Notes

A news scan found the script output: a Python script is running a trading test. The script uses a formula to calculate the entry and exit points for trades: e)-(1 if h>0 else 0);i=j-(h-1 if h>0 else 0). If the current index is not in the quotes index, the script skips to the next iteration. The script then calculates the entry price, stop loss, and target price using the average true range, or atr, which is a measure of volatility. The script checks if the mode is set to 'barrier' and if so, it loops through each bar from the entry point to the exit point, checking for certain conditions such as if the open price is less than or equal to the stop loss or if the high price is greater than or equal to the target price. The script then updates the size and price arrays and calculates the net return for the trade. The script also appends a dictionary to the trades list with information about the trade, including the event date, source, and reason for the trade. The script uses the vbt library to create a portfolio from the orders and calculates the return and drawdown for the portfolio. The results are then written to a json file. The script also prints the results for the base case with a slip of 0.001. The script then calculates the average true range, or atr, for the stocks 'AVO' and 'UPB'.

A Python script is being used to test an automated trading system. The script uses libraries such as numpy, pandas, and vectorbt. It sets a random number generator with a seed of 2026090808. The script defines a function called bars that loads data from a json file, converts the data to a pandas dataframe, and sorts the data by time. The function also renames the columns of the dataframe to open, high, low, close, and volume. The script then loads data for the stocks AVO and SPY. It calculates the average true range, a measure of volatility, over a 14-day period. It also calculates the simple moving average over a 50-day period and the momentum over a 20-day period. The script then loads a list of events from a json file and sorts them by date. The script defines a function called stats that calculates various statistics such as the mean, median, and standard deviation of an array of numbers. The script then runs a series of simulations with different parameters, including the hours of the day to trade, the mode of trading, the amount of slippage, and the split of the data into training and holdout sets. For each simulation, it calculates the statistics of the trades and stores them in a list called results.

A new study is being conducted on AVO, which was selected after the issuer confirmed information on August 25 and September 8 after the market closed, as stated on the company's website. The study reviewed 67 SEC 8K filings, including 23 routine earnings reports from January 2021 to June 2026, excluding the December 19, 2024 report due to a pre-announcement on November 5. The baseline for the study was set prior to the session open and close, using data from 2021 to 2024 for training and 2025 to 2026 for holdout. The sensitivity of the study was set at 3/5 and the stop loss at 1 ATR, with a target of 1.5 ATR, using daily bars and a conservative stop loss strategy. Fees were set at 10 basis points plus a 10 basis point slip on each side, with a stress test of 25 basis points slip. The study only used same-day open-close data until independent AMC timing was verified, and trading was not allowed on September 8 from the prior session edge. The availability of date notices must be verified before promotion, and the source 8K announcement date is not known in advance. A structural break due to the Calavo merger on May 28, 2026, was separately flagged. There was no tuning from the holdout data. The study used vectorbt long only to reconcile compounded returns, with a bootstrap of 100,000, quarter Kelly training, and uncertainty gates. Daily bars were used, but cannot validate 15-minute execution. If any necessary gate fails, there will be no allocation. The experiment has a maximum allowance of 15 minutes.

Before the market opens at 08UTC, the fund manager reviews the plan. Previously, there was no valid edge, so the expectation of staying cash was set aside. The decision was made based on the actual timing of the first news on September 8 and 9, and the executable pre-event window, ignoring the popularity of headlines. The plan has three levels: A, B, and C. Level A looks at early contamination, including clinical releases and regulatory actions. For example, it considers UPB's first release versus a conference, INBX's first release versus a call, and ROIV's early data. It also looks at TLX's early action, the FDA's meeting calendar on September 8 and 9, and amendments versus old deadlines. Level B examines the calendar, including earnings reports from BRZE, CASY, and GME on September 8, and CHWY and ASO on September 9. It also considers global ITX timing and venue, as well as macroeconomic events like the Census on September 8 and the NYFed's expectations. Level C focuses on tradable edges, including evidence from closed trades and execution strategies like fund 49.14 and whole shares. The plan also covers operations, including coverage, revision, and handoff at 12:30UTC. The binding constraint is the expected net event payoff, not the monthly shortfall. The initial discovery should take no more than 10 minutes, and the last 5 minutes of records will be reserved for reporting.

The Catalyst fund is doing research on IONS, MIRM, and GIS. The fund size is 0. The current time is 2026-09-08T07:45:29.185251+00:00. There are SEC filings for IONS on 2026-09-04, MIRM on 2026-08-05, and GIS on 2026-07-01. A news scan found the headline that Novartis' Pelacarsen Phase III Lp(a)HORIZON Trial failed its primary cardiovascular endpoint on 2026-09-04. General Mills is having a conference on September 8, 2026, and will release earnings on September 23, 2026. The fund will review the company's performance, specifically looking for improving organic volume or cash, or new guidance. If the review is positive, the fund will record evidence and consider independent cost analysis. If the review is weak or delayed, the fund will retain its cash. The baseline for GIS is a fiscal year 2027 EPS of 3.00-3.20, organic growth of -1.5 to +0.5%, and savings of 750 million. The fund has a frozen paper window from September 16 to September 21 and will not retune past tests. The fund has completed 3 filing checks, 1 headline lead, and 2 GIS reminders. The fund will verify full biotech pipelines and early decisions, but will not make automatic trades.

The automated trading fund manager reports that as of 06UTC, the cash balance is $49.14 with no orders placed. The final strategy records show rejections for LOVE and SFIX. For SFIX, a paper trade was simulated from September 16 open to September 22 close, and the results will be tracked without re-tuning and without a standing entry. For LOVE, the T5 window has passed, and there are 2 sessions remaining, but it is not possible to borrow the T5 edge. A review of LOVE on September 10 at 12:30UTC compared sales and EBITDA from $157 million to $166 million, with a range of -4 to +2 million in refunds, and $3.6 million in refunds. A review of SFIX on September 24 at 12:30UTC will compare revenue from $322 million to $327 million, EBITDA from $7 million to $10 million, client retention, and cash. Both new monitors are unarmed due to a 20% cap. The collector runs every 240 minutes, has been tested with 1423 bytes, and preserves 17 FDA checks. There is no existing research coverage that has been canceled without replacement. At 08UTC, the actual job delivery will be read, including jobs 48, 67, and 75. Note that the collector output is not guaranteed on the chosen review minute. The primary dates observed are 13294-13295, and the full historical corpus, variant, and simulation outputs have been recorded. Fundamental 1104 replaces 1098, and the LOVE FY guide EBITDA mid-margin is 5.625%. The primary discovery will continue only if new independent information is available, and positive holdouts will not be re-optimized.

A news scan found the reproduction manifest and explicit gaps. The system concatenates same filename chunk bodies with no separator. The full script is test_love.py. The SFIX script is identical except for bars SFIX, events sfix_events.json, and it skips only dates not in q.index with no advance notice gate, and it outputs all prefix sfix_. Event observations contain a complete LOVE/SFIX dated corpus. The market inputs are requests to get data from the site URL '/data/stock-bars' with parameters symbols LOVE/SFIX/SPY, timeframe 1 day, start date 2019-01-01, end date 2026-09-05, limit 10000, feeds ip, adjustment all, and auth Bearer DATA_TOKEN. The system received 1930 bars, and the next token is None. The raw hashes are: love_events.json is 19330a214d26119ffa4684b7333d6d9df6f63121e4ba343a3997475ebad97e4b, sfix_events.json is e709f261c39a1d72bdabf1cf2c1f7f6c1f57ca3f1c419a8a12d52d9868e4365b, LOVE_bars.json is 52b4a5f9cd077fa0c5906a0b627683f9b1b486b088802cea03fc4244633beb62, SFIX_bars.json is a5bee79d0997651142a17afd6fc6efd5a6e92362fb664235d076c4c7b09676d7, sfix_results.json is 0f7468143d266ea0bc76468ed7e42c4baac8f99fef93b5625b0bf2f9d06b27cf, SPY_bars.json is 4e06ff1abf6631e0d953ea0a463eae34e48515def67b5b0fcf5c3bcd3bdedf5e, and results.json is d7330d78728a258673fa476b601a9b4a6c0ad55c54703d37faec2be045d29876. Local research files persist in this workspace. The system recorded full statistics, including OLS regressions and bootstrap methods, and historical SFIX notice lead time, which is 15 minutes, execution time, spread/fees, and early news contamination, which is unresolved. No model is ready for trade. The collector 80 replaces 17 with the same FDA code plus 2 SEC 1 RSS, which is 240 minutes, and 96 runs equal 16 days. The local test is 1423 bytes and exits 0. The monitor 20 cap means new LOVE/SFIX dates are un armed. The system has periodic reminders, but they are not exact time execution. The system will hold cash until an independent validated edge is found.

A script ran with the following results: it checked the stocks SFIX and LOVE. It also looked at the news feed from LOVE at the website https://investor.lovesac.com/news-events/press-releases/rss. The most recent news item was published on a date and had a title, but the details are not specified. The script then printed some reminders about upcoming events for LOVE and SFIX, including a call on September 10 at 08:30 ET and a call on September 24 after the market close at 17:00 ET. The script also mentioned some rules for making decisions, including improving organic growth and cash, and requiring independently validated net event edges. It noted that there are no automatic orders and that the information is not exhaustive.

A news scan found the script output of a python code. The code is doing two separate tasks. The first task is to scrape a website for links related to a meeting. The website is https://www.fda.gov/advisory-committees/advisory-committee-calendar/september-23-2026-molecular-and-clinical-genetics-panel-medical-devices-advisory-committee-meeting. The code is looking for links with labels that contain certain keywords like briefing, meeting materials, presentation, and transcript. It then prints out some information, including the status code of the website request, whether a specific date is present in the text, the number of matching links, and a list of up to 8 links. The second task is to retrieve the latest 8-K filing for a given stock ticker from the SEC website using a node script. The code prints out the ticker, the date of the filing, and the URL of the filing. The code also handles errors and timeouts.

A news scan found the script output: a five-session path bootstrap using only training pre-event normalized OHLC, or open, high, low, close, blocks. The script created a list of blocks, where each block is a set of five consecutive sessions. It then selected 100,000 blocks at random and simulated a trading scenario. The simulation started with a price of 1.001 times the current close price, and a stop loss at 1 times the average true range below the current close price, and a target at 1.5 times the average true range above the current close price. The simulation ran for 5 sessions, and the results were as follows: the mean percentage return was the mean of the returns times 100, the win percentage was the percentage of returns greater than 0, the stop percentage was the percentage of returns that hit the stop loss, the target percentage was the percentage of returns that hit the target, and the time percentage was the percentage of returns that ended at the end of the 5 sessions. The script also calculated the 5th and 95th percentiles of the returns, and the net risk reward ratio. The whole shares were calculated as 49.14 divided by the current close price, and the full fund gap 30 was calculated as 30 percent of 49.14. The results were written to a file called sims.json.

A news scan found the script output: the code is printing the results of a trading simulation. It first prints the base results in json format. The code then calculates the interactions between different indicators, such as trend and momentum, and stores them in a list called inter. This list is then written to a file called interplay.json. The code also calculates various technical indicators, such as the simple moving average and momentum, and stores them in a dictionary called tech. This dictionary is then written to a file called tech.json and printed. The code then resamples the complete training event outcomes and simulates different fractions of the results. The simulations are stored in a list called sims.

A code snippet found the following trading logic: if an event is excluded or not released or the date is not in the index, the code continues to the next event. It calculates the location of the date in the index and checks if the notice date strictly precedes the entry date. If not, the event date is added to the exclusions list. The code then calculates the entry price, stop loss, and target price based on the average true range, which is a measure of volatility. It checks if the mode is set to barrier and iterates over the bars between the entry and exit points to check for gap stops, targets, and stops. The code updates the size and price arrays and calculates the net return. A news scan found the headline is not applicable in this context, as this is a code snippet. The code appends a dictionary to the results list with the trading metrics, including return percentage and drawdown. Finally, it writes the results to two JSON files: research/results.json and research/trades.json.

A Python script is being run at 06UTC. The script imports necessary libraries, including json, numpy, pandas, and vectorbt. It sets a random number generator with a seed of 2026090806. The script defines a function called bars that loads data from a json file and converts it into a pandas DataFrame. The function is used to load data for the stocks LOVE and SPY. The script calculates various technical indicators, including the average true range, simple moving average, momentum, and volatility. It also loads events from a json file. The script defines a function called stats that calculates various statistics, including the win rate, average win, average loss, and mean return. The script then runs a series of tests with different parameters, including the hours parameter, mode, slip, and split. For each test, it calculates the results and stores them in a list. The script uses a list of events to exclude certain dates from the tests.

The fund manager of the Catalyst fund, which is for research only, made some primary checks. The fund is not eligible for approval. The manager compared links with prior notes. On September 8, 2026, at 06:16:02 UTC, the manager checked the FDA website, which is a government website that provides information on food and drug administration, and found no matching links. The manager also checked the SEC filings, which are documents that companies submit to the Securities and Exchange Commission, for SFIX on July 28, 2026, and LOVE on June 15, 2026. Additionally, the manager checked the RSS feed, which is a way to receive updates from a website, for LOVE and found an announcement about a conference call on August 27, 2026. The manager chose to review LOVE on September 10 at 12:30 UTC and SFIX on September 24 at 12:30 UTC. The manager will monitor the companies' growth and cash flow, and will require independent validation of any net events. The manager will not make any automatic orders. The checks are not exhaustive, meaning they are not complete or final.

A test of the collector script was done on September 8 at 06:12 UTC and it finished successfully in 2 seconds, using 1 RSS feed and processing about 1600 bytes of data. The script uses command line interface to get filings from Edgar for the tickers SFIX and LOVE, specifically form 8-K and only one filing at a time. It also sends a request to the Lovesac investor website to get the latest press release, including the date, title, and link of the release. The script saves the data in chunks. Two new monitor requests were rejected due to a limit of 20 armed caps, and no old monitors were stopped to make room. The LOVE review was chosen for September 10 at 12:30 UTC, which is 8:30 ET, and the SFIX review was chosen for September 24 at 12:30 UTC. The script checks for positive organic cash, which means cash that is not from borrowing, and if it is not found, it re-verifies the cash. The existing collector, which is 75, was verified on delivery at 04:15 UTC and finished successfully, including the tickers EGTX and MFN, which are not just registered, but actually have data.

A news scan found the following text: Prior to this, there was no validated edge, so affordability and recovery pressure were set aside. The process started with $49.14 in cash, with the goal of achieving repeatable pre-event base hits, considering the binding constraint of independent net event expectancy. A three-level question tree was used: 1. Event integrity, which included 1.1 fresh earnings beyond existing rejected names, looking at the exact issuer date, earliest release versus call, and amendments, and 1.2 history, examining original earnings sentences, prior notice lead time, and exclusions or preannouncements. 2. Quant edge, which included 2.1 temporal holdout, training through 2023 and holding out 2024 onward with no retuning, 2.2 rule, looking at T5 open to T1 close before release day with fees of 10 basis points plus slippage of 10 basis points each side, 2.3 interplay, considering prior SMA50 and 20-day momentum, SPY regime, volatility, and volume, and 2.4 robustness, looking at T3, T5, and T10 and slippage of 10 and 25 basis points, reporting all and selecting no winner. 3. Feasibility, which included 3.1 price and volume, looking at whole share cap of $49.14, spread, and freshness, 3.2 exits, considering prior ATR14 stop of 1 ATR and target of 1.5 ATR, conservative daily barriers, and 15-minute uncertainty, and 3.3 sizing, using bootstrap event returns and exit odds, and quarter Kelly only if train and holdout were robust. 4. Outer structure, which included 4.1 issuer economics, looking at cash, debt, margins, and who benefits, 4.2 anticipation, considering preannouncement, seasonality, and selection bias, and 4.3 consequences, examining earnings surprise to operating cash to financing and valuation. The resolution was that daily sessions were sufficient to reject, but insufficient to certify 15-minute fills, and no media consensus was acted on. A fresh-name search was chosen before observing price results, and missed whole-market candidates were explicitly not exhaustive.

The fund is called Catalyst Research and its size is 0. The time is 2026-09-08T05:00:37.172407+00:00. A news scan found the headline 'Telix Completes Enrollment in Phase 3 BiPASS' on September 1, 2026, which is about a company called Telix completing a study, a type of research project. Another news scan found the headline 'Mission Produce to Release Fiscal Third Quarter' on August 25, 2026, which is about a company called Mission Produce releasing financial information. A third news scan found the headline 'ABM to Announce Third Quarter 2026 Financial' on August 25, 2026, which is about a company called ABM releasing financial information. A fourth news scan found the headline 'Roivant to Present Topline Results from the P' on September 6, 2026, which is about a company called Roivant presenting results from a study. The SEC, which stands for Securities and Exchange Commission, a government agency, has a document about UNFI from July 9, 2026. The company Roivant will present results on September 8, and a review will be done at 10:30 UTC. The results are about a study called PHocus, which has a primary goal and a secondary goal, and will compare certain things like safety. There are also earnings, which are financial reports, from companies ABM, UNFI, and AVO on September 8 and 9. There are also macro events, which are big events that can affect the economy, on September 8 and 9, including reports from agencies, which are organizations that provide information. All branches, which are parts of a decision-making process, will follow certain rules, including requiring independent information and not making orders. The information comes from 4 RSS, which stands for Really Simple Syndication, a way of getting news, and 1 SEC, which is a government agency.

The fund, called Catalyst Research, made some notes on September 8, 2026, at 4:15:26 UTC. It looked at some documents from the US Securities and Exchange Commission, or SEC, for two companies: HITI on August 31, 2026, and BFRI on August 13, 2026. The fund also noted that Dave & Buster's Entertainment, Inc., which is traded as PLAY, will report its second-quarter results on September 1, 2026. The fund has a list of upcoming events, including reviews of companies like TSM on September 10 and NKE on October 2. It also has a list of new events, including a review of CAN on September 8 and EGTX on September 25. The fund uses a strategy that involves recording and verifying certain events, like changes in cash or clinical evidence, and staying in cash if there are adverse or delayed events. It does not make automatic orders. The fund made a decision about PLAY, rejecting a certain option, and may make an early decision about EGTX. The fund also tracks news and reviews from various sources, including a website called MFN, which has information about EGTX. The fund will review EGTX on certain dates, including September 11 and 25, and will record and validate certain events, like new labels or cash flows.

The complete diagnostic script is divided into three parts, from lines 927 to 929, and the body strings are concatenated without any separator. The collector script, from lines 930 to 931, is similarly handled. The inputs include a request to the Yahoo finance website for EGTX.ST data over a 5-year range with daily intervals, as well as a saved research file named bars.json, which was downloaded on September 8 at 04:02 UTC. The SHA256 hash values for the raw data, test, and results are 8b6cbe497d6638874a7d469d557d3721535f38bba392e8ab9e43117f081edf0c, 89fc30b83366a134bc9ec9ab9fb0bf4a6b7f9e42cd300e850c20afea3051ccca, and b95bb427c6e6a475dc6fd301f0a0d90a69e79595e913aab6f835e2c0f908f73b, respectively. Future data pulls do not need to match the range or vendor revisions. There are 1256 complete bars ending on September 4, and the metadata has a value of 7.65 on September 7 without a daily row. The historical sample is chosen from a single issuer and is not representative of the whole market, and there is no historical sample of the first FDA event. Repeated trades are not independent, and the bootstrap confidence interval does not guarantee coverage because the data is not independently and identically distributed, which is a statistical concept meaning that each data point is independent of the others and follows the same distribution. Tests skip the last H rows in each split to avoid unfinished trades, and there is no holdout selection. Monetary amounts are in Swedish Krona unless explicitly stated in US Dollars, and no foreign exchange conversions are made. No API keys are printed.

A news scan found the script output: d. An early decision on EGTX is possible and not limited by a stop. The process ended after 2 seconds and 1 RSS, and 10 static reviews were considered, but this is not an exhaustive list and the monitors are not armed. A Python script was used to scan the website https://mfn.se/a/egetis-therapeutics and found links, then printed the count and hash of these links, and the latest link. The script also printed that EGTX chosen reviews are from September 11 at 13:50 UTC to September 16 at 19:40 UTC conferences, and from September 25 at 12:30 before the September 28 FDA decision. The script stated that a new label, PRV, or cash will lead to recording and validation, while adverse effects or delays will lead to cash and re-verification, with a 50% net PRV and edge 0, and static reminders that are not armed.

A news scan found the script output: python3 command to run a program. The program imports various libraries, including subprocess, json, and datetime. It prints the current time in UTC. The program defines a function called sec that takes a stock ticker symbol as input and tries to run a node command to get the latest SEC filing for that symbol. If successful, it returns the symbol, the date of the filing, and the URL of the filing. If there is an error, it returns the symbol and an error message. Another function called play tries to get the latest news release from the Dave and Buster's website and returns the date, title, and link of the release. If there is an error, it returns an error message. The program then runs these functions in parallel for the stock symbols HITI and BFRI, and for the play function. It prints the results of these functions. The program then prints a list of static chosen review dates and times for various stocks, including TSM, DATA.L, ET.TO, GIS, and NKE. It also prints a list of new review dates and times for various stocks, including CAN, PLAY, ABVX, EGTX.ST, and 4431.T. The program then prints a message about the strategy for all branches, which includes requiring improved organic cash and credible clinical evidence, and not making automatic orders. It also mentions a specific stock, PLAY, with a return of -13.15% and a holdout return of 6.88% with a confidence interval of -4.48% to 7.25%.

A news scan found the script output: on.dump(tech,open('research/tech.json','w'));print('TECH',json.dumps(tech)). This is a data line. A vector, or a series of numbers, called vec was calculated using data up to '2024-12-31' for 'open', 'high', 'low', and 'close' prices, then divided by a shifted constant, or a number moved by one position, and converted to a numpy array, which is a way to handle numbers in a computer program. The number of simulations, N, was set to 100000 and an array of prices, px, was created with the last price in the constant, c. The start price was set to the first price in px, the stop price was set to the start price minus 14-day Average True Range, or ATR14, which is a measure of price movement, and the target price was set to the start price plus 1.5 times ATR14. An array to track why a simulation stopped, why, was created with all values set to 3. The simulation ran 10 times. In each run, new open, high, low, and close prices were calculated by multiplying px with vec and then the prices were updated based on whether they were active and if they reached the stop or target price. The why array was updated accordingly. After the simulation, the return, rr, was calculated as the final price times 0.9975 times 0.999 divided by the start price times 1.0025 times 1.001 minus 1. The results were stored in a dictionary, sim, which included the number of paths, training days, mean percentage return, win rate percentage, 5th percentile return, 95th percentile return, stop percentage, target percentage, time percentage, stop price in Swedish Krona, target price in Swedish Krona, and net return ratio. The sim dictionary was then written to a file called'research/sim.json' and printed.

A code snippet was found. It appears to be a part of an automated trading system. The system uses a library called vbt, which stands for vectorized backtesting. The code defines a portfolio with a given set of parameters, including initial cash and fees. It then calculates the total return and drawdown of the portfolio. The results are stored in a dictionary called out, which contains various metrics such as return percentage, drawdown percentage, and win rate percentage. The code also uses a function called test, which takes in parameters such as moving average, hold period, slippage, and split. The function returns a dictionary containing the results of the test. The code then runs a series of tests with different parameters and stores the results in a list called results. The results are then written to a file called backtests.json. A news scan found the headline 'BASE' with the following data: a list of dictionaries containing the results of the tests. A dictionary called tech was found, which contains technical indicators such as simple moving average, average true range, and volatility.

A news scan found the script output: import necessary libraries including json, numpy, pandas, and vectorbt. The script uses a random number generator with a seed of 20260908. It loads data from a file named 'bars.json' and creates a pandas DataFrame from the 'quote' indicator. The DataFrame is indexed by a datetime object converted to the Europe/Stockholm timezone. The script then drops any rows with missing values and removes any duplicate index values. It also loads the 'adjclose' indicator and adjusts the 'open', 'high', 'low', and 'close' prices using the adjusted close price ratio. The script calculates the average true range, or atr, which is a measure of volatility, and saves the data to a csv file named 'clean.csv'. The script then defines a function named test that takes four parameters: ma, hold, slip, and split. The function creates a signal based on the moving average and percentage change of the close price. It then splits the data into training and testing sets based on the'split' parameter. The function creates a new index with two slots per session, allowing for same-day exits. It then iterates over the data, checking for the signal and atr values. If the signal is true and the atr value is finite, it sets an entry price, stop loss, and target price. The function then checks for exits based on the stop loss, target price, or time, and updates the exit price and reason accordingly.

A news scan found the collector replacement manifest and verified previous delivery. The actual delivery of Job72 at 02:15:50UTC was verified, with an exit code of 0 and 1349 characters. An earlier registration issue is now resolved. The Active17 schedule is set for 02:30:50 on September 9, and the 48 schedule is set for 03:30:27 on September 7, and the 52 schedule is set for 04:45:28, and the 67 schedule is set for 08:15:24, all in UTC time, but scheduling may drift. A new local collector, research/collector.sh, was tested at 04:06:56 and exited with code 0, with 1720 bytes, preserving the existing HITI, BFRI, SEC, and PLAYRSS, as well as 10 static reviews, and adding a link to MFN Egetis with hash 85e339a790c3 and the latest conference on September 4. The Egetis HTML direct feed returned a 403 error, and the MFN issuer distributor works, but there is no independent corroboration. The new collector will replace the old one only after the tested script is recorded. The old collector runs every 480 minutes, which covers 24 days, until September 28. The full source and old static dates are preserved in the 02UTC collector 30 preservation manifest. New Egetis sources are obs12354 and https://mfn.se/a/egetis-therapeutics. Additional static reviews are set for September 11 at 13:50UTC, September 16 at 19:40UTC, and September 25 at 12:30UTC, with review times chosen after conferences and before September 28, and are not based on news promises. A monitor attempt was rejected due to armed capacity, and it is noted that armed capacity should not be claimed. Existing monitors were not resolved prematurely. The system is flat, and no off-run exit is needed.