG was not NORMAL, did not have a 0.4724 score, did not have 3 qualifying measures, or did not carry the disclosed disagreement flag. It would also fail if a correction established that the disagreement statement did not apply to the scored IG cell.\n\nThe funding portion would fail if the public record were corrected to show that the 2026-08-29 Seiche overlay was not EROSION or that the prior regime was not STRAIN. These are record-based falsifiers, not forecasts about spreads, returns, issuance, or liquidity outcomes.\n\nA later board reading may provide new evidence, but it would not retroactively change what this dated dispatch reported.\n\n## Method and limits\n\nThis article is confined to the supplied public evidence envelope for 2026-08-29. It does not reconstruct withheld percentiles, identify undisclosed qualifying measures, calculate an internal IG span, or create a blended Undertow-Seiche score.\n\nUndertow states that a scored tier requires at least 2 measures with stress percentiles and at least 60 observations each. The public IG metadata reports 3 qualifying measures, but the supplied record provides detailed context for only selected named measures. That boundary prevents a claim about the exact source or size of the internal disagreement.\n\nThe board record is dated 2026-08-29 and reports publication at 2026-08-29T21:53:34+00:00, with source-publication status OBSERVED. No sealed call is attached to this daily reading. Research and market data, not investment advice.\n\n## Sources\n\n- [Undertow dispatch, 2026-08-29](https://liquilens-undertow.com/dispatch/2026-08-29.json)\n- [Undertow](https://liquilens-undertow.com)\n- [Undertow API](https://api.seiche.info/undertow)\n- [Seiche](https://seiche.info)\n- [LiquiLens](https://liquilens.in)\n- [Undertow Telegram bot](https://t.me/Undertow_LiquiLens_bot)", "canonical_url": "https://liquilens-undertow.com/articles/2026-08-29-funding-moves-strain-to-erosion-ig-remains-normal/", "date": "2026-08-29", "dek": "Undertow’s 2026-08-29 board scores 1 of 9 segments: IG is NORMAL with a 0.4724 within-board score and 3 qualifying measures. Seiche’s separate funding overlay changed from STRAIN to EROSION; the readings cannot be blended.", "editorial_memory": {"directives": ["preserve_current_standard"], "memory_fingerprint": "sha256:5d650b303ea0acf976120344065f7bdde7301306e8ac02b6547fcd72e1a2d610", "rubric_version": "mqdnse.editorial-rubric.v1", "source_run_id": "sha256:240a106ec01413cf7a3319673d388d619e94c41aad2e6c2029f550adb0417442", "status": "applied"}, "evidence_asof": "2026-08-29", "evidence_receipt": {"board": "https://api.seiche.info/undertow/board.json", "board_scope": "public free subset only", "cross_product_score": null, "dispatch": "https://api.seiche.info/undertow/dispatch.json", "replay": "https://api.seiche.info/undertow/replay_coverage.json"}, "funnel": {"companions": ["https://seiche.info/articles/", "https://liquilens.in/articles/"], "primary_label": "Inspect the live Undertow board", "primary_url": "https://t.me/Undertow_LiquiLens_bot"}, "generation": {"deterministic_boundary_overlay": [], "dossier_sha256": "ce3746212d837aacbe05d75222fe14c32b8e36c148acb90b73fed078c78adf7c", "editorial_memory": {"directives": ["preserve_current_standard"], "memory_fingerprint": "sha256:5d650b303ea0acf976120344065f7bdde7301306e8ac02b6547fcd72e1a2d610", "rubric_version": "mqdnse.editorial-rubric.v1", "source_run_id": "sha256:240a106ec01413cf7a3319673d388d619e94c41aad2e6c2029f550adb0417442", "status": "applied"}, "fallback_reason": null, "mode": "model_written_reviewed", "passes": 2, "review_model": "openai/gpt-5.6-terra", "review_reasons": ["The draft adds an unsupported claim about a possible future “replay verdict” and what it would establish. The dossier says only that replay was generated; it does not supply a replay verdict or its interpretive status.", "“Contemporaneous” overstates the timing relationship. The board is dated 2026-08-29, but the disclosed IG OAS observation is dated 2026-08-27 and the disclosed dealer-position observation is dated 2026-08-19.", "The Sources section contains six link instances but duplicates the Undertow API URL rather than providing six distinct approved-source links. The repaired version uses the dossier’s canonical dispatch URL plus the approved source set.", "The repair removes unsupported or surplus interpretive language while preserving the required countercase, falsifiers, research boundary, all required headings, and dossier-grounded numerals."], "writer_model": "openai/gpt-5.6-terra"}, "headline": "Funding moves STRAIN to EROSION; IG remains NORMAL", "published_at": "2026-08-29T21:54:15Z", "quality": {"failures": [], "numeric_tokens_checked": 20, "required_headings": 6, "source_links": 8, "status": "PASS", "word_count": 852}, "slug": "2026-08-29-funding-moves-strain-to-erosion-ig-remains-normal", "topic": "IG", "type": "current_analysis", "word_count": 852}, "2026-08-30-undertow-s-jane-street-replay-is-not-a-failed-test-it-is-a-coverage-boundary": {"body_md": "## The read\n\nA historical replay is only as useful as its instrument is specific. On that standard, Undertow cannot claim a result for the registered episode, **July 2025 SEBI order against Jane Street — single-firm withdrawal**. The replay covers the window from 2025-07-03 through 2025-08-15, and its registered expectation is explicit: an INDIA_OPTIONS cell would reach STRAINED or DEFICIENT on at least one as-of inside that window. The verdict is **NOT_RUNNABLE**.\n\nThat is the thesis: NOT_RUNNABLE is a coverage finding, not a market verdict. It identifies the point where a liquidity board must stop rather than substitute a nearby market, a different product, or an attractive retrospective story. Undertow’s board vocabulary contains UST, IG, HY, EQUITY, ETF, FX, CN, CRYPTO and BSTOCK. INDIA_OPTIONS is not among those rows. The replay ledger further says the product observes no market of that name. With no named row and no applicable construction, there are no scored as-ofs, no evidence trace and no honest basis for declaring whether the registered threshold was reached.\n\nThe distinction matters because the headline event is not being replayed through an instrument that merely resembles its subject. The ledger does not say that conditions were calm in India options, that the episode did not affect liquidity, or that a single-firm withdrawal could not have mattered. It says that this particular board cannot establish any of those propositions. A missing measurement is not calm; it is missing measurement.\n\nThe current dispatch reinforces why the boundary should remain visible. On 2026-08-30, one of nine board segments scored: IG read NORMAL with three qualifying measures and a within-board score of 0.4724. UST, HY, EQUITY, ETF, FX, CN, CRYPTO and BSTOCK were still accruing history. Separately, Seiche’s declared funding overlay read EROSION. Those are distinct readings from distinct products. They do not repair the INDIA_OPTIONS gap, and they must not be blended into a cross-product score or used as a proxy replay for a market Undertow does not cover.\n\n## The mechanism\n\nThe registered rationale identifies the economic question: when one firm is barred, does the liquidity it provided leave with it? That is a question about the capacity to transfer risk, not merely about whether a market print moved. A valid replay would require a defined market cell, eligible measures and enough historical observations to apply the board’s rules at each as-of. Undertow has none of that for INDIA_OPTIONS.\n\nIts existing construction is deliberately narrower. A scored tier requires at least two measures with a stress percentile at 60 or more observations each. Measures that are accruing or absent are disclosed rather than rendered as calm. That rule is useful only where the board has a row and an associated measurement design. It cannot turn an unobserved market into a scored segment by analogy.\n\nThe replay record names the practical reason. No Indian tape is collected in this repository. It also distinguishes the FBIL/CCIL/RBI daily snapshot associated with LiquiLens as a different product with a different data plane. That cross-product distinction is not cosmetic. A dataset from another product may address a related institutional or market question, but it is not automatically an input to Undertow’s historical construction. Importing it after the fact would change the instrument being evaluated.\n\nPoint-in-time discipline creates a second boundary. The replay is calculated from the stored pack truncated to observations dated on or before each as-of. The ledger calls this a construction replay on final data, not a live simulation, because it does not hold an ALFRED-style vintage archive and notes that FRED and the NY Fed revise their series. In this case, that limitation does not produce a weak pass or a weak fail. It means there is no eligible INDIA_OPTIONS construction to run at all.\n\nThe appropriate editorial conclusion is therefore restrained: the pre-registered question remains unanswered by Undertow. That is more informative than manufacturing a result from a neighbouring segment, a current funding regime, or an institution-liquidity dataset that belongs to another product.\n\n## The countercase\n\nThe strongest countercase is that the registered test was too narrow for its own episode. A liquidity effect related to the July 2025 SEBI order could have appeared in a channel other than the named INDIA_OPTIONS cell, in measures not collected by Undertow, or at a frequency and depth the board does not observe. The event could still have had meaningful consequences while this replay remained impossible.\n\nThat countercase does not weaken the NOT_RUNNABLE verdict. It explains why the verdict should not be misread as an all-clear. A tool can be insufficient for a question without the question being unimportant. Indeed, the ledger’s refusal to substitute a neighbouring segment is the key methodological safeguard. Replacing INDIA_OPTIONS with one of the existing board rows would make a cleaner article but a less defensible replay.\n\nThere is an opposite risk as well: treating a future data addition as if it validates the existing historical claim. It would not. A new source could enable a new registered construction and a new result, but it would not alter what this replay could establish. The original instrument had no INDIA_OPTIONS row, no scored as-ofs and no evidence record. That remains its historical boundary.\n\nCurrent conditions offer no shortcut. The IG reading and the declared Seiche EROSION overlay may be useful present-tense context, but they are comparisons only. The dispatch explicitly marks Seiche as a funding-context overlay and says it was not used in tiering. Neither current reading is evidence about the historical episode, and neither is a blended verdict.\n\n## What would falsify this\n\nThis editorial makes a narrow, falsifiable claim: the published NOT_RUNNABLE verdict accurately describes Undertow’s registered INDIA_OPTIONS replay for the stated window. That claim would fail if the public replay record showed that INDIA_OPTIONS was in fact a board row, if it contained eligible measures and scored as-ofs for the window, or if a reproducible trace established that the registered construction had run and produced a tiered result.\n\nIt would also fail if the article’s stated limitation were contradicted by the record: namely, if Undertow’s data plane did collect the named market and the ledger’s missing-input explanation were wrong. Nothing in the supplied record supports those alternatives. Instead, it explicitly identifies INDIA_OPTIONS as outside the board vocabulary and records the reason for the absence.\n\nA later expansion of coverage would not falsify the historical description of this replay. It could challenge the product’s future coverage limitation, and it could support a separate test with a separately recorded result. But it cannot retroactively transform NOT_RUNNABLE into PASS, FAIL or a claim about event severity without changing the construction.\n\n## Method and limits\n\nThis is a historical replay editorial, not a forward call. Undertow’s sealed calls are maintained as a separate verifiable record, and the dispatch says a daily board reading is not silently promoted into a forward call. The article therefore does not infer a trade, a forecast or an investment action from the replay or from the current board.\n\nThe method boundary is unusually important here. The replay evaluates a named instrument against a named episode and window. Its verdict describes the instrument’s ability to run the registered test; it does not decide whether history happened, grade every possible market channel, or adjudicate the underlying event. The replay ledger reports no sources used, no evidence and no scored as-ofs because the required segment does not exist in the board.\n\nThe current dispatch is likewise limited. It reports IG as NORMAL and reports a declared EROSION funding overlay, while eight other board segments are still accruing history. The funding composite is not an own-history percentile: the record says its stress percentile is withheld rather than invented. Qualifying measures disagree in at least one scored cell. For each such cell, the free board shows the span as an Agent/Desk tier because disclosing the span would otherwise hand back the percentile of a row it withholds. Those disclosures are reasons for restraint, not invitations to fill gaps with estimates.\n\nResearch and market data only; not investment advice.\n\n## Sources\n\n- [Undertow](https://liquilens-undertow.com) — public product and board context.\n- [Undertow API](https://api.seiche.info/undertow) — approved Undertow record access point.\n- [Seiche](https://seiche.info) — funding context product, used only as a contextual comparison.\n- [LiquiLens](https://liquilens.in) — separate product referenced in the replay limitation.\n- [Undertow on Telegram](https://t.me/Undertow_LiquiLens_bot) — approved Undertow channel.\n- [Undertow public record](https://liquilens-undertow.com) — reference for the published board surface.", "canonical_url": "https://liquilens-undertow.com/articles/2026-08-30-undertow-s-jane-street-replay-is-not-a-failed-test-it-is-a-coverage-boundary/", "date": "2026-08-30", "dek": "The registered INDIA_OPTIONS replay for the July 2025 SEBI order against Jane Street — single-firm withdrawal is NOT_RUNNABLE. That verdict says nothing about the severity of the event; it says Undertow has no valid instrument for measuring that named market.", "editorial_memory": {"directives": ["preserve_current_standard"], "memory_fingerprint": "sha256:ef1ae2bbebda4b997e65a0ca79fdacf4520037039f8a469de049d8cc095c649f", "rubric_version": "mqdnse.editorial-rubric.v1", "source_run_id": "sha256:1ecb7a428475cba104da82fc2e0f671c304c2151eafb5e751459f9a7c793c13f", "status": "applied"}, "evidence_asof": "2026-08-30", "evidence_receipt": {"board": "https://api.seiche.info/undertow/board.json", "board_scope": "public free subset only", "cross_product_score": null, "dispatch": "https://api.seiche.info/undertow/dispatch.json", "replay": "https://api.seiche.info/undertow/replay_coverage.json"}, "funnel": {"companions": ["https://seiche.info/articles/", "https://liquilens.in/articles/"], "primary_label": "Inspect the live Undertow board", "primary_url": "https://t.me/Undertow_LiquiLens_bot"}, "generation": {"deterministic_boundary_overlay": [], "dossier_sha256": "bf3faaa7cab271e54e56d35c3e81a912dad0f79ed9be5c0880ead6285463e82b", "editorial_memory": {"directives": ["preserve_current_standard"], "memory_fingerprint": "sha256:ef1ae2bbebda4b997e65a0ca79fdacf4520037039f8a469de049d8cc095c649f", "rubric_version": "mqdnse.editorial-rubric.v1", "source_run_id": "sha256:1ecb7a428475cba104da82fc2e0f671c304c2151eafb5e751459f9a7c793c13f", "status": "applied"}, "fallback_reason": null, "mode": "model_written_reviewed", "passes": 2, "review_model": "openai/gpt-5.6-terra", "review_reasons": ["The draft incorrectly says that within-cell span information is withheld on the free board. The dispatch says that, where qualifying measures disagree, every such cell’s span is shown as the Agent/Desk tier; what is withheld is the percentile of a row the free board does not disclose.", "The sources list describes the Telegram link as “live product access,” a function not established in the dossier. Source labels must stay within the supplied record.", "Repairs are required because the article must accurately distinguish the disclosed Agent/Desk span tier from the withheld underlying percentile."], "writer_model": "openai/gpt-5.6-terra"}, "headline": "Undertow’s Jane Street replay is not a failed test. It is a coverage boundary.", "published_at": "2026-08-30T21:54:05Z", "quality": {"failures": [], "numeric_tokens_checked": 9, "required_headings": 6, "source_links": 6, "status": "PASS", "word_count": 1402}, "slug": "2026-08-30-undertow-s-jane-street-replay-is-not-a-failed-test-it-is-a-coverage-boundary", "topic": "sebi_janestreet_jul2025", "type": "historical_replay", "word_count": 1402}, "2026-08-31-funding-moves-erosion-to-strain-segment-tiers-hold": {"body_md": "## The read\n\nThe narrow conclusion from Undertow’s published reading is not that investment-grade credit has become illiquid. It is that the board’s sole scored segment is holding a NORMAL tier while the separate Seiche funding overlay has changed from EROSION to STRAIN. That is a meaningful disagreement in the system’s public readings, but it is not a blended verdict and should not be presented as one.\n\nIG is the only segment that scores on the current board. UST, HY, EQUITY, ETF, FX, CN, CRYPTO, and BSTOCK are still accruing history. The IG cell has a within-board score of 0.4728, three qualifying measures, and a NORMAL tier. The funding overlay is declared STRAIN. Seiche’s published note describes its composite index at 45.1 out of 100, with full coverage, while also stating that the index is a calibrated weighted blend of engine sub-scores rather than a percentile of its own history.\n\nThose descriptions answer different questions. Undertow’s IG result is a deterministic board derivation based on its own accrued point-in-time histories and coverage rules. Seiche supplies a funding-context overlay. The dispatch explicitly says the Seiche relationship is overlay-only and is not used in the Undertow tier. The editorial implication is therefore constrained: funding conditions have deteriorated by the funding desk’s declared regime while the only qualifying Undertow segment remains classified NORMAL.\n\nThat is not a contradiction to be averaged away. It is a reason to retain the distinction between a market-tier reading and a funding regime. The public evidence says that both are present at once. It does not authorize a claim that STRAIN has already caused IG liquidity to fail, nor that NORMAL IG conditions invalidate the funding signal.\n\nThe most important qualification sits inside the IG cell. The board says qualifying measures disagree. Where that happens, the public presentation uses an Agent/Desk tier span because disclosing a narrower span would reveal a withheld percentile. The disagreement is evidence, not a gap to be filled with guessed arithmetic. The accessible board establishes that the cell is NORMAL; it does not disclose every underlying qualifying percentile needed to independently reconstruct its internal range.\n\n## The mechanism\n\nThe mechanism visible in the dossier is methodological before it is macro-financial. Undertow compares each measure against its own accrued history, rather than treating unlike raw observations as directly interchangeable. A segment tier is then derived deterministically under the board’s coverage requirements. A measure that is accruing, absent, or withheld is not rendered as calm merely because it cannot be displayed in full.\n\nThat framework matters especially on a day with narrow coverage. Only one of nine segments scores. The other segments are not negative evidence for stress, nor positive evidence for calm; they are explicitly still accruing history. The board therefore permits a precise statement about IG and an equally precise statement about what cannot yet be said about the rest of the cross-market set.\n\nThe disclosed IG components reinforce that boundary. IG OAS, identified as BAMLC0A0CM, has a stress percentile of 0.2066 as of its shown observation date. Its stated limitation is material: the accessible FRED history is capped to a trailing window and omits earlier crisis regimes, including the episodes named in the board note. That limitation means the percentile is a ranking within the available history, not a universal measure of rarity across every prior market cycle.\n\nThe board also lists dealer corporate net positions as a congestion proxy, while withholding its stress percentile. Its published caveat is that the series covers the corporate-credit complex rather than IG alone and includes series-break splicing. That leaves no basis to infer the withheld value, assign a direction to the undisclosed disagreement, or claim a precise transmission from funding into dealer capacity. It does establish why the IG tier requires careful reading: the public label compresses measures with different coverage and measurement limits.\n\nSeiche’s STRAIN regime should accordingly be read beside the IG tier, not inside it. The funding desk’s index is expressly not an own-history percentile, and Undertow’s score is not represented as the Seiche index. Their comparison is descriptive: one desk reports a declared funding regime change; the other retains a NORMAL tier for the one segment meeting its scoring conditions.\n\n## The countercase\n\nThe strongest countercase is straightforward. There may be no actionable cross-market tension at all—only two systems reporting different constructs under different methodologies. The funding overlay can move to STRAIN without producing a contemporaneous change in the IG tier because the products are not designed to measure the same object. That possibility is not a defect in either reading; it is the stated boundary of their relationship.\n\nA second countercase is that the apparent complexity of IG is mostly a public-disclosure constraint. The board identifies disagreement among qualifying measures but withholds the span where publication would expose a free-board-hidden percentile. Readers can observe the NORMAL tier and the fact of disagreement, but they cannot responsibly convert that fact into a claim of concealed deterioration or concealed resilience.\n\nA third countercase is historical-window sensitivity. The disclosed OAS percentile carries its own warning that the available series omits earlier crisis regimes. A low percentile within that window should not be recast as proof that IG is insulated from every form of stress. Equally, the dealer-position proxy should not be relabeled as a pure IG measure when the board says it covers a broader corporate-credit complex.\n\n## What would falsify this\n\nThis editorial’s thesis is falsified if a later published board no longer shows the stated configuration: IG classified NORMAL, Seiche declared STRAIN, and the other named segments still accruing history. It is also falsified if the underlying published dispatch corrects the prior funding regime, IG tier, qualifying-measure count, or the board’s description of disagreement.\n\nThe thesis would weaken if subsequent disclosed evidence showed that the present comparison rested on an invalid clock alignment or on a revision that changes the point-in-time record. The dispatch provides separate event, knowledge, publication, and source-publication times; a later account should use those published clocks rather than silently substitute a different timestamp.\n\nWhat would not falsify the thesis is a later price move by itself. This is a claim about the published configuration of a funding overlay and a liquidity-board tier, not a forecast of returns, spreads, or a recommendation to act.\n\n## Method and limits\n\nThis analysis uses only the published Undertow dispatch dated 2026-08-31 and its stated public-board evidence. It treats the Seiche reading as contextual, as the dispatch requires, and does not calculate a cross-product score. It does not infer withheld percentiles, reconstruct the Agent/Desk span, or turn PARTIAL, absent, or accruing coverage into a calm signal.\n\nThe board’s own limitations govern the interpretation. A scored tier requires at least two measures with a stress percentile at the stated observation threshold; measures still accruing are disclosed rather than treated as complete. The board also states that a tier is a deterministic derivation, not an observed price or trade signal. The dispatch says its daily reading is not silently promoted into a forward call, and its sealed-call record is separate and verifiable.\n\nThis is research and market-data analysis, not investment advice.\n\n## Sources\n\n- [Undertow public site](https://liquilens-undertow.com)\n- [Undertow public API](https://api.seiche.info/undertow)\n- [Seiche funding desk](https://seiche.info)\n- [LiquiLens](https://liquilens.in)\n- [Undertow LiquiLens bot](https://t.me/Undertow_LiquiLens_bot)\n- [Undertow dispatch and ledger access](https://api.seiche.info/undertow)", "canonical_url": "https://liquilens-undertow.com/articles/2026-08-31-funding-moves-erosion-to-strain-segment-tiers-hold/", "date": "2026-08-31", "dek": "1 of 9 segments score today; UST, HY, EQUITY, ETF, FX, CN, CRYPTO, BSTOCK still accrue history. The funding overlay reads STRAIN. Qualifying measures disagree inside at least one scored cell; every such cell's span is Agent/Desk tier today, because the span would hand back the percentile of a row the free board withholds.", "editorial_memory": {"directives": ["preserve_current_standard"], "memory_fingerprint": "sha256:f5da1f6b466f2456a7128f159b24e01e0bfa58a0159b1633b2d9ca2d45eeabc7", "rubric_version": "mqdnse.editorial-rubric.v1", "source_run_id": "sha256:c69e080b54817510bbd8fb0f4ee626bdf978d51c5388e04bb26a65fb525af46e", "status": "applied"}, "evidence_asof": "2026-08-31", "evidence_receipt": {"board": "https://api.seiche.info/undertow/board.json", "board_scope": "public free subset only", "cross_product_score": null, "dispatch": "https://api.seiche.info/undertow/dispatch.json", "replay": "https://api.seiche.info/undertow/replay_coverage.json"}, "funnel": {"companions": ["https://seiche.info/articles/", "https://liquilens.in/articles/"], "primary_label": "Inspect the live Undertow board", "primary_url": "https://t.me/Undertow_LiquiLens_bot"}, "generation": {"deterministic_boundary_overlay": [], "dossier_sha256": "1fd04b84c0eff60140ca69aee2f78cfa5be3dc27e5aeff3f6a732623eb950432", "editorial_memory": {"directives": ["preserve_current_standard"], "memory_fingerprint": "sha256:f5da1f6b466f2456a7128f159b24e01e0bfa58a0159b1633b2d9ca2d45eeabc7", "rubric_version": "mqdnse.editorial-rubric.v1", "source_run_id": "sha256:c69e080b54817510bbd8fb0f4ee626bdf978d51c5388e04bb26a65fb525af46e", "status": "applied"}, "fallback_reason": null, "mode": "model_written_reviewed", "passes": 2, "review_model": "openai/gpt-5.6-terra", "review_reasons": [], "writer_model": "openai/gpt-5.6-terra"}, "headline": "Funding moves EROSION to STRAIN; segment tiers hold.", "published_at": "2026-08-31T21:50:36Z", "quality": {"failures": [], "numeric_tokens_checked": 9, "required_headings": 6, "source_links": 6, "status": "PASS", "word_count": 1230}, "slug": "2026-08-31-funding-moves-erosion-to-strain-segment-tiers-hold", "topic": "IG", "type": "current_analysis", "word_count": 1230}}, "entries": ["2026-08-31-funding-moves-erosion-to-strain-segment-tiers-hold", "2026-08-30-undertow-s-jane-street-replay-is-not-a-failed-test-it-is-a-coverage-boundary", "2026-08-29-funding-moves-strain-to-erosion-ig-remains-normal", "2026-08-28-the-june-2013-cn-replay-is-not-a-stress-verdict-it-is-a-coverage-verdict", "2026-08-27-august-2024-yen-carry-fx-replay-remains-inconclusive-under-undertow-s-coverage-rule"], "latest": "2026-08-31-funding-moves-erosion-to-strain-segment-tiers-hold", "schema": "undertow.articles.v1"}