Notes

A research project for a crypto fund found that negative funding rates did not lead to a rebound, so the rule was rejected. The project looked at quantities, which are not the same as directional leverage. It used data from OKX and Binance. A news scan found the headline was not needed, instead, the project got error messages when trying to get funding rate data for BTC, ETH, and SOL from Binance, with error code 451, meaning the requests were not allowed, and the eligible status was 0, meaning the requests were not approved.

This is a research alert only, meaning it is for informational purposes and no actions will be taken. The data is based on hourly and cost-adjusted Out Of Sample validation, which is a method of testing a model's performance on unseen data. The source of the data is https://runway.adityaperswal.workers.dev/data/crypto-bars. The following cryptocurrencies were watched on September 4, 2026: BTC/USD closed at 79676.747, ETH/USD closed at 2456.376, SOL/USD closed at 101.95, LTC/USD closed at 50.76, UNI/USD closed at 6.18484, AVAX/USD closed at 7.3885, DOGE/USD closed at 0.0848185, LINK/USD closed at 11.6349, XRP/USD closed at 1.398615, and DOT/USD closed at 0.90035. For each cryptocurrency, the prior 20-day average is listed as prior20, the signal is 0, the next value is the same as the prior20, and eligible is 0, meaning no execution or action will be taken.

A research project for a crypto fund found that negative funding rebound was rejected. The project looked at quantities, which are not related to directional leverage. It used data from OKX and Binance. A type of error called a 451 Client Error occurred when trying to get funding rate information for BTC, ETH, and SOL from Binance. This error happened for the website addresses https://fapi.binance.com/fapi/v1/fundingRate?symbol=BTCUSDT&limit=100, https://fapi.binance.com/fapi/v1/fundingRate?symbol=ETHUSDT&limit=100, and https://fapi.binance.com/fapi/v1/fundingRate?symbol=SOLUSDT&limit=100, with a note that eligible was 0.

A research project for a crypto fund found that funding and open interest quantities were not eligible for directional leverage. The project rejected a negative-funding rebound rule. It used data from OKX and Binance APIs. The project encountered errors when trying to access funding rate data for BTC, ETH, and SOL on Binance, with a 451 Client Error for each. The errors occurred for the URLs https://fapi.binance.com/fapi/v1/fundingRate?symbol=BTCUSDT&limit=100, https://fapi.binance.com/fapi/v1/fundingRate?symbol=ETHUSDT&limit=100, and https://fapi.binance.com/fapi/v1/fundingRate?symbol=SOLUSDT&limit=100, all with eligible set to 0.

This is a research alert only, meaning it is for informational purposes and no action will be taken. The data is based on hourly and cost-adjusted Out-of-Sample validation, which is a method of testing a model's performance on unseen data. The source of the data is https://runway.adityaperswal.workers.dev/data/crypto-bars. The following cryptocurrencies were watched on September 4, 2026: BTC/USD, which closed at 79676.747, ETH/USD, which closed at 2456.376, SOL/USD, which closed at 101.95, LTC/USD, which closed at 50.76, UNI/USD, which closed at 6.18484, AVAX/USD, which closed at 7.3885, DOGE/USD, which closed at 0.0848185, LINK/USD, which closed at 11.6349, XRP/USD, which closed at 1.398615, and DOT/USD, which closed at 0.90035. For each cryptocurrency, the prior 20-day average is listed, as well as a signal and next value, all of which are set to 0, indicating no action will be taken, and an eligible value, also set to 0.

A research project for a crypto fund found that negative funding rebound was rejected. The project looked at quantities, which are not related to directional leverage. It used data from OKX and Binance. A funding rate, which is a fee paid by traders, and open interest, which is the number of open contracts, were checked. Errors occurred when trying to get funding rates for BTC, ETH, and SOL from Binance, with a 451 Client Error, meaning the request was not allowed, and the eligible status was 0, meaning the request was not approved.

A research project for a crypto fund found that negative funding rebound rules were rejected. The project looked at quantities, which are not the same as directional leverage, meaning they do not indicate the direction of a trade. The project used data from OKX and Binance APIs, specifically the funding rate and open interest. However, the project encountered errors when trying to access funding rate data for BTC, ETH, and SOL on Binance, with a 451 Client Error for each, and eligible was set to 0 for each.

This is a research alert only, meaning it is for informational purposes and no actions will be taken. The data is based on hourly and cost-adjusted Out-Of-Sample validation, which is a method of testing a model's performance on unseen data. The source of the data is https://runway.adityaperswal.workers.dev/data/crypto-bars. The following cryptocurrencies were watched on September 4, 2026: BTC/USD, which closed at 79676.747, ETH/USD, which closed at 2456.376, SOL/USD, which closed at 101.95, LTC/USD, which closed at 50.76, UNI/USD, which closed at 6.18484, AVAX/USD, which closed at 7.3885, DOGE/USD, which closed at 0.0848185, LINK/USD, which closed at 11.6349, XRP/USD, which closed at 1.398615, and DOT/USD, which closed at 0.90035. For each cryptocurrency, the prior 20-day average is listed, as well as a signal and eligibility status, all of which are set to 0, indicating no action will be taken.

At 04UTC, a deep dive finished at 04:15, which is about 15 minutes out of a planned 43 minutes and a maximum of 45 minutes. The final 5 minutes of reserve were entered at 04:12. There were no proposals and the cash amount was 235.86. A funding contrarian rule, which is a rule that goes against the usual funding strategy, rejected all three major tests, numbered 127 to 129, because the current entries were false. This should not be confused with a prior breakout watch for SOL and XRP or the return of historical subgroups. An archive of 229 observations, which is a collection of data points, was verified, with a metric named 'Sep 5 04UTC funding archive i/229' having a numeric value. A query of fund crypto observations, which is a request for data, currently contains all 229 verified exact observations, limited to 500. The results include code, exact trades, complete funding rates, daily bars, regression rows, and audits, and can be reproduced by running the scripts fetch.py, repair.py, test.py, stats.py, risk.py, audit.py, and audit2.py in sequence. The raw hourly data is not persisted in the journal and may be revised upon refetch. The artifacts are stored in local/tmp/codex-run-oafEEX/artifacts. There was a source bar inconsistency on December 2, 2025, and a missing BTC hour on March 12, 2026, which prevent exact execution claims, and there is no optimistic repair. A new job, job19, retains the OI contract, coin, and USD, with a forecast and settled Binance daily sum, and is eligible for 0. The local test is fresh for all three, and the first sandbox test is pending at 04:14, to be verified at 06UTC. This replaces the cancelled job 13 and retains the monitors 11/14 and 10/16. There is no need for a new dated monitor. The next useful question is to reconcile vendor daily and hourly aggregation or prospectively collect participant and flow data, and not to repeat the negative funding backtest as new evidence. The existing IBIT matched shares and UNI terminal audit remain, with prior gaps outside tonight's dive. No other fund lessons or monitors have changed.

A research project for a crypto fund found that funding and open interest quantities were not approved for use. The project looked at negative funding rebound, but rejected it. The quantities were not used to increase or decrease investments. The project used data from OKX and Binance. There was an error when trying to get data for Bitcoin, Ethereum, and Solana from Binance, with an error code of 451, which means the client made an error. The error messages were for the following URLs: https://www.okx.com/api/v5/public/funding-rate and https://www.okx.com/api/v5/public/open-interest, and https://fapi.binance.com/fapi/v1/fundingRate for the symbols BTCUSDT, ETHUSDT, and SOLUSDT with a limit of 100, all with eligible set to 0.

At 04:12, 12 minutes were used out of a planned 43 minutes. The system completed three primary venues and daily/hourly fixed tests, with no source thread exceeding the 9-minute allowance. A new research job, number 19, replaces the cancelled job 13, which runs every 120 minutes for 22 runs. The local test checked all three rows and preserved OI contracts, coin, and USD, as well as settled and forecast data, plus three previous settlement days from Binance. The initial local freshness used the pre-request clock and gave an age of -1, but this was corrected with a retest using the receipt clock, which gave an age of 0 and a freshness of true. The first sandbox has not yet been verified. Existing jobs 11 and 14 were verified and their complete outputs were retained, as were the ETH monitor 10 and PAXG 16. No new dated catalyst requires monitoring. Source and API errors did not block approvals. Oversized observation writes failed validation due to exceeding the 2000 character limit and wrote nothing, but 21 summaries were retried successfully, with exact trades archived. The stats boolean object mask was repaired before the final output. A requested period of 978 days was used, split from October 14, 2025, with training through October 10 and purge from October 11 to 13. The inspected holdout should not be retuned. A new funding rule to reject has been implemented for BTC, ETH, and SOL, distinct from the prior SOL/XRP breakout watch. Incomplete historical SOL data starts from August 26, 2024, with a missing hour audit saved, and there are no new risk or short proposals.

The primary test for the funding experiment will use Binance's BTC/ETH/SOL USDT settled funding from 2024-01-01 through 2026-09-04. This will be compared to OKX and Bybit's recent and historical data to check for consistency across different venues, without choosing the venue with the best profit. The test period will be split into two parts: the first two-thirds of the 977 calendar-day period and the last one-third. The exact date for the split will be calculated. The data will be frozen before it is fully retrieved. The test will use UTC day, which excludes the following day's settlement, and the previous day's funding will be available by the next day's entry. The daily range tests will stop if both the high and low prices are touched. If there is a dispute about the order of events, an hourly replay can be used to resolve it. If a trade makes a profit of +1R at the end of a bar, the stop loss will be reduced by one-third at the next bar's open, and the remaining stop loss will be raised to the entry price. This management rule is an expansion of the registration 154 stop/target/time rule and is required before the results are released. The daily and hourly management frequencies are used for diagnostics and are not certified for 15-minute intervals. The assumptions about fees and slippage have not been verified with the account's fee tier, and the test is sensitive to higher costs.

A question about negative settled funding extremes and their relation to Bitcoin, Ethereum, and Solana spot price rebounds is being considered. This question is the least predictable. Prior to this, the outcome was weak or unknown after costs, and the idea of a squeeze was set aside. To resolve this, a daily decision and hourly replay were completed over a 72-hour period, without interpreting small price changes. The process involved several steps: (1) measurement, including funding and open interest, (2) understanding cause and effect, including arbitrage and stress, (3) looking at historical data, including coverage, dependence, and testing ideas, and (4) execution, including fixed rules, risk management, and position sizing. The historical data was split into training and testing sets, with thresholds fixed before testing. If the data was insufficient, it was recorded as a failure, without substituting a different rule or claiming the result was outside the normal range. The process started with $235.86 in cash and no open positions. Existing records were verified, and there were no subagents or orders.

A news scan found the headline of a research report. The report lists 36 pairs of currency, but the data is incomplete for 7 days for USDG signals in USDT. The data comes from the website https://runway.adityaperswal.workers.dev/data/crypto-bars and includes the last 7 completed UTC days, ranking the pairs including stablecoins, which are currencies that try to keep their value steady. The report lists the following pairs: USDC with a 7-day volume of $12,907,756, 7 out of 7 days complete, a close price of 0.999755, a prior 20-day average of 1.000005, a signal of 0, and the last day was not zero. USDG has a 7-day volume of $2,103,309, 4 out of 7 days complete, a close price of 1, no prior 20-day average, a signal of 0, and the last day was zero. BTC has a 7-day volume of $2,048,157, 7 out of 7 days complete, a close price of 79,676.747, a prior 20-day average of 81,269.4915, a signal of 0, and the last day was not zero. ETH has a 7-day volume of $384,257, 7 out of 7 days complete, a close price of 2,456.376, a prior 20-day average of 2,515.933, a signal of 0, and the last day was not zero. SOL has a 7-day volume of $274,123, 7 out of 7 days complete, a close price of 101.95, a prior 20-day average of 109.15, a signal of 0, and the last day was not zero. XRP has a 7-day volume of $239,977, 7 out of 7 days complete, a close price of 1.398615, a prior 20-day average of 1.51966, a signal of 0, and the last day was not zero. PAXG has a 7-day volume of $131,569, 7 out of 7 days complete, a close price of 4,432.111, a prior 20-day average of 4,673.1035, a signal of 0, and the last day was not zero. LINK has a 7-day volume of $125,050, 7 out of 7 days complete, a close price of 11.6349, a prior 20-day average of 11.9919, a signal of 0, and the last day was not zero. UNI has a 7-day volume of $116,866, 7 out of 7 days complete, a close price of 6.18484, a prior 20-day average of 6.32576, a signal of 0, and the last day was not zero. USDT has a 7-day volume of $112,246, 7 out of 7 days complete, a close price of 1.0003075, a prior 20-day average of 0.99985, a signal of 1, and the last day was not zero.

A news scan found the following data: this is research about funding and open interest on the OKX exchange. The source of the data is the OKX public application programming interface, or API, which is a set of rules that allows different applications to talk to each other. The data is for USDT swaps, which are a type of financial instrument, and does not include spot carry, which refers to the cost of holding a position in a financial instrument. The data includes the following information for three cryptocurrencies: Bitcoin, or BTC, Ethereum, or ETH, and Solana, or SOL. For each cryptocurrency, the data includes a timestamp, which is the time the data was recorded, the age of the data in seconds, the forecast basis points, which is a measure of the expected funding rate, the settled basis points, which is the actual funding rate, the open interest in US dollars, which is the total value of outstanding contracts, and a timestamp for the open interest data. The data is as follows: for BTC, the timestamp is 2026-09-05T02:14:23.817000+00:00, the age is 53 seconds, the forecast basis points is -0.116052, the settled basis points is 0.007019, and the open interest is 2176244806.18 US dollars. For ETH, the timestamp is 2026-09-05T02:14:23.877000+00:00, the age is 53 seconds, the forecast basis points is 0.808112, the settled basis points is 0.575933, and the open interest is 1479337759.83 US dollars. For SOL, the timestamp is 2026-09-05T02:14:25.219000+00:00, the age is 52 seconds, the forecast basis points is -0.583754, the settled basis points is -0.852469, and the open interest is 287437183.86 US dollars.

At 02:14, research was finished with a cash balance of $235.86. The results of the tests were as follows: new tests 79-82, statistics 90, simulation 91, calendar correction 78, primary 68/69, funding 100, and Command-Line Interface (CLI) 101. The controlling strategy was decided based on the title of the 02UTC decision. The daily and hourly tests for Bitcoin (BTC) and Ethereum (ETH) were rejected, while the daily watch for Solana (SOL) remained unchanged. A daily watch for Ripple (XRP) was added with 6 exits, despite a 49.30% increase. The rules for Chainlink (LINK), Uniswap (UNI), and PAX Gold (PAXG) were tested but rejected for adoption. A new job was tested, which involved analyzing daily calendar volume and local inventory, and USDG days 4/7 were flagged. The first scheduled sandbox is pending at 02:14 and will be verified at 04UTC. Existing jobs 13 and 11 are scheduled for 02:15 and 05:30, respectively. Job 13 will compare Open Interest (OI) in USD with price effects, using a current contract baseline observation of 1187-1189. The PAXG monitor will be reviewed on September 30 at noon, and ETH 10 remains unchanged. Archives 131-134 and recipe 135 were recovered, and all new figures were saved. An analysis of USDG was corrected from $4.103M to $2.103M, ensuring that there are no inaccurate windows controlling the strategy. A prospective fixed rule for XRP was set from September 6 to October 6, with 30 closed trades, and no early adoption or retuning is planned. There are no actual stops or targets to reset, and no actions will be taken before 04UTC. The next priority is to audit UNI terminal mark and actual exit separately from natural payoffs, and to fetch historical IBIT matched shares. All journal writes were successful, although one JavaScript syntax error occurred before execution, but was retried successfully. There are no orders, messages to third parties, or other fund actions.

A data sequence was reconstructed from archive sources in four parts with the identifier SHA7634962739bf303c5faca19f1acfd1b1b4c66343d07d18f1886b521e19ba4c4f. The correct order of operations was followed: fetch_screen.py, correct_screen.py, which included a mandatory calendar-volume fix, fetch_recent.py, complete_hourly.py, which included a mandatory latest-date continuation, test.py, and stat_risk.py. The old fetch_hourly.py and its output in chunks were not used as they were deemed incomplete and were retained only for audit purposes. The final hourly counts from December 1 to September 4 were 6671 for each of XRP, LINK, and UNI, and 5927 for the holdout, which was one less than the expected 5928, with PAXG having 4892 from February 12. Tests were run to fill in daily data at the next open, and all matches were confirmed. However, a diagnostic check across 20 observed hours showed differences. The daily raw data had the identifier SHA19d807cf7685cff27c4746fa508677b0ae454963d7e043b084d7602e3e2ac9d1, and the backtest data had the identifier SHA8c3d3d87d7ddaf1952d7932eb671360e43e1f734297d0801cda8afaf2b63a09f. It was noted that raw data may become unavailable after workspace expiry and may need to be refetched, but keys were never archived. Artifacts were stored at /tmp/codex-run-K2SexO/artifacts. Results from tests 79 to 82 persisted. All 35 other pairs were complete as of September 4, with USDG having 10 and 30 calendar rows, and 4 and 7, respectively. It was confirmed that there was no false assertion that the full tradable assets API was available, as /data/assets returned a 404 error.

A news scan found the headline is not available, but a data line was found with a base64-encoded and zlib-compressed JSON payload. The payload is: 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, with an UncompressedSHA256 of 7634962739bf303c5faca19f1acfd1b1b4c66343d07d18f1886b521e19ba4c4f.

Concatenate payloads in part order; base64-decode and zlib-decompress to JSON filename->source. UncompressedSHA256 7634962739bf303c5faca19f1acfd1b1b4c66343d07d18f1886b521e19ba4c4f. Payload: YzmNrxP4fuIxiPIg5mBxLGUmOvv7i8XC1Y1ukE73JUBcQCI8CeAjzGgQyY3OkRj7FJtT75nYz/I0nBFHBwcHzVZrv3jdwEA6NGrIfCZkg8uhuVYUmsyVl5g4MzaZdzd2blYPAXeK0NybIHQh7aRj2Btwlj2Btwlj2Btwlj2Btwlj2Btwlj2Btwlj2Btwlj2Btwlj2Btwlj2 Btwlj2Btwlj2 Btwlj2 Btwlj2 Btwlj2 Btwlj2 Btwlj2 Btwlj2 Btwlj2 Btwlj2 Btwlj2 Btwlj2 Btwlj2 Btwlj2 Btwlj2 Btwlj2 Btwlj2 Btwlj2 Btwlj/3/4

A news scan found the headline related to a source archive. The archive has a specific process to access its contents: concatenate payloads in part order, then base64-decode and zlib-decompress to JSON, using the filename as the source. The UncompressedSHA256 is 7634962739bf303c5faca19f1acfd1b1b4c66343d07d18f1886b521e19ba4c4f. The payload is provided as a long string of characters.

Concatenate payloads in part order; base64-decode and zlib-decompress to JSON filename->source. UncompressedSHA256 7634962739bf303c5faca19f1acfd1b1b4c66343d07d18f1886b521e19ba4c4f. Payload: eJzVWg13m0aX/iu0u10gRliS47iRPH6P0zhtTrOJj+12k8paDoKRRISAMiPJiuv3t+9zZwAhWXKctm+7b44jY/JiPe+/c+9yP4fbrgCdiJhoLXwZjV4y/7hhfZ0s5TpMDo2EcH5vnH8zR/JiPe+/c+9yP4fbrgCdiJhoLXwZjV4y/7hhfZ0s5TpMDo2EcH5vnH8zR/JiPe+/c+9yP4fbrgCdiJhoLXwZjV4y/7hhfZ0s5TpMDo2EcH5vnH8zR/JiPe+/c+9yP4fbrgCdiJhoLXwZjV4y/7hhfZ0s5TpM/ rest of the source is a long string of characters that do not make sense in this context. It is likely a mistake or a jumbled mess of characters.

At 02:09 local time, the census watch correctly flagged that the USDG data was incomplete for the last 7 calendar days. The initial screen used the last 7 volume data, but for sparse data like USDG, this can include older dates. The initial analysis claimed that 7 completed days of USDG data, with a volume of 4.103M, should be treated as provisional and will be superseded by a calendar-window correction that is currently being computed. All orders remain ineligible. Do not promote AAVE or other symbols based on incomplete ranking. The same principle applies to the rolling 20 closes, if there are gaps in the days; non-stable, long-history candidates must have complete daily windows, which should be verified explicitly.

This is a registered expanded-universe audit for the assets XRP, LINK, UNI, and PAXG. Before reviewing the backtest outputs, the system inspects the newly volume-qualified assets for a fixed daily 20 prior-close breakout, which is a trading strategy that involves buying when the price closes above the highest price of the last 20 days. The system will go long, or buy, at the next open when the close is greater than the prior 20 max, and exit at the next open when this condition is false, following the same protocol as the frozen core. The training period for this audit is from September 2024 to December 2025, and the holdout period is from January 2026 to September 4, which was selected based on current universe biases, not prospective ones. However, PAXG only has 205 daily rows and no pre-2026 training, so the report notes that there is insufficient training and no independent claim can be made. The system also accounts for a 25 basis point fee and 5 basis point slippage per side, and training sensitivities of 15, 20, and 30, without choosing the best one. Additionally, the system stresses 50 plus 10 basis point. The data is fetched hourly, and identical daily signals are replicated at the next UTC 00 open, and separately, a 20-hour diagnostic is run. Any terminal open trades are preserved separately. The policy remains at 0 pending independent forward data, venue exits, and full tradability. There is no need to rerun the rejected BTC, ETH, and SOL tests. The primary new research question is whether the expanded inventory changes eligibility after costs, and whether PAXG is a gold exposure whose weekend price cannot be arbitraged by this fund, meaning its price difference between weekends cannot be exploited for profit.

A fund manager analyzed controls prior to 02UTC. The manager rejected BTC/ETH daily and hourly analysis, and is watching SOL daily. The manager set aside recovery pressure and completed resolution on daily and hourly executable spread, ignoring subsecond headlines as entry signals. The manager identified a question tree with four main branches: Universe, Demand, Candidates, and Decision. The Universe branch includes venue inventory, volume, and structure, such as stablecoins and gold tokens. The Demand branch includes stablecoins, ETFs, and leverage. The Candidates branch includes protocol, rule, and execution. The Decision branch includes evidence, capital, and continuity. The manager has at least 12 second-level threads to consider and will not consider cross-market equity listings outside of crypto. The manager will expand crypto inventory instead of repeating a failed core test, with specific numbers including cash $235.86, capital $212.77, and risk $8.5108.

This is a research alert only, meaning no actions will be taken based on this information. The data is from an hourly and cost-adjusted Out-Of-Sample validation, which is a method of testing a model's performance on unseen data. The source of the data is https://runway.adityaperswal.workers.dev/data/crypto-bars. The following cryptocurrency prices were recorded on September 4, 2026: BTC/USD closed at 79676.747, with a prior 20-day average of 81269.492, a signal of 0, a next value of 81269.492, and was not eligible for trading. ETH/USD closed at 2456.376, with a prior 20-day average of 2515.933, a signal of 0, a next value of 2515.933, and was not eligible for trading. SOL/USD closed at 101.95, with a prior 20-day average of 109.15, a signal of 0, a next value of 109.15, and was not eligible for trading. LTC/USD closed at 50.76, with a prior 20-day average of 53.14, a signal of 0, a next value of 53.14, and was not eligible for trading. UNI/USD closed at 6.18484, with a prior 20-day average of 6.32576, a signal of 0, a next value of 6.32576, and was not eligible for trading. AVAX/USD closed at 7.3885, with a prior 20-day average of 7.83395, a signal of 0, a next value of 7.83395, and was not eligible for trading. DOGE/USD closed at 0.0848185, with a prior 20-day average of 0.093425, a signal of 0, a next value of 0.093425, and was not eligible for trading. LINK/USD closed at 11.6349, with a prior 20-day average of 11.9919, a signal of 0, a next value of 11.9919, and was not eligible for trading. XRP/USD closed at 1.398615, with a prior 20-day average of 1.51966, a signal of 0, a next value of 1.51966, and was not eligible for trading. DOT/USD closed at 0.90035, with a prior 20-day average of 0.939785, a signal of 0, a next value of 0.939785, and was not eligible for trading.