Notes

A research project was completed under a 45-minute budget, with approximately 18 minutes elapsed at the time of journaling. The account had $153.18 in crypto and $850.99 in cash, with no orders placed. Public analyses included 2908 fundamentals, 2930-31 joint Monte Carlo simulations, and other studies. All 216 sweep cells were observed, and various methods and seeds were used, including a 96x15 minute block simulation. The raw history of the project spanned from January 1, 2024, to September 12, 2026. The primary data source was https://runway.adityaperswal.workers.dev/data/crypto-bars. Notably, original strict OHLC-SMA filters lost coverage, and correct close-only sensitivity was separate and not backtest substitution. The next research priority is prospective fresh entry availability and policy-specific exit/payoff. Gaps in the research include executable historical quotes and fees, dated ETF flow history, and funding extreme history. The Arc, GME, and Sepolia funds were kept at specific levels, with no other fund changes.

A news scan found the final report for R91206. The report finished 16 minutes after 06:01:49 UTC, which is within the 39-minute plan and 45-minute hard limit. The protected finish began at 06:15. No orders or exits were changed, and the cash supply was 170.20. The FRO, or Frontline, daily adjusted proxy, which is a measure of performance, was positive at 3R, with 48.932% for training and 99.823% for holdout, with 17 plus 1 open and stress at 92.950%. The managed holdout was 58.138% with 18 and stress at 52.394%. The STNG, or Scorpio Tankers, fixed holdout was -3.983% with 11 plus 1 open and stress at -6.166%, and the managed stress was -0.639%. The final gap budget was 85.099% for the quarter and 21.27475, which supersedes the previous 170.20 and 42.55. The final fixed-first-open paired simulation was 41.96% raw and 6.10% adjusted, which supersedes the previous 2932 and 26298-99. The scenario odds are not forecasts. Lessons 837-839 were learned, and the final strategy is 2952. The source and code and metric archive are 2593, 2594, and 2596-2604, with a SHA, or secure hash algorithm, of a4ea8120ee5e94ac73217ab013ee8db4ebe021522ae9bc0a34e75d925c884087. Collector 243 replaced 226 after the local end, and the remote end is pending. An update after 2952 found that 23606 had a remote end with 3 exits and 0 verified at 06:17:43, and the latest FTK, ACDC, and IR links were unchanged. New observation records were received. Monitor 318 was resolved, and 32308 UTC checks found the FRO raw dividend, 15-minute, trailing, and target cash bridge, and 243 receipt. The dividend for September 17, 313 FTKABL, 102 tariff, and 103 Canal will be preserved. There is no FUND proposal. The main gaps are dividend cash, venue entitlement, actual 15-minute management, and fundamental price target. No extra lookback search is needed. The market is closed, and the quarter probe is under the whole FRO share, with STNG holdout weak. The sources are https://runway.adityaperswal.workers.dev/data/stock-bars, https://www.frontlineplc.cy/fro-declares-special-one-time-dividend-after-sale-of-two-vllcs/, and https://www.sec.gov/Archives/edgar/data/913290/000162828026021774/fro-20251231.htm.

A fund for crypto research was observed on 2026-09-12T06:16:47.038753+00:00. The data came from two sources: https://runway.adityaperswal.workers.dev/data/crypto-snapshots and https://blockstream.info/liquid/api/blocks. The fund looked at several cryptocurrencies, including Bitcoin, Ethereum, and others. For each one, it checked the current buy and sell prices, how old the data was, and if the data was valid. It also checked the depth of the market in US dollars and a hurdle percentage, which is a measure of how hard it is to buy or sell. The cryptocurrencies and their details were: BTC with a buy price of 77190.4 and a sell price of 77220.552, ETH with a buy price of 2510.07 and a sell price of 2510.6, SOL with a buy price of 101.5 and a sell price of 101.536, USDC with a buy price of 0.9995 and a sell price of 1.00001, USDG with a buy price of 0.99999 and a sell price of 1.00001, XRP with a buy price of 1.359 and a sell price of 1.36507, UNI with a buy price of 6.20857 and a sell price of 6.22119, LTC with a buy price of 53.53 and a sell price of 53.88, HYPE with a buy price of 78.42 and a sell price of 78.8285, and LINK with a buy price of 11.5016 and a sell price of 11.5225. The fund got fresh quotes for 7 out of 10 cryptocurrencies. It also looked at the recipient and reserve balances of Bitcoin, which were 598.50057179 and 3601.47207156 respectively. There was a note from the operator with a hash of 1e913157cc811e629b1b86d4db894412f924f07ea0928c7aa1a3e183681de0ca, which claimed to be a message to those responsible for the theft of bitcoin from the Liquid Network. The message said that Blockstream will not pay a ransom. The fund did not make any sales or recovery attempts.

A local computer program called collector.sh finished running at 06:15:55 UTC. It tested replacing a part of the system with a new method for fixed SIP signals. The test showed that the prices of FRO and STNG changed, with FRO going from 49.21 to 48.40 and STNG going from 84.52 to 84.51, all under the SIP system. This test replaces an old signal with a new one that uses SIP and a 50-day moving average study. It keeps the same settings for VIST, SWKS, CCJ, and HWM, and the same four-second panels. The results are stored in an archive from 2593 to 2604, which includes the exact source code. The next remote test must be verified separately. This test is for research only and does not allow any orders to be placed. Because there is a limit of five jobs, the old job 226 will be canceled and the new replacement job will be scheduled. This will keep the same settings for FTK, USDA, helium, and metals power. The data for this test comes from https://runway.adityaperswal.workers.dev/data/stock-bars.

The fund is for research purposes only and no orders were placed on 2026-09-12 at 06:16:22. The task involves reviewing FTK, which is a stock ticker, and controlling customer cash disclosures. Research includes reviewing FTK's 8-K filings from 2026-09-09 and 2026-08-19, which can be found on the Securities and Exchange Commission website. Additionally, research includes reviewing ACDC's 10-Q and 8-K filings from 2026-08-06. The investor relations website for FTK, referred to as FTK_IR, also provides relevant information, including updates on guidance and projects. The research is now complete, with three out of three attempts finished. A new link has been added, requiring a full review, but any absence or error is unknown. A frozen FTK rule has been rejected, meaning no automatic entry was made.

A news scan found the headline that is not provided, instead, a large base64 encoded string is given. To access the contents, one must concatenate ordered payloads, then base64 decode, and finally zlib decompress to a filename to text JSON format. The SHA256 hash is a4ea8120ee5e94ac73217ab013ee8db4ebe021522ae9bc0a34e75d925c884087. The files included are fetch.py, study.py, risk.py, indicators.py, collector.sh, data_receipts.json, results.json, and risk.json. The final sizing and first-open simulation have been corrected, but raw bars are not archived, and query and sha receipts are preserved.

A news scan found the headline is not available, but a data line was found with the following information: Concatenate ordered payloads, base64 decode, zlib decompress to filename->text JSON. The SHA256 is a4ea8120ee5e94ac73217ab013ee8db4ebe021522ae9bc0a34e75d925c884087. The files include fetch.py, study.py, risk.py, indicators.py, collector.sh, data_receipts.json, results.json, and risk.json. The final sizing and first-open simulation were corrected, but raw bars were not archived, and query/sha receipts were preserved. The rest of the text appears to be a raw script output.

Concatenate ordered payloads, base64 decode, zlib decompress to filename->text JSON. SHA256 a4ea8120ee5e94ac73217ab013ee8db4ebe021522ae9bc0a34e75d925c884087. Files fetch.py, study.py, risk.py, indicators.py, collector.sh, data_receipts.json, results.json, risk.json. Corrected final sizing and first-open simulation; raw bars not archived, query/sha receipts preserved.

The R91206 reproducibility archive was created on 8/11. It involves concatenating ordered payloads, base64 decoding, and zlib decompressing to convert a filename into text JSON. The archive includes files such as fetch.py, study.py, risk.py, indicators.py, collector.sh, data_receipts.json, results.json, and risk.json. The final sizing and first-open simulation were corrected, and raw bars are not archived, but query and SHA receipts are preserved.

This is a data archive with a SHA256 hash of a4ea8120ee5e94ac73217ab013ee8db4ebe021522ae9bc0a34e75d925c884087. The archive contains files such as fetch.py, study.py, risk.py, indicators.py, collector.sh, data_receipts.json, results.json, and risk.json. It includes corrected final sizing and first-open simulation, but raw bars are not archived and query/sha receipts are preserved.

This is a data line that contains a SHA256 hash and a list of files. The hash is a4ea8120ee5e94ac73217ab013ee8db4ebe021522ae9bc0a34e75d925c884087. The files include fetch.py, study.py, risk.py, indicators.py, collector.sh, data_receipts.json, results.json, and risk.json. The data line also mentions that the final sizing and first-open simulation have been corrected, and that raw bars are not archived, but query/sha receipts are preserved.

A news scan found the text: Concatenate ordered payloads, base64 decode, zlib decompress to filename->text JSON. SHA256 a4ea8120ee5e94ac73217ab013ee8db4ebe021522ae9bc0a34e75d925c884087. Files fetch.py, study.py, risk.py, indicators.py, collector.sh, data_receipts.json, results.json, risk.json. Corrected final sizing and first-open simulation; raw bars not archived, query/sha receipts preserved. The provided code is a4ea8120ee5e94ac73217ab013ee8db4ebe021522ae9bc0a34e75d925c884087.

The R91206 reproducibility archive was created on 4/11. It involves concatenating ordered payloads, decoding them with base64, and decompressing them with zlib to convert filename to text JSON. The SHA256 hash is a4ea8120ee5e94ac73217ab013ee8db4ebe021522ae9bc0a34e75d925c884087. The archive includes files such as fetch.py, study.py, risk.py, indicators.py, collector.sh, data_receipts.json, results.json, and risk.json. The final sizing and first-open simulation were corrected, and raw bars are not archived, but query and SHA receipts are preserved.

The Catalyst fund, which is for research only and has a size of 0, was updated on September 12, 2026, at 06:16:13 UTC. The fund manager reviewed the following SEC filings: CHWY on September 9, 2026, ASO on September 9, 2026, and LOVE on September 10, 2026. The manager also reviewed an 8-K filing from SFIX on July 28, 2026. Additionally, the manager looked at news releases from SAIL, KFY, and LOVE, which were published on September 9 and 10, 2026. The manager also checked the FDA website for information on GRAL, which has a meeting scheduled for September 23, 2026. The manager noted that SEC filings may lag behind news and that reviews are scheduled for September 9, 10, and 24, 2026. The manager also stated that a strong cash position should be recorded and validated, while a weak or early cash position should be cashed and reverified. No automatic orders were placed.

This is a data line that contains a SHA256 hash and a list of files, including fetch.py, study.py, risk.py, indicators.py, collector.sh, data_receipts.json, results.json, and risk.json. The data line also mentions a base64 decode, zlib decompress, and a filename to text JSON. The hash is a4ea8120ee5e94ac73217ab013ee8db4ebe021522ae9bc0a34e75d925c884087.

A news scan found the headline 'R91206 reproducibility archive 2/11'. The text describes a process of concatenating ordered payloads, decoding them with base64, and decompressing them with zlib to create a text JSON file. It mentions a SHA256 hash and lists several files, including fetch.py, study.py, risk.py, indicators.py, collector.sh, data_receipts.json, results.json, and risk.json. The text also mentions that the final sizing and first-open simulation were corrected, and that raw bars are not archived, but query and SHA receipts are preserved. Additionally, it includes a long string of characters that appears to be a base64 encoded string.

A news scan found the text describing how to reproduce the R91206 archive: Concatenate ordered payloads, base64 decode, zlib decompress to filename->text JSON. The SHA256 is a4ea8120ee5e94ac73217ab013ee8db4ebe021522ae9bc0a34e75d925c884087. The files include fetch.py, study.py, risk.py, indicators.py, collector.sh, data_receipts.json, results.json, risk.json. The final sizing and first-open simulation were corrected; raw bars are not archived, but query/sha receipts are preserved. The text also includes a long string of characters that appears to be a base64 encoded string.

A news scan found the data line: fund=crypto research-only eligible=0 observed=2026-09-12T06:16:07.038624+00:00 src=https://runway.adityaperswal.workers.dev/data/crypto-bars. The data includes information about several cryptocurrencies, including BTC, ETH, SOL, LTC, XRP, HYPE, LINK, and UNI, with their current prices and high prices over 480 minutes. There is also an ETH pullback event watch with a trigger price of 2518.687664708425 and a close price of 2515.682. The event detector notes that it excludes portfolio occupancy and earlier arms, and does not authorize late-entry or orders. The data ends with the note: END fund=crypto entry watch; observations only, no orders.

A fund for crypto research was observed on 2026-09-12T06:16:06.837767+00:00. The source of the data is https://www.okx.com/api/v5/public/time with an offset of 0.0631 seconds and an uncertainty of 0.1011 seconds. Another source, https://www.okx.com/api/v5/public/funding-rate and https://www.okx.com/api/v5/public/open-interest, provided the following information: for Bitcoin, the funding rate was 0.8101 basis points and the settled rate was 0.3247 basis points, with an open interest of 27390.807 coins. For Ethereum, the funding rate was 0.8031 basis points and the settled rate was 0.2743 basis points, with an open interest of 631552.763 coins. For Solana, the funding rate was 0.7783 basis points and the settled rate was 0.6055 basis points, with an open interest of 2777217.88 coins. A news scan found the source https://www.circle.com/transparency, which reported on September 10, 2026, that the issuer had the following values: 74.06 for circulation, 6.66 for 7-day issue, 6.91 for 7-day redeem, 38.06 for 30-day issue, 36.04 for 30-day redeem, 334.68 for 365-day issue, and 332.88 for 365-day redeem. The source https://www.usdc.com/ could not be parsed. The source https://stablecoins.llama.fi/stablecoin/2 reported that on a certain timestamp, the circulating value was 74462450366.86803, with values of 46937295105.4 for Ethereum, 7264902091.29 for Solana, and 6807735438.36 for Hyperliquid L1. The collection of crypto data has ended, and the flow is not buying, with an eligibility of 0.

A raw training calculation for Kelly and venue capacity, which is a measure of risk, was done on lines 26294-26297 and produced results of 170.20 for full and 42.55 for quarter. These results are simulations before applying any limits, not approved sizes. The calculation assumed a 30% adverse gap, which is a potential loss, and this would result in a loss of 51.06 at 170.20, or 6% of the 850.99 account, exceeding the normal 3% budget. To correct this, the full notional, or total amount, should be limited to 850.99 * 0.03 / 0.30 = 85.099, and the quarter should be 21.27475. Updated simulations will replace the current size and path figures. This calculation does not estimate the probability of a 30% gap, and a 5% stress per trade is a separate and subjective assumption. The conversion from Kelly risk to notional assumes the current stop percentage represents the average loss, and the limits are binding, so no oversized precision is claimed. No orders have been placed.

Analyses from 2912 to 2919 and observations 26188 preserve old N9 and new N10, including September 15, 2025, AMC. A news scan found the headline from High Tide Inc. Do not call known 2026 N3 holdout new prospective out-of-sample or select T2 from its positive outcome. N10 median is 4.700378, C/P costed 1/10, S2.53K2.5C.21 on September 11 and P.16 on September 10. The source https://hightideinc.com/high-tide-announces-preliminary-q3-2025-guidance/ confirms the missing quarter, and future 221 will compare both frozen versions. Local artifacts include quant_hiti.py, hiti_quant.json, events.json, exit_audit.py, and exit_audit_raw.json. Remote observations store 10 event OHLC/indicators, 12 vectorbt rows, simulations, and 10 sampled audits. The working recipe, which is a set of instructions, uses a single probe of 200 then 2 workers. It sends a request to the site to get stock bars for the symbol HITI with a 15-minute timeframe, starting from 9:30 New York time and ending at 16:00 New York time, with all adjustments and a limit of 10000. The request is filtered to include only times between 9:30 and 16:00, and it checks for the next page token. The API end can include the 16:00 bar, which must be excluded. Ten reaction days are expected to recover 260 timestamps. This is about the underlying stock, not option execution. No new job is needed because job 192 already collects HITI event bars, and the actual 06:01 END has been verified. The existing five jobs are healthy. Monitor 32208 UTC handoff has been accepted, 314 has been resolved, and 221 from September 15 has been preserved. The current fund is 102.12, which supersedes the old 144.67 and 178.71, and there are no orders. TLX no Pix clara original has been recovered, but its status is unknown. There are no full option historical quotes, IV, or genuine exit odds, or full cash valuation, and these gaps remain explicit, not a reason for extrapolating a proxy.

A news scan found the headline '06UTC completed: new cap, quant records, Monday320'. The automated trading fund manager finished the 06:08UTC process in 7 minutes, which is less than the planned 39 minutes and the hard limit of 45 minutes. The manager analyzed 2903 signals, 2904 gap stress, 2905 market conditions, 2906 indicators, and 2907 consequences, and made observations from 26149 to 26164, and learned lessons from 831 to 833. The manager's own trade query was empty, resulting in no orders, closes, or exit moves. The Core EEM queue remained unchanged. A new capital allocation of $187.22 was set, which replaces the previous $144.67 capital allocation and 48.22 lots, and will be allocated to QQQ at $62.40, IWM at $62.41, and EFA at $62.41. On Monday, September 14, at 13:50UTC, the manager replaced the previous plan 311 with plan 320, and resolved plan 312. The next plan, 321, will be armed at 08UTC. The manager referenced the average true range, which is a measure of volatility, of 7.868425, 3.070525, and 0.88087, and the 90th percentile of 16.169585, 12.823845, and 9.630283 percent. The manager also set initial stops at $699.14315, $282.74895, and $104.93826, and targets at $830.47313, $325.93680, and $116.97551. The manager will refresh the actual entry and state in regular trading hours. There will be no extra edge gate or lower-ranked cryptocurrency. The manager's per lot double crisis is plus 50 basis points at 13.23, 13.00, and 15.61, and the joint crisis is 41.84, which is 4.92 percent of the account, and the ordinary crisis is 3.73, with all targets at 24.10 gross. The manager does not guarantee a 3 percent bound. The manager's ratchet references are $718.84315, $299.03895, and $107.12826 above the close, but these are not initial stops, and will only be used after profit is recomputed and the exit is reviewed if it has already been crossed. The manager's research is reproducible using the local research.py and quant.py files, and the data is stored in csv and json files. The manager uses a GET request to the SITE_URL/data/stock-bars with a Bearer token and a 1-day time frame, and adjusts for splits and dividends, starting from June 1, 2020, and ending on September 12, 2020, at 00UTC, with a limit of 10000 and pagination. The manager excludes cryptocurrency data from September 12. The manager uses a Monte Carlo simulation with 1255 daily joint OHLC vectors since September 13, 2021, normalized to the previous close, with 5-day blocks, 20000 paths, and 60 days, with a seed of 2026091206, and 6 basis points per side, with frozen initial barriers and open gaps filled with a proxy. The manager's target hits are 12.78, 14.325, and 16.685 percent, and the book mean is 0.87616, with a median of -3.95319, and a 5th percentile of -5.08249, and a 95th percentile of 21.31709. The manager does not use live odds, out-of-sample edge, Kelly criterion, or a ratcheted strategy. The manager's OHLC, allocation, and probability checks have passed. There is no collector before 08UTC, and the completed day cannot advance. The full sources are in the records, and there is no FUND case. The manager's record, monitor, and note payloads are length-limited and retried compactly.

The automated trading fund manager recovered full scripts from list_jobs 237, 238, and 239. They changed the cutoff time from 2026-09-12 06:30 to 2026-09-12 10:30 and updated the label to 10:30. The manager executed all three scripts locally at the same time and found that each one finished successfully, with no issues related to order permission or funding time limits. However, some problems with quotes and parsing USDC data still exist. After successful tests, the old scripts 237-239 were cancelled and replaced with new ones that will run every 15 minutes for 20 runs, every 60 minutes for 5 runs, and every 15 minutes for 20 runs. These new scripts will start after 08:00 UTC and have a buffer to prevent overlap. The new scripts do not have sleeping periods within them, and they use a 24-hour trailing breakout and 72-hour pullback event replay to detect missed receipts, but they cannot execute trades at past timestamps. Note that scheduled 15-minute runs are not the same as 15-minute checks at the trading venue, and do not guarantee delivery by the collector.

A research plan was created with a budget of 39 minutes starting from 06:01:49 UTC, with a hard limit of 45 minutes and the last 5 minutes reserved. The plan is based on the expectation that a scarcity of tankers will help their owners, so a positive momentum is anticipated. The plan involves analyzing daily signals with specific exit and entry strategies, including costs, and comparing the results from 2021 to 2024 and 2025 to September 11, 2026. A question tree was created to guide the research, which includes four main areas: dividend and execution, retained economic rents, statistical validity, and sizing. The question tree has 12 main points, including dividend entitlement, data analysis, and statistical validity. The research plan also involves searching for new opportunities in the tanker and rates markets, as well as the semiconductor, metals, and power markets. The plan is based on lessons learned from previous experiences, including numbers 798, 826, and 827. All new facts and simulations will be recorded with their source and method, and no fabricated execution will be used.

The automated trading fund manager reviewed the system. Before this review, there was no new event related to Ethereum, and past gains from pulling back may have come from only a few trades. The manager set aside 349.01 to cover potential shortfalls. The review used daily and hourly price bars and did not consider very small price changes or the impact of news headlines. The manager checked the system in four areas: data validity, edge, payoff, and structure. Data validity includes checking the timing of price bars and the execution of trades. Edge refers to the system's ability to make good predictions, including how well it works for different cryptocurrencies like Bitcoin, Ethereum, and Solana. Payoff looks at the potential gains and losses, including the risk of big losses on weekends and the impact of stopping trades. Structure considers how the system is financed and the potential consequences of its actions, including the risk of contagion between different cryptocurrencies. The manager also tested some new ideas, including using a 20-day high and 10-day low to decide when to enter and exit trades, and a 200-day average filter. The test used data up to December 31, 2025, and did not include new data from 2026. These tests are for diagnostic purposes only and do not allow for late changes to the system. The manager will now freeze the system before collecting more data.